Lumentum’s Q1 provided more confirmation that EML laser shipments are ramping in full force, with another record quarter driven by 100G speeds and an increase in 200G shipments. EMLs have been the primary driver of growth so far for Lumentum, though the supply-demand imbalance is widening due to tight indium-phosphide (InP) capacity. Looking ahead to 2026, InP capacity will be a key factor to focus on as Lumentum is targeting 40% capacity growth over the next few quarters, with the potential for this to drive even stronger revenue growth.
Outside of EMLs, Lumentum is beginning to work on CW lasers for silicon photonics and co-packaged optics. CW laser shipments for 800G have begun with 1.6T eventually layering in to growth next year, regardless if CW lasers or EMLs are the preferred component of choice for 1.6T rates. Management also remains confident in other growth opportunities in co-packaged optics (CPO) and optical circuit switches, though the latter is expected to be a late calendar 2026 story.
On the financials side, the number one item was Q2’s impressive 22% QoQ revenue growth guide to $650 million at midpoint. This is significant as Lumentum is reaching its $600 million quarterly revenue target two quarters ahead of schedule, with this also marking its highest revenue in company history. The 22% QoQ guide would also reflect Lumentum’s fastest sequential growth since the September 2020 quarter.
Electro-absorption modulated lasers (EMLs) have quickly become attractive for AI servers as these components help enable 100G and 200G per lane transmissions, thus enabling 800G and 1.6T data rates for optical transceivers. EMLs also leverage indium phosphide (InP) over silicon as InP reduces power consumption, although it is more expensive at the component level as four EMLs are needed compared to two lower-cost CW lasers for silicon photonics modules.
EMLs are a critical component with Nvidia’s Blackwell generation, as the scale-up in GPU counts per rack from eight to 72 and subsequent increases in bandwidth and switch density will require low-power, efficient high-speed optics. The power advantages over SiPho also come to the forefront as power consumption becomes a central concern in scaling AI data centers, with Blackwell doubling power consumption versus Hopper at 140kW per rack.
EMLs are the main driver for Lumentum’s growth as these are good for short-to-medium reach and a strong choice for 400G and 800G optical transceivers, with the company having begun its 100G EML ramp for these data rates in early 2024. EML laser shipments reached a fresh record in fiscal Q1 2026, driven once again by 100G speeds and an increase in 200G shipments.
More importantly, Lumentum expects calendar 2026 to be another breakout year for laser chip shipments, anchored by a widening supply-demand imbalance, sharp capacity growth and mix shift to higher priced, higher margin 200G products.
One important discussion on EMLs is that the supply-demand imbalance continues to widen, meaning that substantial growth in capacity through 2026 should quickly convert to revenue. CEO Michael Hurlston explained that “last quarter, I think we characterized it as roughly a 20% shortfall relative to total customer demand. Even with the add in supply, I would say that number has increased to 25% to 30%. We are quite a bit short right now relative to the customer demand.”
Lumentum is not the only supplier commenting about this imbalance, with Applied Optoelectronics also echoing this in their Q3 earnings call; however, management hinted that despite industry-wide capacity increases, supply could still lag demand through 2027: “We've also said we see the supply and demand imbalance increasing we're falling further behind. And that accounts for all this other capacity that's being built out here or there and everywhere by our competition. So at least through 2027, we don't believe we catch up. We think we're still behind on supply.”
On the positive side, Lumentum shared that while its indium phosphide fab is fully allocated due to high demand, it has made “better-than-expected progress on yields and throughput and now see a line of sight to add approximately 40% more unit capacity over the next few quarters.” CEO Michael Hurlston clarified at UBS’ tech conference that “we gave in the last earnings call a new benchmark saying, over the next 3 quarters, meaning our December, March and June quarters, we expected to add that 40%. So that's a forward-looking statement where we'd expect an increase in capacity of 40% on what already is a doubled number.”
Breaking this down suggests that the yield and throughput improvements means Lumentum is exceeding linear capacity growth, which can translate to stronger than expected revenue from more capacity going to higher ASP products. It also has strong implications for Lumentum’s margins and EPS, driving strong expansion in operating margins that then flows through to EPS:
“So that 40% increase in indium phosphide capacity is focused on laser chips, which has, as you know, higher gross margins than many more of our other product lines. So as that flows through in the coming quarters, that will have a positive effect on our earnings per share. What you're seeing this quarter is without that increased capacity and increase gross margin contribution from the indium phosphide capacity we talked about.”
Lumentum is also now working on CW lasers for silicon photonics (SiPho) and co-packaged optics (CPO), which are expected to kick in with 1.6T transceivers and layer into topline growth even if CW takes share from EMLs at 1.6T.
Management expects to be well positioned for both EML and CW lasers ramping for 1.6T transceivers, as its capacity is interchangeable between the two components, despite management noting a difficulty in forecasting how the two will ramp – the primary takeaway here is that even if faster data rates such as 1.6T are less dependent on EMLs, management believes there is more than enough content for them to do well:
“On the battle between CW and EML, it appears to us that CW is going to ramp with 1.6T but so will EML. And so the slope of the 2 ramps was hard for us to call but it looks like no matter how you slice it, the numbers will increase. So even if the mix shifts away from EML-based transceivers at 1.6T, the absolute numbers seem to be stratospherically high. And at least in the near term, we see no end in sight. We watch it every day, Chris, just like you're sort of cautioning but I think for the next 6 quarters, we're completely sold out, and we have long-term agreements, as I said, that we've worked out with our customers to ensure that they're going to take any additional capacity we've got online.”
For a bit more on CW lasers, its 70 mW lasers started meaningful shipments this quarter and will be a more reasonable part of the mix in the December quarter, while sampling for 100 mW CW lasers just began. 100 mW lasers are expected to be in full production by mid-year 2026, with Lumentum aiming to integrate these into its own internal transceivers, slated for the June 2026 quarter.
Q2 Outlook of $650M, Two Quarters Ahead of $600M Target
While Q1 produced a solid beat, the most impressive part of the report was Q2’s guidance, with the company forecasting revenue of $630 million to $670 million. This marks a sharp sequential acceleration of nearly 11 points to 21.8% QoQ growth at midpoint and 25.5% QoQ at the high-end of guidance.
On a YoY basis, the midpoint of the guidance points to a more than 3 point acceleration to 61.6% YoY, while the high-end would reflect 66.6% YoY growth.
Just last quarter, Lumentum had projected reaching $600 million in quarterly revenue by the June 2026 quarter (fiscal Q4) or earlier, with the company now two quarters ahead of that target. When looking at the company’s original guidance for the end of 2025, which was $500 million (and satisfied by Q1), Q2’s forecast is 30% ahead of that, reflecting the strength of the AI networking theme and the demand the company is seeing.
For the strong QoQ guide, management said that “the thing that probably caught us flat-footed is the width of the customer demand. It's touching everything. We talked about pump lasers. We talked about narrow linewidth. We talked about the transceivers. We talked about even coherent components. So it is very, very broad-based. And every single one of our segments is up. Every single one of our segments is contributing to the growth that you see.”
To put in perspective how strong Lumentum’s growth curve is, current estimates for the June 2026 quarter sit at $740.3 million, more than 23% ahead of the company’s target revenue. This is also up from $689.9 million on November 7, a 7.3% revision higher in less than one week.
Out of Lumentum’s three outlined growth drivers through 2026 – cloud transceivers, optical circuit switches (OCS) and co-packaged optics (CPO) – only cloud transceivers are expected to meaningfully contribute to Q2’s growth. OCS and CPO are expected to see much stronger growth next year, with ultra-high power lasers for CPO more geared towards 2H 2026. More on this is discussed below.
Lumentum Intentionally Keeping Customer Count Low
Another important discussion circled back to supply allocation and possible customer consolidation. This is not something that is necessarily new to Lumentum, as we had covered in our previous analysis that the company is intentionally keeping customer count low and not taking on new customers in an effort to focus on the highest-margin opportunities.
Analysts had asked if management would use EML supply constraints to drive new transceiver engagements and qualifications and expand the customer base. However, management countered this and said they are actually trying to “consolidate supply and consolidate our customer base around a couple of folks that we think are going to be long-term winners. Those customers in return have given us multiyear commitments that give us a lot of confidence that our business is going to be sustainable even as we continue to ramp capacity through the next probably 6 or 8 quarters.”
Management made sure to emphasize again that they will aim to “allocate our laser capacity based on the profitability metric more than to trying to broaden our transceiver opportunities in 1.6T using our lasers.”
This is a two-edged sword, as multi-year commitments give Lumentum security in the ramp phase with visible, long-term revenue growth, yet it also could increase customer concentration risk by tying Lumentum solely to handful of key customers and limit its opportunities to diversify its customer base. This concentration risk is already becoming a bit more evident, with two customers accounting for 45% of revenue, at 22% and 21% respectively in fiscal Q1. This is up from 31.4% of revenue in fiscal 2025, at 16% and 15.4% respectively for the two largest customers.
Cloud Transceiver Ramp Expected to Begin Next Quarter
Lumentum’s ramp for cloud transceivers is expected to begin next quarter, with management stating that they have a line of sight to transceivers eventually becoming a $250 million/quarter business, or a $1 billion annual run rate; this is double its $500 million annual run rate today. Lumentum does not plan to expand the business beyond that $1 billion run rate, stemming from its gross margin profile.
Cloud transceiver revenue was roughly flat QoQ in Q1 with Lumentum focusing primarily on increasing manufacturing capacity in Thailand to meet rising demand. As a result, management expects to resume growth in Q2 with the upward trajectory accelerating for the next four to five quarters.
Lumentum believes Q2 will serve as a ‘proof point’ that as its new 1.6T and 800G transceivers ramp, it will see “the revenue layering benefits that our larger transceiver competitors have experienced” around the middle of 2026. The ramp of 1.6T will be important to track, as Lumentum has been straightforward about 1.6T margins being “significantly better” than 800G.
However, CEO Michael Hurlston made clear at UBS’ tech conference that Lumentum is “operating meaningfully below the mid-30s in terms of margin” for transceivers as manufacturing is “substandard” on throughput, scrap and yields. He added that there is a path to get to the mid-30s over the next few quarters as production ramps and Lumentum in-sources components (versus virtually zero in-sourced today), but this remains a headwind to the company’s target model of 42%.
Because of the lower-than-target margins, Hurlston explained that while Lumentum has “aspirations to get it to $1 billion annually, to add another $500 million of incremental revenue. But we don't want it to run much higher than that, just given the margin headwinds we see. We think we can manage our business up from a margin perspective if we keep the business to about $1 billion top line. If it gets beyond that, it will be more challenging.”
No Change to Co-packaged Optics Timing, But Demand is Stronger
As we discussed in September for Discovery members, co-packaged optics (CPO) is not contributing to revenue now yet could materialize into one of the biggest opportunities among all of the components and subsystems that Lumentum supplies, as the company says it is enabling Nvidia’s Spectrum-X networking switches. As a reminder, CPO places optical transceivers directly on the chip package, rather than using separate optical modules, resulting in faster data transmission, reduced latency and higher bandwidth. This may be the best of both worlds: the performance of optical yet with reduced power consumption for increasingly power-hungry AI racks.
Management provided a brief update on CPO, noting that the ramp is forecast to begin in the early stages of calendar Q3 2026 with a more meaningful contribution in calendar Q4. Hurlston explained in Q1’s call that the only change is that “demand is stronger than we initially forecast” and “getting better,” though the timing for the ramp is still the same. He clarified further that Lumentum expects “an inflection point on Ethernet-based switches. That's where we see the real step-up where our revenue would become more material” in the second half of 2026, continuing through 2027. Second-gen CPO products for 3.2T speeds are tentatively on deck for 2028.
Ultra-high power lasers are still in the initial production ramp, though Lumentum expects significant growth in shipment volumes in 2H 2026 with accelerating adoption, with this providing further confirmation of the strength of the CPO opportunity. Lumentum had announced the production expansion in early August, giving the company multiple quarters to ramp.
Optical Circuit Switching Also a Late 2026 Story
Lumentum’s second upcoming growth driver, optical circuit switches, are not expected to meaningfully contribute in Q2, rather being a late 2026 story alongside CPO. Optical switches are a new kind of switch for AI clusters that handle the switching optically instead of using transceivers to convert photons to electrons, and back again. Optical switching and CPOs work together to allow for more flexibility for reconfigurations, to reduce energy and complexity while also increasing bandwidth.
Management has outlined confidence in reaching a $100 million quarterly revenue target by the December 2026 quarter, with its two major customers expected to be qualified in the March 2026 quarter with a third customer potentially qualifying in the middle of the year. CEO Michael Hurlston provided more clarity about how Lumentum expects OCS to ramp beginning in this quarter through 2026:
“We outlined sort of a revenue ramp of kind of mid-single-digit millions here in the December quarter, getting to double digit — very, very low double digits in the March quarter and then accelerating to kind of mid $50 million, $60 million in the middle of the year and then getting all the way to that $100 million mark in the December quarter.”
This commentary implies that the largest ramp and impact from OCS will hit in fiscal Q2 2027, with management eyeing tens of millions of QoQ growth in the back half of next year. As seen in the revisions, current estimates only point to $59 million QoQ growth in that quarter, which may underestimate the tailwinds from simultaneous growth in OCS, transceivers and initial CPO growth in the second half of 2026.
Financials
Revenue Growth Maintaining >50% YoY
Lumentum fulfilled its guidance for a >$500 million revenue quarter in calendar 2025, reporting a record $533.8 million in revenue in fiscal Q1, beating estimates by just 1.4%. Revenue growth accelerated 2.5 points to 58.4% YoY though QoQ growth slowed to 11%.
As discussed previously, Lumentum guided for $630 to $670 million in revenue in Q2, accelerating to 61.6% YoY and 21.8% QoQ, whereas consensus estimates were pegged at almost 40% growth to $561.5 million.
Looking ahead, growth is expected to stay strong in Q3 at nearly 61% YoY, but the more impressive number is Q4’s estimated 54% growth, as this comes against a much more difficult comp of 55.9% vs 16.0% for Q3. This underscores the strength of the demand ramp Lumentum is discussing for the back half of calendar 2026.
On an annual view, Lumentum is estimated to report 57.3% growth in fiscal 2026 to $2.59 billion, before slowing to 29.6% YoY to $3.36 billion in fiscal 2027. Revisions are much stronger in fiscal 2027, up $600 million since the start of October versus a $300 million increase for fiscal 2026.
AI Revenue
Lumentum estimates that over 60% of total revenue comes from cloud and AI infrastructure customers, or above $320 million in Q1.
Lumentum changed its reportable segments in Q1, dropping Cloud & Networking and Industrial Tech and instead transitioning to Components and Systems. Components include laser chips, laser subassemblies, line subsystems and wavelength management subsystems, while Systems includes full stand-alone products such as optical transceivers, optical circuit switches and industrial lasers.
Components revenue rose 18.4% QoQ and 63.9% YoY to $379.2 million, fueled by “robust demand inside the data center”, strong momentum for DCI products with narrow linewidth laser assemblies for DCI transmission up 70% YoY, and record EML shipments. Lumentum expects Components to be the cornerstone for revenue growth and profitability while Systems will scale rapidly with transceivers, OCS and other high-performance solutions.
Systems revenue declined (3.6%) QoQ but increased 46.5% YoY to $154.6 million. Cloud transceiver revenue was approximately flat QoQ as Lumentum worked to increase capacity.
For Q2, Lumentum expects approximately half of its sequential revenue growth (or ~$60 million at midpoint) to come from Components, and the other half from Systems, “primarily reflecting the ramp of high-speed optical transceivers for data center applications and to a lesser extent, the early phase of our optical circuit switch ramp.”
GAAP EPS Back to Positive
Lumentum reported a razor thin $0.05 in GAAP EPS, while adjusted EPS of $1.10, up 511% YoY, beating estimates by 6.8%. For Q2, Lumentum guided for adjusted EPS in a wider range of $1.30 to $1.50, up 233% YoY, coming in well ahead of the $1.16 estimate at the midpoint. Fiscal Q3 and Q4 are expected to see adjusted EPS continue to increase, though YoY growth technically is decelerating to 162% in Q3 and 91% in Q4 as comps get more difficult.
Lumentum did not provide a full year adjusted EPS guide, though consensus now sits at $5.63, up from $4.90 and pointing to growth of 173% YoY. Considering Q2’s estimate remains below the midpoint of management’s guidance at $1.38, there is room for upside revisions if Lumentum provides another beat and raise next quarter.
Margins Show Strong Expansion
Gross margin continued to expand both sequentially and YoY, helping drive GAAP operating margin back to positive territory.
GAAP gross margin was 34.0%, in Q1, up nearly 11 points YoY and 0.7 points QoQ. Adjusted gross margin was 39.4%, up 6.6 points YoY and 1.6 points QoQ.
GAAP operating margin was 1.3%, up nearly 26 points YoY and 3 points QoQ. Adjusted operating margin was 18.7%, up 15.7 points YoY and 3.7 points QoQ, ahead of guidance for 16-17.5%. For Q2, management guided for continued adjusted operating margin expansion to 20-22%.
GAAP net margin was 0.8%, up 25.3 points YoY and not comparable QoQ due to an income tax benefit in Q4. Adjusted net margin was 16.2%, up 12.6 points YoY and 3 points QoQ.
Management provided a deeper discussion on margins moving through 2026, with product pricing from supply-demand imbalances serving as a strong lever for margin expansion:
“I think we're moving the margin line up. Pricing, obviously, is a lever. And when you look at that very, very carefully, I think what you see in the guide is some pricing, very targeted price increases happening. I think as you look out next year in 2026, our agreements with customers will include more pricing, more broad-based price increases, just given the supply-demand imbalance.”
CFO Wajid Ali added that margins are benefitting from improved manufacturing utilization, and moving into calendar 2026, gross margins are expected to move up in line with the company’s model from OFC (shown below) as OCS, 1.6T transceivers, and CPO ramp.
Lumentum is currently tracking closer towards the $750 million model by mid-2026, which is expected to see adjusted operating margin above 20% and gross margin approaching 40%. Lumentum is currently ahead of targets for adjusted operating margin per Q2’s guide, which suggests that there could be further upside as higher-margin product ramps, or some stagnation from the initial ramp phases for OCS and CPO to remain within the target ranges.
Cash Flows Muted
Cash flows were rather muted, with operating cash flow margin shrinking both YoY and QoQ.
Operating cash flow was $57.9 million in Q1 for a 10.8% margin, down from 11.8% a year ago and 13.3% in Q4. Free cash flow was ($18.3 million) for a (3.4%) margin, up from (10.2%) a year ago but down from 2.1% in Q4.
Cash and equivalents were $1.12 billion while debt was $3.24 billion.
Inventories for $531.6 million, up more than 13% QoQ, while accounts receivable surged nearly 23% QoQ to $307 million, both aligning with management’s commentary for strong product and revenue ramps over the coming quarters.
Valuation
Lumentum is trading at a stretched 6.9x forward PS multiple, more than double its five year average of 2.9x and above the 4.2x level that shares failed to break past in late 2024 and early 2025.
On the bottom line, however, Lumentum is trading just above its average multiple, currently valued at 45x forward adjusted EPS versus its five year average of 40x. Shares have traded as high as 60x and as low as 18-20x.
Conclusion
A lack of InP capacity is causing a rather substantial shortfall in EML laser supply as demand continues to expand, with Lumentum one of two companies able to meet this demand. Evidence of this tight supply is seen in Lumentum’s QoQ acceleration from 11% in Q1 to 22% guided in Q2, as Lumentum was able to increase InP capacity by 40% a few quarters ago. Looking ahead, an additional 40% capacity growth is coming online over the next few quarters, and higher yields and throughput on the upcoming capacity expansion could translate into a higher revenue growth rate.
Outside of EMLs, Lumentum is beginning to work on CW lasers for silicon photonics and co-packaged optics, which is expected to serve as the company’s next catalyst moving into 2026. This catalyst will hinge on whether Lumentum sees similar qualifications for SiPho and CPO as it has for EMLs for Nvidia’s Blackwell, or if it fails to be chosen as a lead supplier for CW moving through 2026.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Take advantage of the I/O Fund’s largest sale of the year with up to $250 off Advanced Market Signals, which offers Knox’s weekly webinars, real-time trade alerts, and in-depth analysis on the most powerful tech trends. Learn more here.Learn more here.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Lumentum’s Q1 provided more confirmation that EML laser shipments are ramping in full force, with another record quarter driven by 100G speeds and an increase in 200G shipments. EMLs have been the primary driver of growth so far for Lumentum, though the supply-demand imbalance is widening due to tight indium-phosphide (InP) capacity. Looking ahead to 2026, InP capacity will be a key factor to focus on as Lumentum is targeting 40% capacity growth over the next few quarters, with the potential for this to drive even stronger revenue growth.
Outside of EMLs, Lumentum is beginning to work on CW lasers for silicon photonics and co-packaged optics. CW laser shipments for 800G have begun with 1.6T eventually layering in to growth next year, regardless if CW lasers or EMLs are the preferred component of choice for 1.6T rates. Management also remains confident in other growth opportunities in co-packaged optics (CPO) and optical circuit switches, though the latter is expected to be a late calendar 2026 story.
On the financials side, the number one item was Q2’s impressive 22% QoQ revenue growth guide to $650 million at midpoint. This is significant as Lumentum is reaching its $600 million quarterly revenue target two quarters ahead of schedule, with this also marking its highest revenue in company history. The 22% QoQ guide would also reflect Lumentum’s fastest sequential growth since the September 2020 quarter.
Electro-absorption modulated lasers (EMLs) have quickly become attractive for AI servers as these components help enable 100G and 200G per lane transmissions, thus enabling 800G and 1.6T data rates for optical transceivers. EMLs also leverage indium phosphide (InP) over silicon as InP reduces power consumption, although it is more expensive at the component level as four EMLs are needed compared to two lower-cost CW lasers for silicon photonics modules.
EMLs are a critical component with Nvidia’s Blackwell generation, as the scale-up in GPU counts per rack from eight to 72 and subsequent increases in bandwidth and switch density will require low-power, efficient high-speed optics. The power advantages over SiPho also come to the forefront as power consumption becomes a central concern in scaling AI data centers, with Blackwell doubling power consumption versus Hopper at 140kW per rack.
EMLs are the main driver for Lumentum’s growth as these are good for short-to-medium reach and a strong choice for 400G and 800G optical transceivers, with the company having begun its 100G EML ramp for these data rates in early 2024. EML laser shipments reached a fresh record in fiscal Q1 2026, driven once again by 100G speeds and an increase in 200G shipments.
More importantly, Lumentum expects calendar 2026 to be another breakout year for laser chip shipments, anchored by a widening supply-demand imbalance, sharp capacity growth and mix shift to higher priced, higher margin 200G products.
One important discussion on EMLs is that the supply-demand imbalance continues to widen, meaning that substantial growth in capacity through 2026 should quickly convert to revenue. CEO Michael Hurlston explained that “last quarter, I think we characterized it as roughly a 20% shortfall relative to total customer demand. Even with the add in supply, I would say that number has increased to 25% to 30%. We are quite a bit short right now relative to the customer demand.”
Lumentum is not the only supplier commenting about this imbalance, with Applied Optoelectronics also echoing this in their Q3 earnings call; however, management hinted that despite industry-wide capacity increases, supply could still lag demand through 2027: “We've also said we see the supply and demand imbalance increasing we're falling further behind. And that accounts for all this other capacity that's being built out here or there and everywhere by our competition. So at least through 2027, we don't believe we catch up. We think we're still behind on supply.”
On the positive side, Lumentum shared that while its indium phosphide fab is fully allocated due to high demand, it has made “better-than-expected progress on yields and throughput and now see a line of sight to add approximately 40% more unit capacity over the next few quarters.” CEO Michael Hurlston clarified at UBS’ tech conference that “we gave in the last earnings call a new benchmark saying, over the next 3 quarters, meaning our December, March and June quarters, we expected to add that 40%. So that's a forward-looking statement where we'd expect an increase in capacity of 40% on what already is a doubled number.”
Breaking this down suggests that the yield and throughput improvements means Lumentum is exceeding linear capacity growth, which can translate to stronger than expected revenue from more capacity going to higher ASP products. It also has strong implications for Lumentum’s margins and EPS, driving strong expansion in operating margins that then flows through to EPS:
“So that 40% increase in indium phosphide capacity is focused on laser chips, which has, as you know, higher gross margins than many more of our other product lines. So as that flows through in the coming quarters, that will have a positive effect on our earnings per share. What you're seeing this quarter is without that increased capacity and increase gross margin contribution from the indium phosphide capacity we talked about.”
Lumentum is also now working on CW lasers for silicon photonics (SiPho) and co-packaged optics (CPO), which are expected to kick in with 1.6T transceivers and layer into topline growth even if CW takes share from EMLs at 1.6T.
Management expects to be well positioned for both EML and CW lasers ramping for 1.6T transceivers, as its capacity is interchangeable between the two components, despite management noting a difficulty in forecasting how the two will ramp – the primary takeaway here is that even if faster data rates such as 1.6T are less dependent on EMLs, management believes there is more than enough content for them to do well:
“On the battle between CW and EML, it appears to us that CW is going to ramp with 1.6T but so will EML. And so the slope of the 2 ramps was hard for us to call but it looks like no matter how you slice it, the numbers will increase. So even if the mix shifts away from EML-based transceivers at 1.6T, the absolute numbers seem to be stratospherically high. And at least in the near term, we see no end in sight. We watch it every day, Chris, just like you're sort of cautioning but I think for the next 6 quarters, we're completely sold out, and we have long-term agreements, as I said, that we've worked out with our customers to ensure that they're going to take any additional capacity we've got online.”
For a bit more on CW lasers, its 70 mW lasers started meaningful shipments this quarter and will be a more reasonable part of the mix in the December quarter, while sampling for 100 mW CW lasers just began. 100 mW lasers are expected to be in full production by mid-year 2026, with Lumentum aiming to integrate these into its own internal transceivers, slated for the June 2026 quarter.
Q2 Outlook of $650M, Two Quarters Ahead of $600M Target
While Q1 produced a solid beat, the most impressive part of the report was Q2’s guidance, with the company forecasting revenue of $630 million to $670 million. This marks a sharp sequential acceleration of nearly 11 points to 21.8% QoQ growth at midpoint and 25.5% QoQ at the high-end of guidance.
On a YoY basis, the midpoint of the guidance points to a more than 3 point acceleration to 61.6% YoY, while the high-end would reflect 66.6% YoY growth.
Just last quarter, Lumentum had projected reaching $600 million in quarterly revenue by the June 2026 quarter (fiscal Q4) or earlier, with the company now two quarters ahead of that target. When looking at the company’s original guidance for the end of 2025, which was $500 million (and satisfied by Q1), Q2’s forecast is 30% ahead of that, reflecting the strength of the AI networking theme and the demand the company is seeing.
For the strong QoQ guide, management said that “the thing that probably caught us flat-footed is the width of the customer demand. It's touching everything. We talked about pump lasers. We talked about narrow linewidth. We talked about the transceivers. We talked about even coherent components. So it is very, very broad-based. And every single one of our segments is up. Every single one of our segments is contributing to the growth that you see.”
To put in perspective how strong Lumentum’s growth curve is, current estimates for the June 2026 quarter sit at $740.3 million, more than 23% ahead of the company’s target revenue. This is also up from $689.9 million on November 7, a 7.3% revision higher in less than one week.
Out of Lumentum’s three outlined growth drivers through 2026 – cloud transceivers, optical circuit switches (OCS) and co-packaged optics (CPO) – only cloud transceivers are expected to meaningfully contribute to Q2’s growth. OCS and CPO are expected to see much stronger growth next year, with ultra-high power lasers for CPO more geared towards 2H 2026. More on this is discussed below.
Lumentum Intentionally Keeping Customer Count Low
Another important discussion circled back to supply allocation and possible customer consolidation. This is not something that is necessarily new to Lumentum, as we had covered in our previous analysis that the company is intentionally keeping customer count low and not taking on new customers in an effort to focus on the highest-margin opportunities.
Analysts had asked if management would use EML supply constraints to drive new transceiver engagements and qualifications and expand the customer base. However, management countered this and said they are actually trying to “consolidate supply and consolidate our customer base around a couple of folks that we think are going to be long-term winners. Those customers in return have given us multiyear commitments that give us a lot of confidence that our business is going to be sustainable even as we continue to ramp capacity through the next probably 6 or 8 quarters.”
Management made sure to emphasize again that they will aim to “allocate our laser capacity based on the profitability metric more than to trying to broaden our transceiver opportunities in 1.6T using our lasers.”
This is a two-edged sword, as multi-year commitments give Lumentum security in the ramp phase with visible, long-term revenue growth, yet it also could increase customer concentration risk by tying Lumentum solely to handful of key customers and limit its opportunities to diversify its customer base. This concentration risk is already becoming a bit more evident, with two customers accounting for 45% of revenue, at 22% and 21% respectively in fiscal Q1. This is up from 31.4% of revenue in fiscal 2025, at 16% and 15.4% respectively for the two largest customers.
Cloud Transceiver Ramp Expected to Begin Next Quarter
Lumentum’s ramp for cloud transceivers is expected to begin next quarter, with management stating that they have a line of sight to transceivers eventually becoming a $250 million/quarter business, or a $1 billion annual run rate; this is double its $500 million annual run rate today. Lumentum does not plan to expand the business beyond that $1 billion run rate, stemming from its gross margin profile.
Cloud transceiver revenue was roughly flat QoQ in Q1 with Lumentum focusing primarily on increasing manufacturing capacity in Thailand to meet rising demand. As a result, management expects to resume growth in Q2 with the upward trajectory accelerating for the next four to five quarters.
Lumentum believes Q2 will serve as a ‘proof point’ that as its new 1.6T and 800G transceivers ramp, it will see “the revenue layering benefits that our larger transceiver competitors have experienced” around the middle of 2026. The ramp of 1.6T will be important to track, as Lumentum has been straightforward about 1.6T margins being “significantly better” than 800G.
However, CEO Michael Hurlston made clear at UBS’ tech conference that Lumentum is “operating meaningfully below the mid-30s in terms of margin” for transceivers as manufacturing is “substandard” on throughput, scrap and yields. He added that there is a path to get to the mid-30s over the next few quarters as production ramps and Lumentum in-sources components (versus virtually zero in-sourced today), but this remains a headwind to the company’s target model of 42%.
Because of the lower-than-target margins, Hurlston explained that while Lumentum has “aspirations to get it to $1 billion annually, to add another $500 million of incremental revenue. But we don't want it to run much higher than that, just given the margin headwinds we see. We think we can manage our business up from a margin perspective if we keep the business to about $1 billion top line. If it gets beyond that, it will be more challenging.”
No Change to Co-packaged Optics Timing, But Demand is Stronger
As we discussed in September, co-packaged optics (CPO) is not contributing to revenue now yet could materialize into one of the biggest opportunities among all of the components and subsystems that Lumentum supplies, as the company says it is enabling Nvidia’s Spectrum-X networking switches. As a reminder, CPO places optical transceivers directly on the chip package, rather than using separate optical modules, resulting in faster data transmission, reduced latency and higher bandwidth. This may be the best of both worlds: the performance of optical yet with reduced power consumption for increasingly power-hungry AI racks.
Management provided a brief update on CPO, noting that the ramp is forecast to begin in the early stages of calendar Q3 2026 with a more meaningful contribution in calendar Q4. Hurlston explained in Q1’s call that the only change is that “demand is stronger than we initially forecast” and “getting better,” though the timing for the ramp is still the same. He clarified further that Lumentum expects “an inflection point on Ethernet-based switches. That's where we see the real step-up where our revenue would become more material” in the second half of 2026, continuing through 2027. Second-gen CPO products for 3.2T speeds are tentatively on deck for 2028.
Ultra-high power lasers are still in the initial production ramp, though Lumentum expects significant growth in shipment volumes in 2H 2026 with accelerating adoption, with this providing further confirmation of the strength of the CPO opportunity. Lumentum had announced the production expansion in early August, giving the company multiple quarters to ramp.
Optical Circuit Switching Also a Late 2026 Story
Lumentum’s second upcoming growth driver, optical circuit switches, are not expected to meaningfully contribute in Q2, rather being a late 2026 story alongside CPO. Optical switches are a new kind of switch for AI clusters that handle the switching optically instead of using transceivers to convert photons to electrons, and back again. Optical switching and CPOs work together to allow for more flexibility for reconfigurations, to reduce energy and complexity while also increasing bandwidth.
Management has outlined confidence in reaching a $100 million quarterly revenue target by the December 2026 quarter, with its two major customers expected to be qualified in the March 2026 quarter with a third customer potentially qualifying in the middle of the year. CEO Michael Hurlston provided more clarity about how Lumentum expects OCS to ramp beginning in this quarter through 2026:
“We outlined sort of a revenue ramp of kind of mid-single-digit millions here in the December quarter, getting to double digit — very, very low double digits in the March quarter and then accelerating to kind of mid $50 million, $60 million in the middle of the year and then getting all the way to that $100 million mark in the December quarter.”
This commentary implies that the largest ramp and impact from OCS will hit in fiscal Q2 2027, with management eyeing tens of millions of QoQ growth in the back half of next year. As seen in the revisions, current estimates only point to $59 million QoQ growth in that quarter, which may underestimate the tailwinds from simultaneous growth in OCS, transceivers and initial CPO growth in the second half of 2026.
Financials
Revenue Growth Maintaining >50% YoY
Lumentum fulfilled its guidance for a >$500 million revenue quarter in calendar 2025, reporting a record $533.8 million in revenue in fiscal Q1, beating estimates by just 1.4%. Revenue growth accelerated 2.5 points to 58.4% YoY though QoQ growth slowed to 11%.
As discussed previously, Lumentum guided for $630 to $670 million in revenue in Q2, accelerating to 61.6% YoY and 21.8% QoQ, whereas consensus estimates were pegged at almost 40% growth to $561.5 million.
Looking ahead, growth is expected to stay strong in Q3 at nearly 61% YoY, but the more impressive number is Q4’s estimated 54% growth, as this comes against a much more difficult comp of 55.9% vs 16.0% for Q3. This underscores the strength of the demand ramp Lumentum is discussing for the back half of calendar 2026.
On an annual view, Lumentum is estimated to report 57.3% growth in fiscal 2026 to $2.59 billion, before slowing to 29.6% YoY to $3.36 billion in fiscal 2027. Revisions are much stronger in fiscal 2027, up $600 million since the start of October versus a $300 million increase for fiscal 2026.
AI Revenue
Lumentum estimates that over 60% of total revenue comes from cloud and AI infrastructure customers, or above $320 million in Q1.
Lumentum changed its reportable segments in Q1, dropping Cloud & Networking and Industrial Tech and instead transitioning to Components and Systems. Components include laser chips, laser subassemblies, line subsystems and wavelength management subsystems, while Systems includes full stand-alone products such as optical transceivers, optical circuit switches and industrial lasers.
Components revenue rose 18.4% QoQ and 63.9% YoY to $379.2 million, fueled by “robust demand inside the data center”, strong momentum for DCI products with narrow linewidth laser assemblies for DCI transmission up 70% YoY, and record EML shipments. Lumentum expects Components to be the cornerstone for revenue growth and profitability while Systems will scale rapidly with transceivers, OCS and other high-performance solutions.
Systems revenue declined (3.6%) QoQ but increased 46.5% YoY to $154.6 million. Cloud transceiver revenue was approximately flat QoQ as Lumentum worked to increase capacity.
For Q2, Lumentum expects approximately half of its sequential revenue growth (or ~$60 million at midpoint) to come from Components, and the other half from Systems, “primarily reflecting the ramp of high-speed optical transceivers for data center applications and to a lesser extent, the early phase of our optical circuit switch ramp.”
GAAP EPS Back to Positive
Lumentum reported a razor thin $0.05 in GAAP EPS, while adjusted EPS of $1.10, up 511% YoY, beating estimates by 6.8%. For Q2, Lumentum guided for adjusted EPS in a wider range of $1.30 to $1.50, up 233% YoY, coming in well ahead of the $1.16 estimate at the midpoint. Fiscal Q3 and Q4 are expected to see adjusted EPS continue to increase, though YoY growth technically is decelerating to 162% in Q3 and 91% in Q4 as comps get more difficult.
Lumentum did not provide a full year adjusted EPS guide, though consensus now sits at $5.63, up from $4.90 and pointing to growth of 173% YoY. Considering Q2’s estimate remains below the midpoint of management’s guidance at $1.38, there is room for upside revisions if Lumentum provides another beat and raise next quarter.
Margins Show Strong Expansion
Gross margin continued to expand both sequentially and YoY, helping drive GAAP operating margin back to positive territory.
GAAP gross margin was 34.0%, in Q1, up nearly 11 points YoY and 0.7 points QoQ. Adjusted gross margin was 39.4%, up 6.6 points YoY and 1.6 points QoQ.
GAAP operating margin was 1.3%, up nearly 26 points YoY and 3 points QoQ. Adjusted operating margin was 18.7%, up 15.7 points YoY and 3.7 points QoQ, ahead of guidance for 16-17.5%. For Q2, management guided for continued adjusted operating margin expansion to 20-22%.
GAAP net margin was 0.8%, up 25.3 points YoY and not comparable QoQ due to an income tax benefit in Q4. Adjusted net margin was 16.2%, up 12.6 points YoY and 3 points QoQ.
Management provided a deeper discussion on margins moving through 2026, with product pricing from supply-demand imbalances serving as a strong lever for margin expansion:
“I think we're moving the margin line up. Pricing, obviously, is a lever. And when you look at that very, very carefully, I think what you see in the guide is some pricing, very targeted price increases happening. I think as you look out next year in 2026, our agreements with customers will include more pricing, more broad-based price increases, just given the supply-demand imbalance.”
CFO Wajid Ali added that margins are benefitting from improved manufacturing utilization, and moving into calendar 2026, gross margins are expected to move up in line with the company’s model from OFC (shown below) as OCS, 1.6T transceivers, and CPO ramp.
Lumentum is currently tracking closer towards the $750 million model by mid-2026, which is expected to see adjusted operating margin above 20% and gross margin approaching 40%. Lumentum is currently ahead of targets for adjusted operating margin per Q2’s guide, which suggests that there could be further upside as higher-margin product ramps, or some stagnation from the initial ramp phases for OCS and CPO to remain within the target ranges.
Cash Flows Muted
Cash flows were rather muted, with operating cash flow margin shrinking both YoY and QoQ.
Operating cash flow was $57.9 million in Q1 for a 10.8% margin, down from 11.8% a year ago and 13.3% in Q4. Free cash flow was ($18.3 million) for a (3.4%) margin, up from (10.2%) a year ago but down from 2.1% in Q4.
Cash and equivalents were $1.12 billion while debt was $3.24 billion.
Inventories for $531.6 million, up more than 13% QoQ, while accounts receivable surged nearly 23% QoQ to $307 million, both aligning with management’s commentary for strong product and revenue ramps over the coming quarters.
Valuation
Lumentum is trading at a stretched 6.9x forward PS multiple, more than double its five year average of 2.9x and above the 4.2x level that shares failed to break past in late 2024 and early 2025.
On the bottom line, however, Lumentum is trading just above its average multiple, currently valued at 45x forward adjusted EPS versus its five year average of 40x. Shares have traded as high as 60x and as low as 18-20x.
Conclusion
A lack of InP capacity is causing a rather substantial shortfall in EML laser supply as demand continues to expand, with Lumentum one of two companies able to meet this demand. Evidence of this tight supply is seen in Lumentum’s QoQ acceleration from 11% in Q1 to 22% guided in Q2, as Lumentum was able to increase InP capacity by 40% a few quarters ago. Looking ahead, an additional 40% capacity growth is coming online over the next few quarters, and higher yields and throughput on the upcoming capacity expansion could translate into a higher revenue growth rate.
Outside of EMLs, Lumentum is beginning to work on CW lasers for silicon photonics and co-packaged optics, which is expected to serve as the company’s next catalyst moving into 2026. This catalyst will hinge on whether Lumentum sees similar qualifications for SiPho and CPO as it has for EMLs for Nvidia’s Blackwell, or if it fails to be chosen as a lead supplier for CW moving through 2026.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Credo dropped another silly-good earnings report. This company simply won't stop shattering estimates and leaving analysts scrambling to revise their models. Fiscal Q2 revenue reported growth of 272% YoY and 20% growth QoQ for revenue of $268 million – beating estimates for revenue of $235 million and growth of 226%. The company is GAAP profitable with an operating margin of 29.4% and an adjusted operating margin of 46.3%.
However, if you thought this quarter’s 14% beat coming in $33 million over expectations was impressive, next quarter’s guide is insane. Credo guided $340 million at the midpoint vs $247.6 million expected — a 37% beat, or roughly $92 million above consensus.
Credo is capable of this strong performance due to the reliability of active electric cables (AECs). The company continues to carve out a name for itself in mission critical interconnect features such as reliability, signal integrity, latency and reach. According to management: “At 100 gig per lane today and 200 gig per lane tomorrow ZeroFlap AECs deliver up to 1,000x better reliability than traditional laser-based optical modules while consuming roughly half the power.
In addition, management stated they had four hyperscalers contribute more than 10% of revenue with the fourth in full volume ramp and the fifth starting to contribute initial revenue. This is up from three in fiscal year 2025. This diversification helps quite a bit as the lead customer cooled off in the recent quarter while another customer stepped up in revenue percentage.
This particular call was also loaded with details on the future road map with information on three new growth pillars that represent “multi-billion dollar market opportunities.” Make sure to read this section if you’re curious about what Credo has planned to expand their dominance to scale-up interconnects.
AECs Lead in Front-End and Scale-Out Connectivity
Last quarter, I made it a point to highlight the CEO’s comments on AEC reliability, because those remarks addressed why Credo continues to deliver 200%+ year-over-year growth:
“Again, reiterating that if you have got a single link flap in, say, 10,000 or 100,000 or 1 million GPU cluster, it brings the entire cluster down because there's no redundancy from that NIC to tour connection. And so we're actually seeing the TAM expanding. And I think for the first time in history that you're seeing copper replacing optical connections. So we're quite bullish on the market generally.”
This was repeated again this quarter with the magical words “expansion of AEC TAM,” which implies Credo is expanding its market as AECs become desirable for lengths of up to 7 meters (copper was traditionally used under 3 meters).
“When you're installing a 100,000 GPU cluster, link flaps can delay time to stability and time to revenue. And when you're training a model costing tens of millions of dollars, link flaps can have a significant impact on overall uptime and productivity. It is this step function improvement in reliability and power efficiency that's driving the expansion of the AEC TAM in the 100 gig and now 200 gig per lane generations. And we expect that trend to continue as customers densify racks and push cluster scale to new levels.”
The TAM is also expanding as AECs are used for front-end network connections, scale-out (or back-end) network connections and also for replacing chassis backplanes (in-rack cabling).
Perhaps most importantly, AECs may see an opportunity for scale-up networking (also back-end networking) yet what is unique about the scale-up networking opportunity is that Credo currently does not see any revenue yet here, creating yet another expansion of TAM should AECs pass qualification: “I would say the one remaining application that will be high volume is with the scale up network as that network goes rack-scale and then ultimately goes row-scale depending on the density and the number of racks that are being deployed.”
Aggressive 3-Year Product Road Map
There was brand-new information on the earnings call on how Credo plans to approach CY2026-2028 with some fairly aggressive product lines.
In the opening remarks, the CEO detailed the following:
ZeroFlap optics combine the reliability from AECs with an optical DSP and switch level SDK to integrate with their customer’s software. This allows observability data to mitigate system failures from faulty link flaps. According to management: “Our ZF optics solutions expand our addressable market to any length of connection within the data center. We anticipate initial revenue in fiscal '27 and long term, a market that will be a multibillion-dollar opportunity.”
Credo is also developing high-performance micro LED technology along with partner Hyperlume. The first product will be a pluggable optical solution that uses micro LEDs as the light source to product “active LED cables” or ALCs. The result will be ALCs that offer the same reliability and power efficiency as AECs yet can reach up to 30 meters. According to management, “We plan to sample the first ALC products to lead customers during our fiscal '27 with initial revenue ramping in fiscal '28. We believe the ALC TAM will ultimately be more than double the sizes of the AEC TAM.”
OmniConnect gearboxes are the third growth vector and will target the XPU market (or ASICs market) that uses 112G VSR SerDes for increases DDR memory capacity and throughput. According to management: “Weaver allows designers to move to commodity DDR memory and achieve up to 30x more memory capacity and 8x the bandwidth […] We anticipate initial revenue in our fiscal '28 with significant scaling thereafter.”
It’s important to note that Credo is future-proofing by designing optical solutions for the ZF flaps and ALCs. Per the Q&A session: “And I would say that, yes, ALCs as well as ZF optics, those are both optical solutions. But the OmniConnect family will be initially copper-based and then longer term, we'll offer near package optics options with that.”
Perhaps most importantly, Credo stated their goal with these new products is to move from a $1 billion annual revenue threshold to $5 billion (although there has to be quite a bit of solid execution in-between): “We've been working on these things for 18 months or so. But now being able to talk about it, I think it shows that their path to a much more diversified company long term as we think about moving the company from that $1 billion threshold of revenue annually to $5 billion and beyond over the next several years.”
We will be closely monitoring the execution around these new products in the coming quarters.
Financials
Stellar Revenue Growth of 272%
Credo’s Q2 FY2026 ending Oct 2025 revenue grew by 272.1% YoY and 20.2% QoQ to a record $268 million, beating estimates by a solid 14.1%. The robust growth was primarily led by continued strong demand for its power-efficient high-speed AI connectivity solutions, particularly its Active Electrical Cable (AEC) product line.
The company’s CEO, William Brennan, said in the earnings call, “These are the strongest quarterly results in Credo's history, and they reflect the continued build-out of the world's largest AI training and inference clusters. AI clusters are no longer measured in tens of thousands of GPUs. They're now measured in hundreds of thousands and soon millions.”
The company’s four hyperscale customers each contributed more than 10% of total revenue. The fourth hyperscaler is now in full volume ramp, and a fifth customer started contributing initial revenue in the recent quarter. The CFO, Daniel Fleming, said in the earnings call, “The largest was 42% of revenue, and that was the customer that we've, in the past, said we expect to be the largest customer this fiscal year. The second largest was 24%, which have to be our first hyperscaler to ramp a few years back. Third largest was 16%, which was our largest customer in Q1. And the fourth was 11%, which is our newest hyperscaler that we've discussed in the past.” Management expects revenue diversification to strengthen further with the fourth customer surpassing the 10% revenue for this fiscal year.
Management also provided a strong guide for the next quarter of $335 million to $345 million, representing a YoY growth of 151.8% and 26.9% QoQ at the midpoint. Notably, this guidance crushed analyst estimates by an extraordinary 37.3%, highlighting the company's robust outlook.
Management expects strong growth to continue and the CFO said in the earnings call, “As we look toward the end of fiscal year '26 and into fiscal '27, we expect sequential revenue growth in the mid-single digits, leading to more than 170% year-over-year growth in the current fiscal year. We expect each of our top 4 customers from Q2 to grow significantly year-over-year in fiscal year '26.”
Product Revenue Growth of 278%
Credo’s product revenue grew by 278% YoY and 20% sequentially to $261.3 million. This stellar performance was primarily driven by the Active Electrical Cable (AEC) product line. The AEC product line achieved new record revenue levels after posting strong double-digit sequential growth, fueled by substantial YoY growth across four hyperscale customers. Management also highlighted that customer forecasts have strengthened across the board in the past months.
IP License revenue grew by 128% YoY and up 12% QoQ to $6.7 million. The revenue growth decelerated from 152% YoY in FQ1 and accounted for only a small 2.5% of total revenue.
Strong Margins
Credo reported strong profits that exceeded management guidance. During the earnings call Q&A, management reiterated that the long-term adjusted gross margin to be in the range of 63% to 65%.
Vijay Rakesh (Analyst)
“Got it. And then longer term, as you — you're obviously seeing a pretty strong AC ramp. How should we look at the gross margin profile as optical DSPs are starting to ramp as well? Just longer term, how to look at gross margins?”
Daniel Fleming (CFO)
“Yes. We've been very consistent in saying our long-term expectation for gross margins is in the 63% to 65% range. So we are clearly at a point in time right now where we're a bit above that, but we don't expect that to be the case longer term. If you look at the more medium term, probably we guided to 65% at the midpoint. So we'll be kind of near that high end of that long-term expectation. But just longer term, I expect that to settle down into an area that historically, companies like us have been in.”
Gross profits grew by 298% YoY to $181.1 million with a gross margin of 67.5%, up 430 basis points YoY and up 10% basis points sequentially and higher than the guide of 64.5%. The adjusted gross margin was 67.7%, higher than the guidance of 65%. Management expects gross margin to be 64.8% and adjusted gross margin to be 65% in the next quarter.
The operating margin was 29.4%, up 41.1 percentage points YoY and up 2.2 percentage points sequentially, driven by strong operating leverage. It was above the guide of 23.2%. Adjusted operating margin was 46.3% compared to 11.5% in the same period last year and 43.1% in the previous quarter. Management’s operating margin guide for the next quarter is 30.1% and the adjusted operating margin is 44.4%.
Net margin was 30.8% compared to (5.9%) in the same period last year and 28.4% in the previous quarter. Adjusted net margin was 47.7% compared to 17% in the same period last year and 44.1% in the previous quarter.
Adjusted EPS beat of 35.3%
Credo’s GAAP EPS was $0.44 compared to ($0.03) in the same period last year, beating the estimates by 45.1%. Adjusted EPS grew by 857% YoY to $0.67, beating estimates by 35.3%. Analysts expect adjusted EPS to grow by 104.4% YoY to $0.51 in FQ3 and 53% YoY to $0.54 in FQ4.
Cash Flow and Balance Sheet
Credo has strong cash flow driven by growth in profits.
FQ2 operating cash flow grew by 500% YoY to $61.7 million with an operating cash flow margin of 23% compared to 14.3% in the same period last year.
FQ2 free cash flow was $38.5 million compared to ($11.7 million) in the same period last year and $53.1 million in the previous quarter. Free cash flow margin was 14.4% compared to (16.2%) in the same period last year and 23.8% in the previous quarter. Free cash flow was down sequentially due to higher capex, driven primarily by investments in production mask sets.
Cash and short-term investments were $813.6 million compared to $479.6 million in the previous quarter, and the increase was primarily from the proceeds of the ATM (at-the-market) equity offering. Credo received $384.6 million in net proceeds through the issuance of 2.7 million shares. The company announced in October that it entered into an equity distribution agreement with Goldman Sachs to raise money from time to time with a total offering of $750 million. Credo remains debt-free.
In September, the company also acquired Hyperlume, a developer of miniature light-emitting diode (microLED) based optical interconnect technology for chip-to-chip communication, for a total purchase consideration of $92 million.
The inventory was $150.2 million, up from $116.6 million in the previous quarter, suggesting strong future growth expectations.
Conclusion:
All around, Credo offered an earnings report that helps confirm the #1 leading trend in my Q4 Top 15 AI Stocks report, which was AI networking, is fully in play. Nvidia’s 162% growth in the networking segment was a nice clue, as well, that Credo would deliver tonight. I spoke about that here with Charles Payne.
It’s a good feeling when you work hard at identifying a thesis and it plays out. The beat this quarter and the strong guide next quarter suggests we are on the right track. However, AI networking will challenge even the most detailed analysts as it’s rapidly evolving, with new suppliers being qualified and new standards emerging in close succession. This is the best part of tech investing — finding the disruptors, and Credo clearly demonstrated tonight that they are one of them.
Equity Analyst Royston Roche contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund own shares in “CRDO” at the time of writing and may own stocks pictured in the charts.
Applied Optoelectronics missed revenue by $1.2M for revenue of $118.6M expected compared to $119.9M reported. The miss was due to a timing issue with management stating data center revenue was “a touch below our expectations, largely due to the timing of certain shipments at quarter-end. In particular, we had approximately $6.6 million in shipments of 400G transceivers to a large hyperscale customer, which was not able to be turned into revenue during the quarter due to various shipping and receiving delays and which we have booked in Q4.”
Although the headline numbers are causing an aftermarket selloff, the call was quite clear that the company AOI (Nasdaq: AAOI) is preparing to grow shipments significantly. Most importantly, there were discussions of an “imminent” 800G qualification coming in the next few weeks with additional hints of very strong QoQ data center growth next quarter. The earnings call signaled an important inflection point in Q4 that is not accurately depicted in Q3 numbers. We cover this and more below!
The Importance of the 800G Qualification
AOI is expected to become a large supplier for 800G and 1.6T optics, especially for its customer Amazon with a deal worth $4B over ten years. However, the 800G qualification is expected to expand beyond Amazon with management stating: “we believe we are near the final stages of qualification with several customers. We expect qualification in the near term based on conversations that we are having with our customers, and we continue to believe that we will produce meaningful shipments of 800G products in the fourth quarter.”
When asked how soon the qualifications could occur, management used the word “imminent.” This is noteworthy because AOI faced timing delays in Q3, which meant the quarter reflected very little contribution from AI data center revenue. In fact, roughly 83% of data center sales came from 100G products, while only 9% came from 200G and 400G transceivers. If we take management’s commentary at face value, this sets up a potential inflection point for AOI, as 400G — and especially 800G — products carry higher sales prices that can accelerate growth. With qualifications now described as “imminent,” AOI may be entering the part of the product cycle where higher-speed optics begin to materially impact the top line.
This understanding was echoed by management with the following statement: “Looking ahead to Q4, we expect a substantial sequential increase in our data center revenue driven by growth in 400G revenue, as well as layering in some increased 800G revenue.” Later, management quantified their expectations: “That means the data center growth should be a lot, okay, since the revenue increased by about 10% compared to Q3. That means data center revenue will increase by $25-$40 million in Q4.” Given that data center was $43.9 million this quarter, that would imply 74% data center growth at the midpoint – a sharp contrast to the (2%) decline QoQ in data center revenue this quarter.
This is supported by additional color in terms of where the company is now on shipping volumes compared to where they expect to be by year-end and mid-year 2026:
Right now, we’re only talking about maybe 10,000, 20,000. It’s a volume still far away from, quite away from. That’s why I say by end of December, we should have 100,000 per month. By end of June next year, we have 200,000 per month.”
AOI is Building USA’s Largest 800G and 1.6T Laser Production Capacity
As covered previously, in OFC in April, AOI outlined one of the most impressive capacity expansion stats that we have seen: an 8.5x increase for 800G and 1.6T products by the end of the year, with management reaffirming in both Q1 and Q2 that they remain on track to reach said target. Capacity for these high-speed products is split between two facilities, one in Taipei, Taiwan and the other in Sugar Land, Texas. Overlaying potential 3x growth industry this year with 8.5x capacity growth for AOI implies the company is eyeing market share gains into year end persisting through 2026.
If 8.5X growth was not enough, AOI is going further and aiming to double capacity again by mid-2026, stating in Q2’s call that they are expecting to be able to produce >200K 800G/1.6T products per month, with the majority produced in Texas. This corresponds to annual production of ~2.4 million 800G/1.6T products, up more than 16x from less than 150K annually prior to these expansion plans.
The fact AOI is on the verge of a sharp ramp was a central theme on the call with management stating:
As a reminder, we expect this will culminate later this year with what we believe will be the largest domestic production capacity for 800G or 1.6 terabit transceivers, approximately 35,000 transceivers per month, or roughly 35% of our overall capacity for these advanced optical transceivers. Notably, we will be able to accommodate this expansion in our current Texas facility footprint. Further, by mid-2026, we continue to expect to be able to produce over 200,000 pieces per month, with the majority produced in Texas.”
The only change in tone the I/O Fund team could pick up on is the Texas facility is expected to now produce 35% of overall capacity compared to commentary in the previous earnings call that Texas would contribute 40% of overall capacity. In Q2, it was stated:
We continue to expect to exit this year with a production capacity of over 100,000 units of 800G transceivers per month, with 40% of this production being done in the US”
Last week, the company announced its $150M investment for expansion in Texas to increase its USA production over a five-year period:
The expansion project, when complete, will have the largest production capacity for AI-focused datacenter transceivers in the U.S.”
Timing for 400G, 800G and 1.6T
I’m earmarking AOI to (hopefully) make a splash starting next quarter and into Q2 2026. It’ll be an interesting three quarters as 400G is expected to carry the revenue in Q4, then 800G in early 2026 with 1.6T taking effect by Q2.
Here is what was stated on the earnings call:
If you look at our guidance, again, just kind of go back to the segment guidance that we gave. It implies a dramatic ramp in data center revenue in the fourth quarter. We didn’t give annual guidance for next year, but we certainly believe that’s the beginning of a sustained ramp. I think we’re exactly in sync with what you described. We’re seeing that ramp first at 800G, but as we talked about, later next year, we expect 1.6 to be a strong contributor as well.”
Financials
Revenue
AOI reported $118.6 million in revenue in Q3, slightly below estimates for $119.8 million and at the lower end of management’s guidance for $115 to $127 million. This represented growth of 15.2% QoQ and 82.1% YoY, decelerating from 137.9% YoY in the second quarter, driven by strong cable TV demand and the ramp of 1.8 GHz amplifier products as data center revenue was soft.
For Q4, AOI guided for revenue between $125 and $140 million, up 11.7% QoQ and 32.1% YoY, another sharp deceleration though this comes against much tougher comps. Management expects to recognize 800G revenue in the fourth quarter: “we continue to believe that we will produce meaningful shipments of 800G products in the fourth quarter.”
Key Segments
CATV (Cable TV):
CATV revenue surged 237.1% YoY and 26.1% QoQ to a record $70.6 million, with management characterizing demand as “exceptionally strong.”
Data Center:
Data center revenue was $43.9 million, up 7.3% YoY but down (1.9%) QoQ, with management explaining that this “came in a touch below our expectations, largely due to the timing of certain shipments at quarter end due to various shipping and receiving delays.” AOI also added that it is seeing increased orders for 100G and 400G products from several large customers, and expects increased demand for both through the end of the year.
According to the opening remarks, the split across products was “In the third quarter, 83% of data center revenue was from 100G products, 9% was from 200G and 400G transceiver products, and 7% was from 10G and 40G transceiver products.”
Telecom/Other:
Telecom revenue rose 33.7% YoY and 92.8% QoQ to $3.74 million, while other revenue was $0.35 million.
Margins
AOI showed a marginal sequential improvement in GAAP operating margin despite GAAP gross margin contracting, though the company is not meaningfully closer to GAAP profitability.
GAAP gross margin was 28.0%, down 2.3 points QoQ but up 3.6 points YoY. Adjusted gross margin was 31%, at the high end of guidance and up 0.6 points QoQ and 6 points YoY.
GAAP operating margin was (15.3%), a slight improvement from (15.5%) in Q2 and up more than 10 points YoY. Adjusted operating margin was (8.7%), improving 1.8 points QoQ and 9.2 points YoY.
GAAP net margin was (15.1%), down from (8.8%) in Q2 but up from (27.3%) in the year ago quarter. Adjusted net margin was (4.6%), below guidance for (3.3%) but marking an improvement from (8.6%) in Q2 and (13.5%) in the year ago quarter.
For Q4, management guided for adjusted gross margin to be 29-31%, down 1 point QoQ but up 1.3 points YoY at midpoint. Adjusted net margin was guided at (4.5%), approximately flat QoQ.
EPS
AOI met adjusted EPS estimates this quarter at ($0.09), though Q4’s guidance missed as the company is still forecasting a small loss whereas estimates were expecting a shift to profitability, albeit at a thin $0.03.
Q3 GAAP EPS was ($0.28), improving from ($0.42) in the year ago quarter but widening from ($0.16) in Q2. This missed estimates for ($0.10).
Q3 adjusted EPS met at ($0.09).
Q4 adjusted EPS was guided to be ($0.13) to ($0.04), short of consensus for $0.03.
Cash and Balance Sheet
Cash flows significantly improved from Q2, and inventories rose sharply once again, likely in preparation for the ramp of 800G products.
Q3 operating cash flow was ($28.5 million) for a (24%) margin, improving from a (63.6%) margin in Q2 but down slightly from (22.2%) a year ago.
Q3 free cash flow was ($57.5 million) for a (48.5%) margin, improving from (101.3%) in Q2 but still lower from (32%) a year ago.
Cash and equivalents totaled $150.7 million and debt totaled $192.1 million.
Inventories were $170.2 million, up 22.5% or $31.3 million QoQ. Since Q1, inventories have risen by ~$68 million.
Conclusion:
If the market were always on our side, investing would be easy. What we’re seeing this quarter is a reminder that even when a company’s story remains intact, the market can still get the jitters. This stock, however, continues to get a green light from me across the board — the headline “miss” was a non-issue (on the contrary – it’s a boon for the strong QoQ commentary on Q4). Looking ahead, shipments are on track to increase 16X over the next year, kicking off soon with key 800G qualifications only weeks away. Management is feeling comfy enough to include 800G in the Q4 guide — a meaningful signal. Of course, nothing is ever certain in investing. But based on what I heard this evening, this is not a report I’m concerned about.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund own shares in Applied Optoelectronics at the time of writing and may own stocks pictured in the charts.
Vertiv will not win any hypergrowth stock awards, especially as management has previously offered CAGR guidance of 15% to 17% through 2029. Rather, it’s where Vertiv is positioned as an AI infrastructure partner especially as the trend turns toward modular infrastructure that makes this a stock to watch. Essentially, all roads point toward Vertiv’s power and thermal solutions becoming increasingly important for future generations of rack scale solutions, with the company already preparing 800V DC solutions for Nvidia’s Rubin Ultra platform due in 2027.
Revenue
Vertiv reported revenue up 29% YoY and 1% QoQ to $2.676 billion, well ahead of its original guidance for 23% growth in the third quarter. This was driven by 43% YoY growth in the Americas on accelerated AI demand and 20% growth in APAC.
For Q4, Vertiv guided for revenue to be $2.81 billion to $2.89 billion, up 6.5% QoQ and 18-22% YoY at the $2.85 billion midpoint. While this was ahead of previous guidance for $2.735 to $2.815 billion, this would still represent a nine point deceleration on the topline at midpoint. Management expects Americas revenue to be up high-30s, APAC up mid-single digits and EMEA down high single digits but up mid-teens QoQ.
The strong outperformance in Q3 also led to Vertiv hiking its FY25 revenue guidance from $10 billion at midpoint to $10.2 billion at midpoint, pointing to organic growth of 26-28% YoY. Management did not provide any direct insight into FY26, though they did say that based on the “substantial backlog and clear visibility of pipeline, we anticipate continued significant organic sales growth in 2026,” with EMEA potentially reaccelerating in 2H 2026.
AI Revenue Metrics
Vertiv’s backlog rose ~30% YoY and 12% QoQ to $9.5 billion, reaccelerating from 21% YoY growth last quarter. More importantly, the $1 billion sequential increase in backlog was the largest in more than two years. However, one of the stronger metrics was order growth, with Vertiv reporting organic orders up 60% YoY and 20% QoQ in Q3. This drove a ten point rebound in TTM organic order growth to 21% YoY, from 11% in Q2.
However, starting in Q4, Vertiv will no longer report on quarterly orders and backlog information, and instead will report a new metric “projected full year orders.”
The following was stated in Q2: “Beginning on our Q4 and full year 2025 earnings call, we will provide projected full year orders rather than quarterly orders and backlog information. We believe this better aligns with how we run our business. We will provide updates on the full year projections quarterly as we progress through the year and as we deem necessary.” This could create a boost to Vertiv’s stock to remove the lumpiness from quarterly reports and to also be more forward looking in terms of visibility offered to investors.
Earnings
Vertiv reported adjusted EPS up 63% YoY to $1.25 in the quarter, beating the $0.99 estimate by 25%. GAAP EPS of $1.02 beat estimates by 16.7%. For Q4, adjusted EPS was guided to decelerate to 27% growth to $1.26 at midpoint.
For the full year, Vertiv raised its adjusted EPS forecast to $4.07 to $4.13, up from its prior view for $3.75 to $3.85. At midpoint, this represented a nearly 8% hike, now pointing to 44% YoY growth versus 33% previously.
Margins
Vertiv reported expanding margins across the board in Q3, though Q4 is expected to be approximately flat for adjusted operating margin.
Gross margin was 37.8%, up 1.3 points YoY and 3.8 points QoQ.
GAAP operating margin was 19.3%, up 1.4 points YoY and 2.5 points QoQ. Adjusted operating margin was 22.3%, up 2.2 points YoY and 3.8 points QoQ, driven by tariff mitigation efforts and strong execution addressing operational inefficiencies.
Net margin was 14.9%, up 6.4 points YoY and 2.6 points QoQ.
For Q4, adjusted operating margin was guided to be up 0.9 points YoY and approximately flat QoQ at 22.4%, as “progress addressing operational inefficiencies [is] offset by acceleration in growth investments and negative impact from new tariffs.” This is a rather steep decrease from Q2’s guidance for 23.6%, which would’ve been its best adjusted operating margin print since going public in 2020.
For FY25, Vertiv slightly raised its adjusted operating margin forecast by 0.2 points at midpoint to 20.2%, representing YoY expansion of 0.8 points. This is strong as it comes in the face of “significant headwinds from tariffs and operational inefficiencies driven by supply chain actions to mitigate tariffs.” Tariff impacts are expected to be materially offset exiting Q1 ’26.
Cash
Vertiv reported strong cash flows in Q3, with operating cash flow of $508.7 million, up nearly 36% YoY. OCF margin was 19%, up 1.8 points YoY and 6.8 points QoQ.
Q3 adjusted free cash flow was $462 million, up 32% YoY. Adjusted FCF margin was 17.3%, up 1.1 points YoY and 6.8 points QoQ. Q4 adjusted FCF was guided to be $496 million for a 17.4% margin, up marginally from Q3. Vertiv boosted its adjusted FCF guidance by $100 million, now forecasting $1.5 billion for the year, up from $1.4 billion previously. This corresponds to a 14.7% margin.
Accounts receivable dipped (1%) QoQ to $2.81 billion, while inventories rose less than 2% YoY to $1.43 billion.
Cash, equivalents and investments totaled $1.94 billion, while debt totaled $2.90 billion.
Valuation
Vertiv is trading at peak multiples on the top line, and slightly below peak on the bottom line. Vertiv’s forward PS is 7.2x, above its late 2024 peak of 6.8x, and substantially higher than its April low at 2.2x forward PS.
On the bottom line, Vertiv is just below peak multiples, at 47.3x forward earnings versus its peak at 52.5x.
Notable Risks
Vertiv’s extended valuation is a primary risk as the company contends with a sharper deceleration on the top line heading into Q4, as well as a sharp deceleration in EPS growth from 63% in Q3 to 27% in Q4. Margins are also a line item to watch, considering management had guided for a Q4 adjusted operating margin of 23.6% back in Q2 but then subsequently cut that guide to 22.4% in Q3.
Conclusion:
The current quarter was not a showstopper as we prefer to be allocated more heavily to stocks that are showing signs of imminent Blackwell participation. However, Vertiv remains one to watch as the backlog increasing 12% QoQ and order growth increasing 20% QoQ could be signaling an inflection.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Applied Optoelectronics (AOI) (NASDAQ: AAOI) is a lesser-known optical component and transceiver supplier. The small-cap has recently caught our attention for its hyperscaler deals and its ambitious capacity growth targets, aiming to become one of the largest domestic manufacturers for 800G transceivers.
The VSCEL laser, EML chip, and optical transceiver supplier has provided some of the strongest commentary for future growth, with management outlining 8.5x capacity growth by year-end and substantial capacity growth in 2026. This builds upon a comment slipped into Q1’s call that AOI believes it can become the largest 800G/1.6T optics supplier to Amazon, after signing a deal in March that management says is worth more than $4 billion over ten years.
AOI also stood out from its 40% QoQ growth in data center revenue in Q2, though this came after a soft Q1 affected by inventory digestion at its largest hyperscale customer and seasonality. Interestingly, data center was not the core driver of growth in Q2, with cable TV taking the reins with 8x YoY growth in the quarter. Cable TV is expected to provide an additional lever of growth come 2026, with management tentatively outlining early visibility into $300M+ in demand.
The crux with AOI is that cash flows are much softer than more-established AI networking stocks, and GAAP margins are still negative. However, it’s hard to argue with the implications from the profound expansion in capacity and how optics growth can transform margins.
Plotting Out the Optical Opportunity
Here’s a quick recap from our first thematic coverage on optical interconnects from over a year ago and why optical components would become a necessity as AI workloads rise.
Copper had long been standard for data center interconnects, but it cannot support network speeds of 800 gigabits (800G) to 1.6 terabits (1.6T) over long distances due to substantial signal loss. This isn’t to say copper is dead – Nvidia’s GB200 NVL72 utilized copper over optics (with more than 2 miles of copper cabling in the rack) to reduce power consumption by 20 kw (the system still draws 120kw of power). According to a representative from Marvell’s Cloud Optics division, “optical is the only technology that can give you the bandwidth and reach needed to connect hundreds and thousands and tens of thousands of servers across the whole data center.”
Optical transceivers are crucial in enabling high-speed data transfer, by transmitting and receiving data from optical (light) signals to electrical signals. In data centers, optical interconnects and transceivers are becoming the de facto standard to handle AI workloads, since they can function at significantly higher speeds than copper (currently at 800G+ speeds and moving quickly to 1.6T), with longer range, higher data capacity, and lower latency with minimal signal loss. One drawback, however, is that due to the electronic complexity of optical products, costs are higher as well as power consumption versus copper.
To read the full analysis on optical interconnects, click here.
Major optics manufacturer Innolight reportedly raised its forecast for global 800G unit shipments by 50%, now seeing 15 million shipments globally for 2025, up from 10 million previously. Other reports out of China were more optimistic, with one firm projecting >16 million 800G shipments this year with 1.6T demand exceeding 5 million units, and another projecting demand for 17 million 800G and 4 million 1.6T units. FS had placed a rough estimate of 5 million shipments for 800G products in 2024, so these projections would represent >3X growth YoY if they come to fruition.
Nvidia’s math also aligns with the upper end of those demand scenarios for >800G transceivers for scale-out connectivity, with the GPU leader claiming a 1:6 GPU-to-transceiver ratio for 100K 4-GPU servers. Nvidia states that this scenario, with 400K total GPUs, would require 2.4 million transceivers.
Nvidia estimates that optical module requirements for scale-out networks may be at a ratio 1:6 per GPU. Source: Nvidia via The Next Platform
While a majority of Nvidia’s shipments are likely to focus on the 72-GPU rack-scale solutions rather than smaller servers, a rough estimate for 750,000 servers, or 3 million GPUs (less than half of Nvidia’s estimated 2025 unit shipments), would project 18 million in transceiver demand solely to meet Nvidia’s needs.
For intra-rack connections for the GB200 NVL72, reports suggest that “the ratio of GPUs to 1.6T optical transceivers is 1:2 in the dual-layer InfiniBand network and 1:3 in the three-layer InfiniBand network. Compared to the DGX H100, the NVIDIA GB200 NVL72 architecture doubles the port speed for servers and switches, significantly advancing the adoption of 1.6T transceivers and increasing the need for the 1.6T network.” Assuming roughly 30,000 NVL72 units in demand, this would project to between 4.32 million to 6.48 million 1.6T transceivers for intra-rack connections.
AOI Targeting 8.5X Capacity Growth for 800G/1.6T Products by Year-end
At OFC in April, AOI outlined one of the most impressive capacity expansion stats that we have seen: an 8.5x increase for 800G and 1.6T products by the end of the year, with management reaffirming in both Q1 and Q2 that they remain on track to reach said target. Capacity for these high-speed products is split between two facilities, one in Taipei, Taiwan and the other in Sugar Land, Texas. Overlaying potential 3x growth industry this year with 8.5x capacity growth for AOI implies the company is eyeing market share gains into year end persisting through 2026.
Below is how AOI’s management is charting out the growth in capacity, with consistent monthly expansion taking it to its 8.5x target:
Capacity growth figures are relative to a March 2025 baseline for Taiwan production and a September 2025 baseline for Texas production. Source: AOIAOI
Note that these figures are cumulative, meaning that AOI is essentially targeting 2.35x more capacity by June from its March baseline in Taiwan for its 800G 2xFR4/DR8 product, before scaling to 3.65x by August and so forth. This is also the case in Texas, where AOI is targeting the start of production in September (100%), before scaling 4x by December. Equipment that was ordered in Q1 was said to be arriving as of Q2, further supporting this expansion effort.
Putting this in perspective, the 8.5x growth is targeting approximately 100K monthly capacity globally of 800G/1.6T products by year-end, according to management. This suggests that capacity earlier in 2025 was roughly 12K per month. AOI expects ~40% of 800G production to be domestic at its Texas facility, which they believe will make them one of the largest domestic manufacturers. More importantly, AOI says they can accommodate this expansion under its current facility footprint, saving on capex.
Management Eyeing 2X Growth in 2026
If 8.5X growth was not enough, AOI is going further and aiming to double capacity again by mid-2026, stating in Q2’s call that they are expecting to be able to produce >200K 800G/1.6T products per month, with the majority produced in Texas. This corresponds to annual production of ~2.4 million 800G/1.6T products, up more than 16x from less than 150K annually prior to these expansion plans. Compared to market leaders Innolight and Eoptolink, where monthly capacity is estimated at 500K and 300K units respectively (for annual capacity of ~9.6M), AOI still lags but is quickly catching up. The company believes it holds an advantage from its vertically-integrated model with in-house laser and subassembly manufacturing, along with substantial factory automation that allow it to quickly ramp capacity to this degree.
Based on Nvidia’s calculations, AAOI alone may be able to support 400K GPU shipments at full capacity, or more than 5% of UBS’ estimate for Nvidia’s GPU shipments of 7.4 million next year. More importantly, none of this capacity so far, not even the 8.5X growth, has translated into revenue, with 800G contribution immaterial to revenue as of Q2 as products remain in the qualification stage.
While pricing is unknown, an analyst had questioned management about the ramp of 800G/1.6T products in Q1 and penciled in $0.75/gig: “But I guess rough math, yeah, $0.75 a gig, that would be well over $100 million in a quarter. Am I doing something wrong in that thinking?” The question was quickly shot down by management, yet assuming pricing at this degree as the market becomes more competitive through 2026 provides an interesting look into the potential growth ahead for AOI.
Assuming by Q3, AOI is operating at max capacity of 200,000 pieces per month, with 80% of those being 800G and 20% 1.6T. This would roughly project ~$96 million monthly revenue for 800G products and ~$48 million monthly revenue for 1.6T products, assuming capacity is sold out.
On a quarterly basis, this would be north of $430 million in quarterly revenue simply from >800G products, not including 400G or cable TV, already more than double current consensus estimates for $194 million in Q3 and $217 million in Q4.
More Clarity on 800G Ramp, 1.6T
Considering the combination of commentary for >16X capacity growth for 800G and 1.6T products in just one year and the fact that the opportunity is squarely ahead, tracking the timing and scale of this revenue ramp is critical. Based on commentary from Q2, 800G accounted for, at maximum, 1% of data center revenue, considering shipments were only for qualification purposes.
Management said that they “continue to believe that we will produce meaningful shipments of 800G products sometime in the second half of 2025, likely in late Q3 or Q4. The schedule is constrained by our ability to build and qualify production capacity. We believe that the demand for 800G is strong, and we expect that when our production is ready, we will see a fairly quick ramp in revenue for 800G.”
The timing of the ramp comes down to two factors: AOI has to have meaningful production capacity available in order to quickly shift to volume production, and also wait for its prospective customers to finish production qualification. October and November are expected to see larger jumps in production capacity, with Taiwan capacity rising from 5x to 8.3x, and Texas up to 3x from its baseline, suggesting the ramp is imminent yet pending the conclusion of qualification.
AOI believes that one of its major hyperscale customers is in the final stages for securing 800G qualification, having recently audited and approved AOI’s Taiwan factory for 800G production. This adds an additional layer of confidence that the revenue ramp for 800G is near, as management expects that this customer could become a >10% customer in Q3.
For 1.6T, AOI expects to kick off volume manufacturing around June to July 2026, noting that it is already working with several customers and expects to have a minimum of three tier 1 customers. All three are pre-existing customers, so this could include Microsoft, Amazon or possibly even Oracle.
Another reason that the ramp of 800G and 1.6T is important is tied to margins: management explained that for 1.6T, gross margins should be above 40%, while 800G margins should be close to 40%. As discussed further below in the Financials section, AOI’s current gross margin is hovering at 30%, meaning a strong ramp for these products will drive gross and likely operating and net margins higher – management has guided for 35% to 40% gross margins by the end of 2026 and these products will be crucial to reach that.
AOI shares surged in early March when the company announced that it signed a 10-year supply deal with an Amazon subsidiary, widely understood to be AWS, while offering the company warrants for 7.94 million shares.
This is not AOI’s first long-term agreement with a major hyperscaler for data center products – the company signed a five-year supply agreement with Microsoft in 2023 for >400G products through 2028, and expects revenue from this contract to increase in 2025 relative to 2024. Terms for the Microsoft deal were not disclosed.
AOI has disclosed additional details about the Amazon deal: 1.32 million shares vested upon the signing of the agreement, but the remaining 6.62 million shares may vest over the next 10 years “dependent on aggregate purchases by Amazon of $4 billion of our products over this time period.” CEO Thompson Li said in Q1 that he believes the deal could be “much more than $4 billion in the next 10 years.” This would correspond to annual revenue on average of at least $400 million solely from Amazon.
Q2 had quite an interesting snippet from management regarding the deal:
“Our belief, our expectation, is that we can grow to be, the largest supplier of 800G and faster, higher data rate optics for Amazon. Now, that's not guaranteed by any means, but I don't see any reason why we couldn't be there. And that would imply a market share; typically, they're going to have two or three suppliers, so that would be, maybe 30, maybe even up to 40%.”
This quote is very important as it indicates the Amazon deal could create a long-term, high-volume customer contributing hundreds of millions in annual revenue, it also signals product validation from an industry leader and could lead to future supply deals, assuming capacity supports it.
Significant Customer Concentration a Risk to Consider
AOI exhibits a higher degree of risk related to its significant customer concentration, with its top ten customers accounting for 98% of revenue in Q2, a slight increase from 94% in the year ago quarter.
AOI’s largest customer was in its CATV segment, contributing 54% of total revenue, while its second largest customer was in its data center segment and contributed 34% of total revenue. These two customers alone, likely Digicomm and Microsoft based on prior revenue trends, combined for 88% of revenue. As mentioned previously, management believes they could have a third >10% customer in Q3, a major hyperscaler for 400G and 800G products.
Oracle was previously a >10% customer as recently as 2024, contributing 12.4% of total revenue last year, up from 8.8% in 2023. Microsoft had contributed nearly 44% of revenue as AOI’s largest singular customer in 2024, while Digicomm accounted for 34%, suggesting the recent growth in CATV has swapped its position.
High China Revenue Concentration
AOI has elevated China revenue exposure, which raises risk considering that geopolitical tensions are flaring up again. In Q2, China contributed $62.4 million in revenue, up 403% YoY and accounting for nearly 61% of revenue, with this likely driven by the growth in cable TV. Taiwan revenue was $39.5 million, up 37% YoY to 38% of revenue.
On the other hand, AOI is aiming to reduce its China content in its transceivers to near zero as they scale production, with current China content at <10% for its 800G and 1.6T products. AOI says that a majority of its key components for these products, such as laser chips, are already manufactured in the US.
Financials
Revenue
AOI reported revenue of $102.95 million in Q2, up 138% YoY and 3% QoQ, representing a modest (2.7%) miss versus consensus estimates. This was a slight deceleration from 146% growth in the previous quarter. Cable TV was the primary growth driver with revenue up 862% YoY to $56 million, while data center revenue rose 30% YoY to $44.8 million.
For Q3, AOI guided for revenue to be between $115 million to $127 million, pointing to 86% YoY growth at midpoint, though on a sequential basis this is a sharp uptick to 17% QoQ growth. Management expects that there may be some initial contribution from 800G modules late in the quarter: “Looking ahead to Q3, we expect a sequential increase in our datacenter revenue, driven by continued growth in our 100G and 400G products with the possibility of layering some additional increased 800G revenue late in the quarter.”
For fiscal 2025, revenue is currently projected to be $467.3 million, up 87.4% YoY, before slowing to 55.9% growth in fiscal 2026 to $728.4 million. However, there is potential that the 800G and 1.6T capacity ramp meaningfully accelerates AOI’s run rate by the end of 2026.
Key Segments:
CATV (Cable Television)
As noted above, CATV revenue in Q2 was up 862% YoY but down (13%) QoQ from a record Q1, in line with management’s expectations as production was retooled for Motorola-style amplifiers. The YoY growth was driven by the ramp of its 1.8 GHz amplifiers.
AOI also finished testing and certified its Motorola- and GameMaker-style amplifiers for Charter Communications, who will deploy AOI’s 1.8 GHz amplifies and QuantumLink remote management software. For Q3, AOI expects CATV revenue to be record or near-record.
Data Center
For Q2, data center revenue was $44.8 million, up 30.4% YoY and 39.8% QoQ. The QoQ increase comes after a (27.6%) QoQ decline in Q1 from inventory digestion at one of AOI’s largest hyperscaler customers and seasonality. Data center revenue has been lumpy so far, though growth is expected to pick up in Q3 and into Q4 as 400G ramps and 800G begins contributing.
Revenue for 10/40G products rose 68% YoY and 26% QoQ to $4.0 million, or 10% of data center revenue
Revenue for 100G products rose 25% YoY and 25% QoQ to $31.4 million, or 70% of data center revenue.
Revenue for 200G/400G products rose 43% YoY and 180% QoQ to $8.9 million, or 20% of data center revenue. AOI also completed its first volume shipment for high-speed single-mode 400G transceivers to a recently re-engaged hyperscaler, with increased sequential demand from other hyperscalers arising in Q2.
Telecom
Telecom and other revenue was down (45%) YoY to $2.1 million. Management expects telecom revenue, which was $1.9 million, to fluctuate from quarter to quarter.
Margins
AOI is not the strongest on margins, with operating margin widening for a second consecutive quarter on increased expenditures related to customer qualification projects for 800G and 1.6T products.
AOI reported GAAP gross margin of 30.3%, up more than 8 points YoY but down 0.3 points QoQ. Adjusted gross margin was 30.4%, with management guiding this to be essentially flat next quarter at 30.3%. Management has a target of 35% to 40% by late 2026 to 2027, emphasizing that they will need a few quarters for higher-margin 800G and cost reduction efforts from shifting to larger wafers to be seen. In the near-term, the ramp of single-mode 400G transceivers, which carry higher ASPs and gross margins, may provide a tailwind for expansion.
GAAP operating margin was (15.5%), a notable improvement from over (60%) in the year ago quarter, but widening from (6.5%) in Q4 and (8.9%) in Q1. Adjusted operating margin as (10.5%), down from (4.8%) in Q1 but up from (37.6%) a year ago.
Net margin was (8.8%), improving from (9.2%) in Q1 and more than (60%) a year ago. Adjusted net margin was (8.6%), down from (0.9%) in Q1 but improving from (25.1%) a year ago. For Q3, management’s guide implies adjusted net margin improving to (3.3%) at midpoint.
EPS
AOI reported a ($0.16) loss per share in Q2, missing estimates for ($0.07), as net margin did not improve much sequentially. For Q3, AOI guided for ($0.03) to ($0.10), a decent sequential improvement, likely driven by the ramp in 400G and a potential rebound in operating margin
Looking ahead, AOI is expected to shift to GAAP profitability by Q4 and expand earnings through mid-2026, with current estimates pointing to $0.03 in Q4 and rising to $0.19 by Q2 ’26, driven by the ramp of 800G.
Cash
Cash flows are also quite weak, though this has been mostly from surging accounts receivables and capex to support capacity expansion, a solid signal for the upcoming 800G ramp in conjunction with management’s commentary.
Operating cash flow in Q2 was ($65.5 million), widening from ($50.9 million) in Q1 as inventories and accounts receivables surged, up $36 million and $40 million QoQ respectively. OCF margin was (63.6%), down from (4.6%) a year ago and (51%) in Q1.
Free cash flow was ($104.3 million), widening from ($87.2 million) in Q1. Capex for the quarter was $38.8 million as AOI is quickly purchasing equipment to meet its aggressive capacity expansion targets for year-end and mid-year 2026. Full-year capex guidance is $120 to $150 million in preparation for increased 400G, 800G and 1.6T production.
Inventories were $138.9 million, up more than $36 million from Q1, driven by raw material purchases for use in production over the next few months.
Accounts receivable were $211.5 million in Q2, up $40 million QoQ. Digicomm accounted for $171.6 million of these receivables, tied to upcoming cable TV growth: “We talked about the dynamics, because we wanted to get some of the cable TV products in particular into the country and ready to be staged, ready for customer acceptance, we have offered some extended payment terms to certain customers in that channel chain to be able to accommodate that additional amount of revenue so that it's here when the customers need it.”
Cash and equivalents totaled $87.2 million, while debt was $188.2 million. During Q2, AOI completed its ATM program and raised $98 million net of fees, which it will utilize for equipment and machinery for its capacity expansion in Taiwan and Texas.
Earnings Q&A:
Gross Margin Increase Driven by Wafer Sizing
A five to ten point expansion in gross margins from 30% to 35-40% in four to six quarters (end of 2026 to early 2027) is a tall task to achieve while greatly expanding production capacity, but management provided some insights into the different levers available to reach this target. First, a shift from 2-inch wafers to 3-inch wafers is expected to bring substantial cost savings, along with 800G products ramping and improvements in cable TV:
Q, Simon Leopold, Raymond James:
Where I was trying to go with the question was to try to get a better sense of one of the elements to help the gross margin move towards that long term of 40%. So what I was trying to tease out in this question was the degree that you're outsourcing today versus a change towards more vertical integration in the future as a lever for gross margin improvement. So maybe the question is off-base and maybe I'm going down the wrong path. More bluntly, what will help the gross margin improve?
A, CEO Thompson Lin:
“Yes. I think the key is wafer, okay? Right now, we are doing 2-inch wafer. But we're going to 3-inch wafers. The cost will reduce by, I don't know, 50%, 60%. Then we'll go to 4-inch wafer by end of next year. This is a major, much bigger cost savings than what you're talking about. I think right now, yes, we're only maybe using 30% to 40% of our lasers. We were using, I would say, 2/3 of AOI lasers, okay? It will depend on customer by customer [basis]. Some customers prefer all the AOI lasers. Some customers prefer 50-50, okay? So that's why it's different. But more importantly, the cost funnel of AOI lasers changed from 2-inch to 3-inch to 4-inch.”
Management added that they need a few quarters to see a bigger impact from 800G products, which will have gross margin near 40%, and increased software revenue mix in cable TV, such as the QuantumLink remote management software deployment with Charter Communications.
Cable TV Demand to Reach $300-350M
Though the data center opportunity is what we’re most interested in, AOI’s commentary on cable TV pointed to potential 40% growth in fiscal 2026. Management said they have a “pretty clear line of sight” into channel inventory and demand, which supports this confidence in reaching $300 to $350 million in cable TV revenue next year. For perspective, cable TV revenue is on a ~$240 to $260 million run rate with $120 million in revenue in 1H:
CEO Thompson Li:
“So right now, next year, we are very comfortable, besides Charter, we should have more than 10 customers next year. And right now, based on the feedback from this customer, I think the real demand from these customers next year is, I would say, minimum $300 million to $350 million. … But the demand is pretty big, all right? Just next year, that's the number we see right now, $300 million to $350 million of real demand, all right, for this customer in, I would say, U.S., Canada, all right?”
CW Laser Production Increasing
AOI has been rather quiet about CW lasers for silicon photonics, with the focus primarily on the 800G transceivers, though management discussed increasing capacity for the high-power CW lasers to 2.5 million per month, or 30 million annually, by “sometime next year.”
We have previously discussed the importance of CW lasers for SiPho in our Lumentum analysis, Lumentum at Inflection Point with 20% QoQ Growth in AI-Related Segment, where Lumentum said it was “having challenges actually getting enough CW lasers for our own transceivers.” This will give AOI easily enough capacity to meet its planned transceiver capacity growth in a vertically-integrated manner:
“We are increasing our high-power CW laser for silicon photonics to maybe 2.5 million lasers per month by sometime next year. So right now, our in-house capacity is 100G EML. We should have 200G EML sometime soon, next year, for sure, the high-power laser for silicon photonics and VCSEL, the other new project, the 200G photo detector, okay? So this is all manufactured, 100%, in Houston.”
Conclusion
The primary risks with AOI stem from its margin profile as it is much weaker than more established networking players as well as the hypergrowth players we track. While there is an expectation for margins and EPS to improve through 2026 as 800G/1.6T products ramp, this will need to be proven over the coming few quarters. The weak cash flows also present a risk if there is a prolonged recovery, given the thinner cash position AOI has, while China concentration also poses a larger red flag given geopolitical tensions may be rising again.
However, it’s hard to argue with the strong growth expected across the optical transceiver industry emerging through 2026 from 800G and 1.6T products, combined with AOI’s 17X increase in capacity through mid-2026 for said products. Rough math suggests these products could generate more than double current quarterly revenue estimates when selling out at full capacity, while cable TV is expected to contribute nicely to growth in 2026 as well.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Every Thursday at 4:30 pm Eastern, the I/O Fund team holds a webinar for premium members to discuss how to navigate the broad market, as well as various stock and crypto entries and exits. Beth Kindig offers weekly deep dives including lesser-known cryptocurrencies and AI stocks, plus the team offers trade alerts. The I/O Fund team is one of the only audited portfolios available to individual investors. If you’d like to subscribe to the Advanced Market Signals plan, email us at premium@io-fund.compremium@io-fund.com.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Since launching the Discovery tier less than a year ago, The I/O Fund’s internal process for identifying new winners has greatly improved. Based on the research we produced from this tier, we added stocks like Bloom Energy, Core Scientific and Oklo to our portfolio, and these new additions became some of our biggest wins (thus far) in 2025.
As we continue to build out the Discovery library, we’d like to make it as easy as possible for our readers to follow along. We want to cast a wide net, explore and leave no stone unturned – therefore, we anticipate our coverage to include dozens of stocks over time. However, we also want to make sure we get down to brass tracks by providing you a clear takeaway by ranking what we’ve dug up every quarter. The ranking will also help clarify which ones we are eyeing an entry for and what setups Knox will cover in his Discovery Tier webinars. This will be called the I/O Fund’s Top 10 New Ideas List.
The ranking we provide is an estimate, which means all 10 stocks are of interest. Given the nature of momentum stocks, the ranking could shift quickly. Please check back to the Discovery Ranking list provided on the Dashboard for any changes in the interim.
Themes:
AI Networking:
Networking is at the heart of the new architecture that Nvidia is shipping now as the increased bandwidth is instrumental in driving higher performance. For example, the NVL72 systems will deliver 4X faster training and 30X faster inference compared to HGX systems. Notably, this is accomplished with many more GPUs from eight to 72 per system.
To support the new systems, the NVLink domain moves from supporting eight GPUs to 72 GPUs and 36 CPUs with a speed of 1.8 TB/s with 18 NVSwitch ASICs, up from four in the HGX/DGX systems due to increasing the number of GPUs but also due to doubling the per-GPU links. The 5th generation of NVLink supports up to 576 GPUs compared to the fourth generation of up to eight GPUs.
Scale-up refers to increasing the number of GPUs in an AI system. Proprietary NVLink remains the highest performance option for scale-up interconnects, although PCIe and scale-up Ethernet are also used. For the cabling, copper is used for intra-rack scale-up with up to 5,184 cables per system. Future generations of NVLink are likely to integrate optical I/O so that GPUs can communicate across racks without requiring costly retimers.
GB200 NVL72 with 72 GPUs and 36 CPUs has 18 NVSwitch chips and 72 InfiniBand NICs for scale-out networking and 36 Bluefield-3 Ethernet NICs for front-end networking. Compare this to the HGX systems with 8 GPUs and the DGX systems with 8 GPUs and 2 CPUs has 4 NVSwitch chips and 8 InfiniBand NICs.
This means the new architecture that Nvidia is shipping now results in 9X more GPUs, 4.5X more NVSwitches, 9X more InfiniBand NICs and 18X more front-end NICs. Each GPU requires its own InfiniBand link for scale-out whereas NVSwitch components grow faster than GPU count as each GPU must talk to every other GPU, therefore, it has more of an exponential growth.
Although NVLink is proprietary, it acts as a bellwether for the importance of AI networking and lesser-known suppliers. Generally speaking, what we can see from looking more closely at Nvidia’s networking fabric is that networking components are increasing 5X to 9X, and in some cases up to 18X.
Scale-out racks refer to connecting multiple racks across a cluster. InfiniBand switches and Ethernet switches are used for this purpose. As you can see below, Nvidia offers scalable units called SuperPODs that offer tens of thousands of GPUs. For a very large SuperPOD with 16,834 GPUs and 2048 nodes, there would be hundreds of InfiniBand switches required (or a hyperscaler can also use Ethernet switches) and extensive cabling is also required.
Also consider that as the networking and interconnects market matures, there will be new opportunities to participate, for example, co-package optics are expected to be introduced for the Rubin generation of GPUs in 2026-2027.
Regardless of the exact networking-to-compute ratio, as we scale up AI systems, the architecture becomes a networking and interconnect problem that must continually be solved for, and we want to be correctly positioned within the networking supply chain. Therefore, AI networking will remain a key focus for the I/O Fund’s Top 10 New Ideas in the coming quarters and years. You can expect our team to deliver additional names in this trend and to increase our allocation as needed.
AI Energy:
Data center power demand is expected to grow at an accelerated clip through the end of the decade and beyond, with more powerful GPUs and surging growth in inference two main drivers.
McKinsey projects data center capacity will rise ~2.5x to 219 GW by 2030, up from 82 GW in 2025, with AI contributing 70% of that demand. This corresponds to total capacity growth of 137 GW over the next five years, with 112 GW coming from AI.
Goldman Sachs estimated global data center power usage at 55 GW in early 2025, far below BCG’s 82 GW figure. However, GS projects power usage to reach 84 GW in 2027 and increase further to 122 GW by 2030, corresponding to total growth of just 67 GW. However, considering 2025 and 2026 capex spend could support >20 GW of new capacity, this forecast may understate the pace of capacity growth.
As that analysis pointed out, Nvidia’s Blackwell lineup is bringing a significant increase in power consumption, nearly double the H200’s 70 kW at 120 kW for the GB200 NVL72 and 140 kW for the upcoming GB300 racks.
Beyond Blackwell, Nvidia’s future design lineup shows continual increases in power consumption. Its Rubin generation is expected to boost thermal design power (TDP) by 50% over Blackwell at up to 180 kW per rack, with the upgraded Vera Rubin then doubling this to 360 kW per rack by 2027.
In its largest configuration, the Vera Rubin NVL576, dubbed the ‘Kyber’ rack, could draw as much as 600 kW (0.6 MW), or 5x that of the GB200 NVL72 in just a two-year design timeframe. These figures do not include networking, interconnects, cooling and other hardware, which will further boost power draw per rack.
This rapid increase in power consumption per GPU generation is critical, as existing infrastructure is simply unable to meet these escalating power demands. For example, Applied Digital pointed out that nearly 70% of current data centers “contain racks requiring between four and nine kW of power, and less than two percent of data centers have racks with greater than 50kW.”
For comparison, Super Micro’s GB200 NVL72 SuperCluster requires 132kW, while the upcoming Kyber rack could more than quadruple that to 600kW. Because Blackwell-based servers are now 15x to 30x the power density, cooling and power delivery strategies have to be redesigned, as liquid cooling now becomes a necessity.
This sharp rise in power density means current infrastructure may be unable to transition from 4-9kW racks to >130kW racks without incurring significant retrofitting costs, while building new infrastructure bypasses that hurdle and allows for optimization for high-powered racks.
For example, Vantage’s upcoming 1.4 GW campus in Texas for Oracle is designed for ultra-high density racks up to 250kW, yet this will not be enough power to able to host NVL576 racks in just two to three years’ time. Additionally, a former Microsoft Azure AI executive reportedly said he estimated that the “requirements in terms of power for the data center would probably at least double every three years and maybe exponentially so over a period of time,” further reinforcing this.
The push to 600kW racks over the next few years means this is not a transient problem, rather it is one that the industry will continue to face, meaning continuous new construction may be needed to handle surging power demand.
In the latest earnings call with CoreWeave, management agreed with the analysis we have presented to Pro Members, stating: “at the end of the day, right now, it's the powered shells that are the choke point that is causing the struggle to get enough infrastructure online for the demand signals that we are seeing […]”
Takeaway: When building our portfolio, we have to balance many things when it comes to AI energy stocks. Time to power is paramount as some energy infrastructure is 5-10 years out from commencing operations and generating revenue. Secondly, many energy stocks require significant cash to secure energy sources, get government approval and build the underlying infrastructure, which also includes taking into consideration regional differences, transmission capacity and distances from generation sites to metro areas.
AI Data Layer:
The initial years of AI development were compute intensive to where training created a compute hierarchy. We still see evidence of this hierarchy as companies with access to GPUs, networking and energy have an advantage, and the barrier to entry is high in both costs and by the limited supply preventing widespread access.
Nvidia has enjoyed a near-monopoly by being the parallel processing leader decades before AI took over as the primary market that demanded massive compute and data throughput for matrix multiplications and vector math. The lack of supply has afforded the world’s largest companies a head start in training and deploying models while startups and enterprises patiently wait for access. As more frontier LLMs are deployed by R&D labs and Big Tech, the emphasis moving forward will be on inference rather than only on training models.
The inference market is when enterprises and companies sitting on large private data sets will be able to increase the accuracy of open sourced models and licensed models. There will be an important shift to where companies that can offer domain-specific data in various industries, such as finance, healthcare, manufacturing, will do well by optimizing processes to generate more revenue and achieve better margins. It will start with the Fortune 500, the Global Fortune 2000 and well-funded startups. Meanwhile, investors should also not overlook the fact that R&D labs are growing closer to cracking the consumer market, as well, with apps such as OpenAI’s ChatGPT and Sora, or Perplexity’s search.
While training is benefiting those who sell the compute or own the infrastructure (and we will continue to own these stocks), there will also be a shift toward companies that manage the data pipes by sitting across the many database and software services that enterprises use. Think of all the ERP systems, CRMs, legacy databases, etc, where private data is stored. There will be an emphasis toward combining the data, keeping it private, yet utilizing it to increase the quality of inference.
Takeaway: Compute will continue to drive the scale for inference; data will drive the quality of inference. Therefore, a key focus for I/O Fund’s Top 10 New Ideas list over the next few quarters (and years ahead) will be AI data stocks that help private enterprises use their valuable data to feed data-hungry reasoning models. We will also, in tandem with AI networking and AI energy, be looking to build our portfolio with exposure to stocks that will participate in the AI inference market, which spans hardware, software, the data layer, and more.
The stocks below are new ideas and at time of publishing, the I/O Fund does not own these stocks although they are under strong consideration for the portfolio. To find out the stocks the I/O Fund owns, subscribe to our Pro tier for Research or our flagship tier Advanced with additional research, real-time trade alerts, allocations to stocks and weekly webinars.
CoreWeave: $20B+ in New Deals, Targeting Aggressive Power Growth
CoreWeave brands itself as the world’s first “AI hyperscaler” as they offer both infrastructure and a software platform for developing large language models and deploying them. Being dubbed an AI infrastructure player means CoreWeave must offer a compelling value proposition to attract business from arguably the largest competitors in the world – AWS, Microsoft Azure and Google Cloud.
One of their primary value propositions is offering bare metal servers, as the company does not need to offer shared GPUs like the hyperscalers. By stripping away the virtualization layer, raw performance goes up for R&D labs, who do not need to want to lower performance for workload flexibility. Although CoreWeave offers shared infrastructure in terms of storage and networking, the company’s key differentiation from the Big 3 is by offering dedicated bare-metal access. CoreWeave also offers proprietary software to help achieve higher total system performance and more favorable uptime relative to competitors.
CoreWeave has already reached a $5 billion run rate with 470MW (~20%) of its 2.2GW contracted power active and operational, leaving an additional ~1.73GW to be developed. The company is planning to have 900MW active by year-end, hence the need for high capex to support growth. At full-scale, the company may be able to support a $25 billion revenue run rate, aligning with FY29’s revenue estimate.
Overall Revenue Growth Overall Revenue Growth
CoreWeave reached a new milestone of over $1.0 billion in revenue in Q2 2025, growing 206.7% YoY and 23.6% QoQ to $1.21 billion, driven by strong demand for the company’s AI cloud infrastructure services. Revenue growth is expected to be strong in the coming quarters, driven by the robust demand due to training and inference workloads.
Management revenue guidance for Q3 is in the range of $1.26 billion to $1.30 billion, representing YoY growth of 119.2% and 5.5% QoQ at the midpoint. Revenue growth is expected to show a 20% acceleration QoQ in Q4 with revenue growing 139.5% YoY and a further 16 percent acceleration QoQ in Q1 2026, highlighting large deals signed in the recent quarters.
Looking forward, revenue is expected to grow by 174% YoY to $5.26 billion in the year 2025 and 129.6% YoY to $12.08 billion in 2026.
Key Metric Key Metric
CoreWeave reported backlog of $30.1 billion at the end of Q2, up 86% YoY driven by the company’s strategic deal with OpenAI in March 2025 and a subsequent expansion deal in May. CoreWeave expects 50% of this backlog, or ~$15 billion, to convert to revenue over the next 24 months, providing a strong degree of visibility into future growth.
However, CoreWeave’s backlog has now likely surpassed $50 billion, considering the company signed an additional expansion deal with OpenAI worth $6.5 billion and a large-scale deal with Meta worth $14.2 billion, both lasting through 2031.
Earnings Earnings
CoreWeave reported GAAP loss per share of (-$0.60) in Q2 compared to the consensus estimate of (-$0.49), with the miss stemming from higher operating expenses, particularly technology and infrastructure expenses.
CoreWeave is not expected to shift to GAAP profitability until FY27: analysts expect GAAP loss per share of (-$2.67) for this year, followed by (-$0.90) for 2026, before shifting to positive GAAP EPS of $1.59 in 2027.
Margins Margins
While gross margin has expanded over the last two quarters, operating margin remains thin from heightened expenditures and aggressive investments in data center capacity and servers.
Gross margin was 51.2%, up from 50.7% in Q1 and 43% in Q4. This is now slightly below the highest gross margin that CoreWeave has reported publicly at 53.7%.
Operating margin was just 1.6% in Q2, inflecting from a (2.8%) margin in Q1 but substantially lower than the 20% margin in the year ago quarter as tech and infrastructure expenses have surged more than 260% YoY.
Net margin was (24.0%) in Q2, marking a slight improvement from (32.1%) in Q1, with this pressured heavily by high interest payments on debt.
Cash Cash
CoreWeave’s business model is based on aggressive capacity expansion, currently fueled primarily by debt. As a result, cash is rather thin and gets spent quickly, and free cash flow is widely negative.
CoreWeave reported $1.15 billion in cash and equivalents (excluding $0.56 billion in restricted cash and equivalents), though CoreWeave updated in an 8-K related to its now upsized $1.75 billion raise that total cash will be closer to $5 billion.
Operating cash flow was ($251.3 million) for a (21%) margin in Q2, widening from ($117.8 million) in the year ago quarter. Free cash flow was ($2.7 billion) for a (223%) margin, widening slightly from ($2.36 billion) in the year ago quarter.
Debt was reported at $11.05 billion in Q2, with $3.62 billion being current. Current debt is likely closer to $12 billion now, with a majority (~$6.7 billion) tied to its two existing delayed draw term loan facilities. If its new DDTL is drawn upon, debt could rise to $14.6 billion. On the other hand, the company is trying to reduce its cost of capital by raising cash through secured debt, using its highly valuable GPUs as collateral.
Valuation Valuation
Given its limited history on the public markets, CoreWeave’s valuation comps are limited. The company is trading above its average forward PS multiple since IPO of 9.6x, currently valued at 12.8x. This is approaching its peak forward PS multiple so far of 16.8x.
Notable Risks Notable Risks
Based on company guidance, CoreWeave’s capex for the second half of the year is expected to be >$15 billion, with cash on hand only covering one-third of that at maximum. The majority of this capex, around >$12 billion, is expected to hit in Q4 due to timing of standing up new infrastructure, placing emphasis on finding funding to finance this spending. Debt is expensive for CoreWeave, with recent raises at >9% rates, making tapping the debt markets a more expensive endeavor – interest expenses were 22% of revenue in Q2.
Customer concentration also presents a risk, though this is now beginning to ease as other anchor customers Meta and OpenAI ramp. Microsoft had accounted for 72% of revenue in 1H, with the former two providing much needed diversification away from Microsoft.
IREN: Aggressively Building an AI GPU Cloud, Targeting $500M ARR by Q1
There are a few key things that separate IREN from other Bitcoin miners. The first is that IREN is still a Bitcoin miner whereas others have retrofitted their Bitcoin operations for the data center entirely or plan to very soon as the operations were not profitable. In contrast, IREN is able to turn a profit from Bitcoin mining and plans to use that cash to help fund its AI data center expansions.
Secondly, IREN can offer a hybrid mix of both colocation and AI cloud services, which is essentially bare metal servers without a hypervisor or virtual layer. This is attractive to hyperscalers as virtualization can lead to performance loss. For the AI cloud, IREN functions similar to CoreWeave as the GPUs are leased “as a service.” For colocation, IREN provides the facility, power and cooling for customers to deploy and manage their own hardware.
IREN is working to rapidly build out its GPU fleet considering its power pipeline totals 3GW, with the firm now owning more than 23,000 GPUs consisting of primarily Blackwells, more than doubling this month. IREN does have capacity to support >660K GPUs in total, though this will likely cost upwards of $60 billion in full.
Overall Revenue Growth Overall Revenue Growth
IREN delivered Q4 revenue of $187.3 million, up 229% from the $56.8 million earned in the year-ago period and up 65% sequentially from Q3’s $113.6 million. This kind of sequential growth is rarely seen outside of hypergrowth SaaS, let alone in a miner. The bulk of Q4 revenue came from Bitcoin at $180.3 million.
IREN reported FY25 revenue of $501.0 million, up 168% year-over-year from $187.2 million, underscoring one of the fastest growth rates in the sector. Looking ahead to 2026, IREN is estimated to report 113% growth to $1.09 billion, with growth each quarter expected to be >80% or higher.
AI Segment Revenue Growth AI Segment Revenue Growth
IREN’s AI cloud business is currently a small contributor to growth, with revenue of just $7 million in fiscal Q4. While this represented growth of 180% YoY and 94% QoQ, the segment accounted for less than 4% of total revenue in the quarter. For FY25, AI cloud revenue was $16.4 million, more than 5x higher than its FY24 revenue of $3.1 million.
In accordance with its doubled fleet at 23K GPUs, IREN unveiled a new AI Cloud annualized revenue (ARR) guidance, now targeting >$500 million in ARR by Q1 2026. This is more than double its guidance from August for $200 million to $250 million by year-end 2025, based on a fleet size of 10.9K GPUs. IREN said in early October that it has secured customer contracts for the 10.9K fleet representing $225 million in ARR, and that it remains on track to reach its Q1 target.
Earnings Earnings
IREN reported Q4 GAAP EPS of $0.66, though much of this (~$0.54) was attributed to a $147.7 million gain on financial instruments.
IREN’s EPS flipped positive in FY25, with diluted EPS of $0.39 versus a ($0.29) loss in FY24, marking its first full-year profit on a per-share basis, though GAAP EPS is heavily influenced by fair-value accounting marks. Looking forward, analysts expect GAAP profitability to more than triple through FY26 to $1.23.
Margins Margins
Operating margin shifted to positive in FY25, with Q4 seeing operating margin reach double-digit territory, a stark contrast to other miner peers.
Gross margin was 71.8% in Q4, reflecting benefits from scaling mining operations and disciplined power cost management. FY25 gross margin was 68.3%, up nearly 15 points YoY and signaling more room for expansion with Q4’s print.
Operating margin was 11.0% in Q4, down from 20.1% in the prior quarter. For FY25, operating margin was 3.5%, shifting positive from a (14.6%) margin in FY24.
Net margin was 94.4% in Q4, though stripping out the substantial gain on financial instruments would place net margin at 15.6%. For FY25, net margin was 17.3%, including 15.5 points contribution from gains on financial instruments.
Cash Cash
While operating cash flows were solid, free cash flow was largely negative at more than 2.5x revenue due to elevated capex.
Operating cash flow was $245.9 million in FY25, up from just $52.2 million in FY24. This equated to a strong 48.9% margin, up from 27.9% in the prior year. However, free cash flow widened from ($427.7 million) in FY24 to ($1.13 billion) in FY25, with IREN spending $574 million on PP&E and $799 million on GPUs and hardware.
IREN reported $564.6 million in cash against $962.8 million in convertible debt in Q4. IREN also priced an additional $875 million in convertible debt in early October, which will likely go towards additional GPU purchases.
Valuation Valuation
With its strong multi-month rally, IREN is now trading at peak forward PS multiples at 16.5x, far above its historical average 3.9x forward multiple.
On the bottom line, IREN is trading at the upper end of its forward PE range over the past year at 70x forward earnings. This is notably elevated from early August’s 17x forward PE multiple.
Notable Risks Notable Risks
IREN marks the first (and perhaps only) attempt across Bitcoin miners to double up as a neocloud, which may carry substantial capital risks to self-fund GPU fleet expansion. Analysts expect IREN to grow its GPU fleet by 5x by year-end 2026, though this firm’s estimate hinged on IREN taking on $6 billion in new debt related to GPU purchases. This would quickly consume cash and possibly could cost hundreds of millions quarterly in interest, depending on terms, which would be hard to cover with cash flows.
Additionally, IREN is targeting a rather aggressive ramp in its AI cloud business, which relies on internal projects for utilization, hourly rental rates, and on-time GPU delivery, all factors that may change quite quickly. Monthly AI cloud revenue has still yet to reach an inflection point as of August, suggesting little to no new GPU deliveries have occurred since June.
Palantir: AI Platform Drives Eight-Quarter Revenue Acceleration
The difference between Palantir and other AI-enabled database competitors is that Palantir is able to answer questions a model cannot answer. Traditional business intelligence companies require a complete data set whereas Palantir is able to tackle situations where there is not a complete data set. You can think of the competitive advantage as being actionable depth, which Palantir has described as “the reasoning that goes into decision-making, not just data.”
Palantir’s Artificial Intelligence Platform (AIP) integrates generative AI with operational data and workflows, and when combined with Palantir’s other platforms Foundry and Apollo, it provides an AI service mesh that can run hundreds of microservices, scale compute through its Rubix engine and orchestrate updates through Apollo.
Additionally, Palantir’s knowledge graph referred to as Ontology is a distinct advantage. The graph offers better context than a large language model would on its own – or as Palantir states, it’s “the reasoning that goes into decision-making.”
Overall Revenue Growth Overall Revenue Growth
Palantir cracked the $1 billion quarterly revenue milestone in Q2, with revenue of $1.003 billion in the quarter. Growth accelerated nine points to 48% YoY, and seven points to 14% QoQ. For Q3, Palantir guided for growth to accelerate slightly to 50%, what would mark its ninth straight quarter with accelerating revenue growth.
Management also boosted its full-year revenue growth forecast by ~$250 million, from $3.9 billion to $4.15 billion, corresponding to growth of 45% YoY. This marks a sharp acceleration from 29% growth in FY24.
AI Segment Revenue Growth AI Segment Revenue Growth
Palantir’s US Commercial segment is the primary vector for its AIP-driven growth, with the company seeing robust momentum from the segment. US Commercial revenue grew 93% YoY (a 22 point sequential acceleration) and 20% QoQ to $306 million.
Palantir also scored US commercial remaining deal value of $2.79 billion, up 145% YoY and 20% QoQ, while US commercial total contract value (TCV) rose 222% YoY to $843 million. US commercial customers rose 64% YoY to 485.
For the full year, Palantir boosted its guide to >85% YoY growth to $1.302 billion, compared to its Q1 guide for 68% growth to $1.178 billion.
Earnings Earnings
Palantir reported adjusted EPS of $0.16 in Q2, up 78% YoY, and GAAP EPS of $0.13, both coming in ahead of estimates. For Q3, Palantir is expected to report adjusted EPS growth of 67% YoY to $0.17.
For fiscal 2025, Palantir is estimated to see adjusted EPS rise 57% YoY to $0.64, before slowing to 32% growth to $0.85 in fiscal 2026.
Margins Margins
Palantir is a standout in terms of margins, as the company continues to drive operating margin expansion while accelerating revenue growth. This helps the company’s Rule of 40 metric, which stands at 94 as it combines adjusted EBITDA margin with revenue – or more than double the ideal 40 that many SaaS companies set out to achieve yet cannot due to a lack of GAAP margins.
Gross margin was 80.8% in Q2, down from 81% in the year ago quarter but up marginally from 80.4% in Q1.
Operating margin was 26.8%, expanding significantly from 15.5% in the year ago quarter and from 20% in Q1. Adjusted operating margin was 46.3%, up from 37.4% in the year ago quarter and 44.2% in Q1; Palantir guided for continued strength in Q3 at a 45.6% margin.
Net margin was 32.6%, up nearly 13 points from 19.8% in the year ago quarter and up from 24.2% in Q1.
Cash Cash
Palantir also stands out for its ridiculously strong cash flows, with margins above 50%, putting it in rare territory for a high-growth AI stock when combined with its operating and net margin profile.
Operating cash flow was $539.3 million in Q2 for a margin of 54%, well above its 20% margin from the year ago quarter but slightly below its peak OCF margin of 58% in Q3 2024.
Adjusted free cash flow was $568.8 million for a 57% margin, again up substantially from 22% a year ago but below its peak FCF margin of 63%. Palantir raised its adjusted FCF guidance for FY25 by $200 million, from $1.6-1.8 billion to now $1.8-2.0 billion, or a margin of 45.8% at midpoint.
Cash and equivalents totaled $6.0 billion, while debt was zero.
Valuation Valuation
On the topline, Palantir trades at 105.4x forward PS, far above its 5-year average multiple of 33.4x and well above the second most expensive software stock in Cloudflare at 35.7x forward PS. Palantir is in uncharted territory as it is at its peak multiple ever sustained for this metric.
On the bottom line, Palantir trades at 288x forward PE, again at its peak and far above its average of 103.4x. On an FCF basis, the company trades at 273.3x, also above its average multiple of 154.5x.
Notable Risks Notable Risks
The valuation with Palantir is a gamble as the company is attempting to set a new bar for AI software, with >100x forward sales multiples only achieved for short fashion in 2021 for a handful of prior market darlings, whose stocks have yet to return to those prices. Palantir’s elevated valuation may also present a risk if/when the company reaches peak revenue growth, as it cannot accelerate the topline forever.
Lumentum: Cloud & Networking Growth Accelerates to 16% QoQ
Optical interconnects are a trend the I/O Fund has been tracking for more than a year, as these interconnects help data centers accelerate data throughput between and inside the data center between servers or racks, while reducing latency and power consumption. Lumentum supplies components for datacom transceivers (including VSCELs, CW lasers for silicon photonics and EML-based lasers) and optical interconnects with tech that has caught the attention of heavyweight Nvidia.
While we have been closely monitoring Lumentum and waiting patiently for their EML lasers for 200G to ship, the company’s most recent report provided confirmation of the EML laser ramp, as EMLs achieved an all-time high for shipments with revenue more than doubling versus its June 2024 baseline. Management also cited a “substantial” 200G EML order to be fulfilled in the December quarter, although offered little additional clarity on the size of the order.
Lumentum also remains quite confident in its growth opportunities from EML lasers, optical circuit switches and co-packaged optics heading into 2026. The company provided a new $600 million quarterly revenue target it expects to reach by next June or earlier, up from $424 million in the current quarter.
Overall Revenue Growth Overall Revenue Growth
Lumentum reported Q4 revenue of $480.7 million, beating analyst estimates by a modest 2.29%. This was a notable uptick on the top-line, with growth of 13.1% QoQ, accelerating 7 points, and 55.9% YoY, accelerating 42 points, beginning to support the thesis that Lumentum is past the cyclical low it experienced in FY24.
Guidance for Q1 FY26 came in at $510 to $540 million, representing 55.8% growth YoY and a slight moderation to 6% QoQ growth. Management attributes the strong forward revenue guide to surging AI workloads and the “shift toward high-speed photonics for hyperscale cloud operators.”
FY25 revenue came in at $1.65 billion, with growth rebounding to 21% YoY from FY24’s (23%) YoY decline. FY26 is expected to see revenue growth accelerate nearly 20 points to almost 40% YoY, with current estimates at $2.29 billion.
AI Segment Revenue GrowthAI Segment Revenue Growth
Lumentum’s Cloud and Networking revenue came in at $424.1 million in Q4, representing 66.5% YoY growth. Additionally, the segment’s QoQ growth of 16.1% accelerated sharply from 7.6% QoQ in Q3.
Management cited a few factors behind the outperformance in Q4: strong hyperscaler demand driving more than 50% QoQ growth in cloud modules, all-time high in EML shipments, and strong transceiver demand. Of these, EML drove the results this current quarter: “In Components, we achieved an all-time high in EML shipments, nearly doubling the revenue compared to our June quarter 2024 baseline.”
For Q1, management expects Cloud & Networking to be up QoQ, and based on guidance, this would likely correspond to approximately 10% QoQ growth at midpoint. Management indicated there was potential for strong growth starting in the upcoming December quarter: “Recently, we received a substantial order for 200-gig lane speed EML chips, which we expect to fulfill in the December quarter. Overall, we expect 2026 to be a breakout year for laser chip sales of both 100-gig and 200-gig lane speeds.”
Earnings Earnings
Lumentum reported adjusted EPS of $0.88 in Q4, beating estimates of $0.81 and improving against the $0.57 reported in Q3 and $0.06 reported in the year ago quarter. Management guided for $0.95 to $1.10 in adjusted EPS in Q1, up nearly 470% YoY, while Q2 is expected to see 176% growth to $1.16.
Looking ahead, Lumentum is expected to see adjusted EPS rise at a 77% CAGR through fiscal 2027, rising from $2.06 in FY25 to $4.86 in FY26 to $6.48 in FY27.
Margins Margins
The re-acceleration in revenue drove solid in GAAP gross margin to 33.3%, higher than the 28.8% percentage seen in Q3 and double the 16.6% margin from Q4 FY24. Non-GAAP gross reiterates this story, as Q4’s 37.8% margin continued to expand versus the prior quarter comp of 35.2% and prior year comp of 27.8%.
Lumentum showed continued progress on operating margins as well in Q4 with a (1.7%) operating margin, compared to (8.9%) in Q3 FY25 and (43.3%) in Q4 FY24. Non-GAAP operating margin of 15.0% expanded nicely compared to prior quarter of 10.8% and prior year quarter of (5.1%). Most of the profit margin is driven by the Cloud and Networking segment, which boasted a margin of 23.6% in Q4, up more than 13 points YoY.
Q4’s adjusted net income margin of 13.1% reflects continued operational improvements and the fourth consecutive quarter of sequential improvement (from 9.6% in Q3, 7.5% in Q2, and 3.6% in Q1).
Cash Cash
Cash flows have been lumpy, though Q4 saw Lumentum report the highest operating cash flow margin in the past two years. Despite this, free cash flows are pressured as Lumentum reinvests to expand manufacturing capacity.
Operating cash flow increased 80% YoY to $64 million, representing a 13.3% margin. This expanded nearly 2 points from 11.5% in the year ago quarter and was a notable uptick from (0.4%) in Q3. For FY25, operating cash flow was up ~5x to $126.4 million, for a 7.7% margin, improving from a 1.8% margin in FY24.
Free cash flow declined (7%) YoY to $10.1 million, for a 2.1% margin, down from 3.5% in the year ago quarter. As a result of elevated capex, FY25 free cash flow was ($104.7 million), for a (6.3%) margin, improving only slightly from (8.3%) in FY24.
Cash and equivalents of $877.1 million in Q4, largely in line with the $866.7 million reported in Q3. Debt remained largely consistent with the prior few quarters at $2.57 billion.
Valuation Valuation
Lumentum is trading at peak multiples on a forward PS basis, at 5.2x, nearly twice its average 5-year multiple of 2.8x, where it was valued at as recently as June.
On the bottom line, Lumentum is trading below average multiples, at 35.3x forward PE versus its 5-year average of 40x and its 2025 peak at 50x in January. However, similar to the topline, this has expanded significantly since June, when Lumentum traded at 19x forward PE.
Notable Risks Notable Risks
Customer concentration is a risk present in Lumentum, as two customers currently represent 31% of total revenue. When asked about customer concentration for specific products, such as cloud modules and OCS, management stated that due to their being capacity constrained, it was unlikely they would take on new customers.
Trading at peak multiples on the top-line also presents a risk, as the company is expected to see quarterly revenue growth peak in fiscal Q1 at 56% and then decelerate to ~40% for the next two quarters, an unfavorable position to be in with an elevated valuation.
GE Vernova: Aiming for 60GW Backlog, Supplying AI Data Centers
GE Vernova is part of the spinoff that General Electric first announced in 2021 and later completed in 2024. The company broke up its three biggest segments into separate units: GE Healthcare was spun off first in 2023, GE Vernova for the Energy business was spun off in 2024 and began trading as GEV, and GE Aerospace was the business that remained with the existing stock ticker GE. At the time the Energy segment was split up, it was seeing $33 billion in revenue and was helping to generate 30% of the world’s electricity with 55,000 wind turbines and 7,000 gas turbines.
This year, the company is expected to report $37 billion in revenue with strong earnings growth of 45.3%. The stock is not a hypergrowth profile, but rather, it is a quality, defensible position that could do well during any periods of doubt in the broader AI trend.
The defensibility is particularly attractive when you consider that gas turbines is the crux of the issue for expanding gas power plants. According to Bloomberg’s calculations, more than “$400 billion worth of gas-fired power plants through the end of the decade are in jeopardy of delay or cancellation because of the lack of capacity to meet future turbine orders.” The same article points toward GE Vernova filling orders as far out as 2030.
GE Vernova is the world’s largest gas turbine supplier at 25% ahead of Schneider at 24%. Even still, GEV nearly tripled its gas turbine equipment orders this past quarter – a statement that has us sitting up in our seats. Per the earnings call: “Power orders grew 44%, led by Gas Power equipment nearly tripling year-over-year.” In a bid to supply options quickly to alleviate bottlenecks, GEV is also shipping aeroderivative gas turbine packages and doing extensive R&D on a small modular reactor (SMR) design.
Overall Revenue Growth Overall Revenue Growth
GE Vernova is on a path to revenue acceleration in 2026, as it is a major beneficiary of the spending surge of hyperscalers stemming from the increasing energy requirements from the global AI infrastructure build-out.
The company’s Q2 revenue grew by 11% to $9.11 billion, beating estimates by 3.6%. Organically, revenue grew by 12% YoY to $9.04 billion, primarily due to higher equipment and services revenue. Analysts expect revenue to grow by 2.7% YoY in the next two quarters and revenue growth to accelerate to 8.8% in Q1 2026.
On the back of strong demand for power and equipment, management has raised its full-year revenue guidance and expects 2025 organic revenue to come at the high end of the guidance of $36 billion to $37 billion. The power segment organic revenue guide is increased to 6-7%, up from the previous single-digit guide, and the electrification segment is expected to grow 20%, up from the previous mid-high teens percentage. On a side note, the wind segment is expected to be down mid-single digits due to the more challenging market conditions.
Revenue growth is set to accelerate over the next three years. Analysts expect a 6.4% increase in 2025, bringing the total revenue to $37.17 billion, a 5.1% growth in 2024. Momentum is projected to build further, with revenue climbing to $40.7 billion in 2026, up 9.5% and to $45.5 billion in 2027, up 11.9% YoY.
Key AI Metric Key AI Metric
In Q2, GEV signed 9GW of new gas equipment contracts with 2GW going directly to orders and 7GW going into what’s called a slot reservation. During the quarter, the company also converted 3GW into orders and shipped 5GW of equipment. GEV is also witnessing robust demand for its aeroderivative technology to support data centers, securing 27 aeroderivative units in the recent quarter compared to only one in the same period last year.
Overall backlog is now up to 55GW, ahead of the 50GW guided in April, including 29GW in backlog and 25GW in slot reservation agreements (SRA), up from 21GW. GEV is expecting the backlog will reach 60 GW by the end of this year. There was a discussion on the call that this represents 3 years of backlog: “And we've talked about the fact that we'll get to at least 60 gigawatts by the end of the year. So that's directionally 3 years of backlog.” In other words, the chances that GEV is not a significant player in supplying energy to data centers for many years to come is nil.
Earnings Earnings
Q2 GAAP EPS came at $1.86, beating estimates by a solid 14.3% driven by profitable volume, better pricing, and productivity gains. Analysts expect GAAP EPS of $1.85 for Q3 2025 compared to (-$0.35) in the same period last year. They expect GAAP EPS to grow 95.4% YoY to $3.38 in Q4 and 137.4% YoY to $2.16 in Q1 2026.
Analysts continue to expect strong EPS growth in the coming years. For the year 2025, analysts expect GAAP EPS to grow 45.3% YoY to $8.11, and 58.9% and 41% YoY in the subsequent two years, reaching $18.16 in 2027.
Margins Margins
GEV has rather thin margins, with operating margin in the mid-single digits, but management expects that higher turbine prices and strong demand will lead to improved margins as the backlog approached 60GW by year-end.
Gross margin was 21% in Q2, flat YoY but up 2 points from Q1. Operating margin was 5.6%, rebounding 4.7 points QoQ but down 2.4 points YoY. Net margin was 5.6%, up 2.3 points QoQ but down 10.2 points YoY.
Adjusted EBITDA margin was 8.5%, up more than 2 points YoY, with the improvement driven by volume and price, offsetting tariff impacts and investments. While GEV did raise its full-year adjusted EBITDA margin guide to 8-9%, management was clear to say that this included approximately one point of negative EBITDA margin related to tariffs.
Cash Cash
Management raised full-year free cash flow guidance from a range of $2.0 billion to $2.5 billion to a new range of $3.0 billion to $3.5 billion, primarily driven by a higher profit outlook and increased down payments due to rising orders. Year-to-date, GEV has generated $1.17 billion in free cash flow, implying a strong acceleration of free cash flow in the second half of the year to $2.1 billion.
Q2 free cash flow of $194 million represented a sharper decline versus $821 million in the same quarter last year, though GEV had benefited from a $300 million arbitration in the comparable quarter.
GEV has $7.9 billion in cash and equivalents and no debt.
Valuation Valuation
GEV is trading near peak multiples on the top line, but there is more room on the bottom line as multiples are not nearly as stretched.
The stock is trading at approximately 4.6x forward revenue, just below its peak of nearly 4.9x and above its average 2.6x multiple, though data is limited as the spinoff has only traded publicly for a year and a half.
On the bottom line, GEV trades at 83.3x forward earnings, slightly above its average 73.9x multiple, though shares have traded in as wide a range of 36x to 137x over the past year.
Notable Risks Notable Risks
Wind is providing a notable drag on growth and earnings, with GEV stating that they have lost approximately $300 million in the segment through the first half but expect to be closer to breakeven in the second half. Q2 adjusted EBITDA for Wind widened approximately ($50 million) YoY to ($165 million). Additionally, Q3 growth was guided to be down mid-teens YoY, but excluding a one-time settlement in the year ago quarter, growth would be up low single digits.
Applied Optoelectronics (AOI) is a lesser-known optical component and transceiver supplier, though the small-cap has recently caught our attention for its ambitious capacity growth targets, aiming to become one of the largest domestic manufacturers for 800G transceivers.
The VSCEL laser, EML chip, and optical transceiver supplier has provided some of the strongest commentary for future growth, with management outlining 8.5x capacity growth by year-end and another 2x capacity growth in 2026. This builds upon a comment slipped into Q1’s call that AOI believes it can become the largest 800G/1.6T optics supplier to Amazon, after signing a deal in March that management says is worth more than $4 billion over ten years.
AOI also stood out from its 40% QoQ growth in data center revenue in Q2, though this came after a soft Q1 affected by inventory digestion at its largest hyperscale customer and seasonality. Interestingly, data center was not the core driver of growth in Q2, with cable TV taking the reign with 8x YoY growth in the quarter. However, management is eyeing 800G transceivers to begin ramping as early as late Q3, supported by that substantial capacity growth come 2026.
The crux with AOI is that cash flows are much softer than more-established AI networking stocks, and GAAP margins are still negative, though it’s hard to argue with the implications from the profound expansion in capacity and how optics growth can transform margins.
Overall Revenue Growth Overall Revenue Growth
AOI reported revenue of $102.95 million in Q2, up 138% YoY and 3% QoQ, though this was a modest (2.7%) miss versus consensus estimates. This was a slight deceleration from 146% growth in the previous quarter. Cable TV was the primary growth driver with revenue up 862% YoY to $56 million, while data center revenue rose 30% YoY to $44.8 million. Telecom and other revenue was $2.1 million, down (45%) YoY.
For Q3, AOI guided for revenue to be between $115 million to $127 million, pointing to 86% YoY growth at midpoint, though on a sequential basis this is a sharp uptick to 17% QoQ growth. Management expects that there may be some initial contribution from 800G modules late in the quarter: “Looking ahead to Q3, we expect a sequential increase in our datacenter revenue, driven by continued growth in our 100G and 400G products with the possibility of layering some additional increased 800G revenue late in the quarter.”
AI Revenue Growth AI Revenue Growth
AOI’s data center revenue rose 40% QoQ and 30% YoY to $44.8 million, though the high sequential growth rate (the highest among AI networking stocks we track) was due to a soft comp in Q1, where revenue declined (29%) QoQ on inventory digestion at one of AOI’s largest hyperscaler customers and seasonality.
>800G revenue has not yet begun to appear as AOI has not yet begun shipments, with volume only going to qualification this quarter and accounted for 1% of data center revenue at maximum. Revenue from 200G/400G products accounted for 20% of revenue, or ~$9 million, while 100G accounted for 70% of revenue, or $31.4 million.
For Q3, AOI guided for a sequential increase in data center revenue, saying that “400G is picking up so strong in Q3, Q4.” There is also the possibility that some initial 800G revenue begins layering in late in Q3, but primarily in Q4 and into 2026 as capacity expands and customers move ahead with projects.
Earnings Earnings
AOI reported a ($0.16) loss per share in Q2, missing estimates for ($0.07), as margins did not improve much sequentially. For Q3, AOI guided for ($0.03) to ($0.10), a decent sequential improvement, likely driven by the ramp in 400G.
Looking ahead, AOI is expected to shift to GAAP profitability by Q4 and expand earnings through mid-2026, with current estimates pointing to $0.03 in Q4 and rising to $0.19 by Q2 ’26, driven by the ramp of 800G.
Margins Margins
AOI is not the strongest on margins, with operating margin widening for a second consecutive quarter on increased expenditures related to customer qualification projects for 800G and 1.6T products.
AOI reported GAAP gross margin of 30.3%, up more than 8 points YoY but down 0.3 points QoQ. Adjusted gross margin was 30.4%, with management guiding this to be essentially flat next quarter at 30.3%. Management has a medium-term target of 40%, emphasizing that they will need a few quarters for higher-margin 800G and cost reduction efforts from shifting to larger wafers to be seen.
GAAP operating margin was (15.5%), a notable improvement from over (60%) in the year ago quarter, but widening from (6.5%) in Q4 and (8.9%) in Q1. Adjusted operating margin as (10.5%), down from (4.8%) in Q1 but up from (37.6%) a year ago.
Net margin was (8.8%), improving from (9.2%) in Q1 and more than (60%) a year ago. Adjusted net margin was (8.6%), down from (0.9%) in Q1 but improving from (25.1%) a year ago. For Q3, management’s guide implies adjusted net margin improving to (3.3%) at midpoint.
Cash Cash
Cash flows are also quite weak, though this has been mostly from surging accounts receivables and capex to support capacity expansion, a solid signal for the upcoming 800G ramp in conjunction with management’s commentary.
Operating cash flow in Q2 was ($65.5 million), widening from ($50.9 million) in Q1 as inventories and accounts receivables surged, up $36 million and $40 million QoQ respectively. OCF margin was (63.6%), down from (4.6%) a year ago and (51%) in Q1.
Free cash flow was ($104.3 million), widening from ($87.2 million) in Q1. Capex for the quarter was $38.8 million as AOI is quickly purchasing equipment to meet its aggressive capacity expansion targets for year-end and mid-year 2026.
Cash totaled $87.2 million versus $188.2 million in debt, with AOI recently completing its ATM program in Q2, raising $98 million net cash that will help support its capacity expansion and R&D.
Valuation Valuation
AOI is trading a 4.2x forward revenue multiple, well below its 8x peak at the end of 2024 but also far above its low of 1.2x in April, emphasizing the volatility that stems from its small-cap profile with strong growth and weak cash.
On the bottom line, shares are trading at 37.5x estimated EPS of $0.85 in 2026, with this expected to be the company’s first year of GAAP profitability since 2017.
Notable Risks Notable Risks
The primary risks with AOI stem from its margin profile as it is much weaker than more established networking players as well as the hypergrowth players we track. While there is an expectation for margins and EPS to improve through 2026 as 800G/1.6T products ramp, this will need to be proven over the coming few quarters. The weak cash flows also present a risk if there is a prolonged recovery, given the thinner cash position AOI has.
Micron: HBM, LP Server DRAM Driving Strong Growth
Thematic: 9/10
Fundamentals: 6/10
Valuation: 7/10
Brief Overview: Brief Overview:
Micron is a primary beneficiary of rapidly increasing dollar content of high-bandwidth memory (HBM) chips with each new generation of GPUs. AI training and inference rely heavily on HBM for the massive memory bandwidth that complex models require. AI servers also use more DRAM and NAND than a traditional server. These are reasons that Micron’s cyclical fundamentals could become more secular as the AI economy is built out.
In fiscal 2025, Micron’s HBM, high-capacity dual in-line memory modules (DIMMs) and low-power (LP) server DRAM revenue reached $10 billion, up more than fivefold from the prior year, while HBM alone reached $2 billion in revenue in Q4.
Management expects robust AI server demand, the shift to HBM4 and tight DRAM supply to remain tailwinds to growth and profitability moving through 2026: “we expect healthy demand supply environment in 2026 for overall DRAM, and that bodes well for profitability of DRAM, profitability of HBM and of course, profitability of non-HBM as well, which is experiencing tight supply.”
What Micron will need to answer is if the cyclical nature of the memory market will smooth out as the dollar content of memory is rapidly increasing. Data center is already proving to be a strong driver of growth for Micron, accounting for 56% of sales in FY25, up from 35% in FY24. On a dollar basis, data center revenue surged 137% YoY to $20.75 billion.
Overall Revenue Growth Overall Revenue Growth
Micron reported record Q4 revenue of $11.32 billion, driven by DRAM products (and within that HBM) with revenue up 27% QoQ to $9 billion. Growth accelerated nearly 10 points sequentially to 46% YoY, and on a sequential basis, growth was 22% QoQ, a six point acceleration. Micron guided to a fresh record in Q1 at $12.5 billion at midpoint, pointing to 44% YoY growth and 9% QoQ.
For FY25, revenue rose 49% YoY to $37.38 billion, driven primarily by DRAM and HBM revenue, which rose more than 62% YoY to $28.58 billion. HBM reached an annualized run rate of $8 billion in Q4, with HBM share to grow again in Q1 and HBM4 capacity in discussions to be sold out for calendar 2026. Micron has not provided a full-year guide for revenue, but current consensus estimates call for 43% growth to $53.5 billion in revenue.
AI Revenue Growth AI Revenue Growth
In fiscal 2025, Micron's data center reached a record 56% of company revenue, with growth primarily driven by DRAM products and aided by data center SSDs and NAND components. Overall, data center revenue increased 137% YoY to $20.75 billion.
Micron’s Cloud Memory Business Unit (CMBU), which consists of its HBM, high-capacity dual in-line memory modules (DIMMs), and low-power server DRAM solutions, saw revenue surge 257% YoY in fiscal 2025 to $13.57 billion, or YoY growth of nearly $10 billion. Micron’s other data center unit, its Core Data Center Business Unit (CDBU), consists primarily of data center SSDs and NAND components. This unit saw revenue growth of 45% YoY to $7.23 billion.
For Q4, CMBU revenue rose 214% YoY to $4.54 billion, while CDBU revenue declined (23%) YoY. CMBU revenue growth was driven by HBM and strong bit shipment growth, though Micron offered no commentary behind the decline for CDBU.
Earnings Earnings
Micron is expected to see earnings double this fiscal year as margins have swiftly recovered from late 2023 and early 2024. In Q4, Micron reported adjusted EPS of $3.04, up 157% YoY and beating estimates by 6%.
For Q1, Micron guided for adjusted EPS to be $3.75, +/- $0.15, more than 23% ahead of guidance and corresponding to YoY growth of 110%. Earnings growth is expected to reaccelerate to 155% in Q2 but then decelerate to 126% in Q3. For the full year, Micron is expected to see 100% YoY growth to $16.63.
Margins Margins
Micron’s margin turnaround story has been impressive, with gross margin up more than 55 points over the last two years and operating margin up more than 66 points.
GAAP gross margin in Q4 was 44.7%, up 7 points QoQ and 9.4 points YoY, aided by strong growth in CMBU which carried a 59% gross margin in the quarter, DRAM pricing and favorable product mix. For Q1, gross margin was guided to be 50.5% at midpoint, a nearly 6 point sequential expansion and up more than 12 points YoY.
Operating margin was 32.3%, up 9 points QoQ and 12.7 points YoY, again aided by CMBU which carried a 48% margin in the fourth quarter. For Q1, Micron expects operating margin to be 38.6%, up more than 6 points QoQ and more than 13.5 points YoY, signaling some tailwinds from operating leverage from CMBU.
Net margin was 28.3% in Q4, up 8 points QoQ and nearly 17 points YoY. The trajectory within gross and operating margins suggests that there is a path for net margin to potentially expand to the mid-30% range in FY26.
Cash Cash
Operating cash flow was $5.73 billion for Q4, up 68% YoY and more than 24% QoQ. OCF margin was 50.6%, up 1 point from Q3 and up 6.7 points YoY. For the year, Micron generated operating cash flow of $17.53 billion, more than doubling from $8.51 billion in fiscal 2024, with OCF margin expanding 13 points to 46.9%.
Adjusted free cash flow was $801 million in Q4, shrinking from $1.95 billion in Q3 on surging capex. Adjusted FCF margin was 7.1%, up from 4.2% in the year ago quarter but down from 21% in Q3.
Micron reported total cash and equivalents of $11.9 billion and total debt of $14.6 billion.
Valuation Valuation
Despite its recent rally, Micron trades at reasonable multiples, well below its peak from 2024. On the top line, Micron trades at 4x forward revenue, 10% above its average 3.6x multiple and far below its peak of 6.8x from mid 2024.
On the bottom line, Micron trades at 11.6x forward earnings, though its 43.9x average is skewed higher by 2024’s >100x multiples when margins were razor-thin. Since the start of 2025, Micron has traded as high as 16x forward earnings and as low as 8x.
Notable Risks Notable Risks
Micron’s growth to this point and beyond has been centered around HBM, both on the top and bottom lines. CMBU is the only unit that sees operating margins above the corporate total, at 48% versus 25%, 29% and 20% for its other segments, meaning that future operating margin expansion will be tied solely to growth from CMBU, and felt more the faster CMBU grows. This may mean that margin and earnings upside in 2026 may be limited come 2027.
Talen: $18 Billion, 17-Year Nuclear Deal With Amazon
Talen is an independent power producer with its power assets primarily located in the PJM region, with more than 10GW of generation capacity with 2.2GW nuclear. With assets primarily located in Pennsylvania, Maryland and now Ohio, Talen has exposure to growing data center regions, having already locked in a long-term power agreement with Amazon to fuel data centers in Pennsylvania.
Talen’s deal with Amazon not only locked in substantial power generation for Amazon’s data center assets in a time the industry is facing a power crunch, but also locked in substantial revenue and free cash flow generation for Talen. The deal is expected to add a visible 50% uplift in cash flows as the deal ramps into full capacity by 2032, while providing a repeatable colocation model for Talen’s other assets to meet rising data center demand in the PJM region.
Talen is also expanding its power production portfolio with recent acquisitions of the Freedom Energy Center and Guernsey Power Station for ~$3.8 billion gross, expected to add 2.8 GW of combined-cycle natural gas generation capacity in the PJM region. The two plants are suitable for hyperscale data center supply, and are also expected to be immediately accretive to cash flows.
Overall Revenue Growth Overall Revenue Growth
Talen’s Q2 revenue rose 29% YoY and 62% QoQ to $630 million, positively impacted by a $176 million gain on derivatives (compared to a $76 million gain in the year ago quarter). Revenue from contracts with customers was $409 million, up 18% YoY but down (37%) QoQ. This shows the quarterly fluctuations that can stem from commodity contract hedging. For the first half of the year, Talen’s revenue was $1.02 billion, up 2% YoY, though revenue from customer contracts was $1.06 billion, up 40% YoY.
Looking ahead, Talen has some visibility into 2026 revenue from the PJM auction in July, where Talen cleared 6.7 GW of capacity translating to ~$805 million in capacity revenue for the 2026-27 planning year lasting June 2026 through May 2027. Talen expects 2026 capacity revenue of ~$747 million, up ~124% from $333 million projected in 2025. For additional perspective, capacity revenues were just $88 million in Q2, or ~22% of total contracted revenue with customers.
Key AI Metric Key AI Metric
Talen’s AI ties are to Amazon, having expanded and signed a 17-year power purchase agreement through 2042 with the hyperscaler to power its data center adjacent to the Susquehanna plant and potentially other facilities in the Pennsylvania region.
Talen signed the 1.92 GW agreement worth $18 billion in June, and expects to deliver the first 240MW by mid-2026, scaling to 360MW by mid-2027 and 480MW by mid-2028. Full volume is expected to be reached no later than 2032 with potential to accelerate this timeline, with price escalators through 2042.
Though Talen has not provided a view into how it believes revenue will ramp, at face value the deal is worth more than $1 billion in average annual revenue, though it will take up to seven years to reach full volume. Talen has provided a glimpse into how the deal will accrete to FCF in the ramp stage. The company expects $1.55 in FCF from Amazon in fiscal 2026, before rising to $4.00 to $5.75 by fiscal 2029 (with 840-1,200MW delivered), and to $7.00 to $8.25 by fiscal 2032 (1,600MW to full capacity).
Earnings Earnings
Talen reported GAAP EPS of $1.50 in Q2, rebounding from ($2.94) in Q4 as sharper losses on commodity contracts ate into revenue. This is also not comparable to the year ago quarter where Talen reported $7.60 in EPS, as this included an approximate $9.40/share benefit from the sale of its Cumulus data center to Amazon.
For fiscal 2025, Talen is expected to report $5.39 in GAAP EPS, signaling a strong second half of the year, with 2026 EPS projected to surge 285% to $20.74.
Margins Margins
Talen’s gross margin (operating revenues including derivatives minus energy expenses) was 60% in Q2, down from 63.3% in the year ago quarter. Operating margin was 10.5%, up from 5.5% in the year ago quarter and (27.2%) in Q1, driven by the adverse revenue impact from derivate losses.
Net margin was 11.4% in Q2, which does not compare to the asset-sale impacted margin from the year ago quarter of 92.8%. Q1’s net margin was (34.6%), dragged down by derivatives.
Cash Cash
Talen’s operating cash flow was ($184 million) in Q2 for a (29.2%) margin, bringing 1H operating cash flow to ($65 million) for a (6.4%) margin, down from $150 million for a 15% margin in the year ago period.
Adjusted FCF was ($78 million) in Q2 for a (7.6%) margin, with 1H adjusted FCF just $9 million. Talen maintained its guidance for $450 to $540 million in adjusted FCF for the year, implying a significantly stronger second half of the year.
For fiscal 2026, Talen is guiding to $980 million to $1.18 billion in adjusted FCF, more than doubling YoY, and for fiscal 2027, adjusted FCF of $1.055 billion to $1.425 billion. This would represent over 250% growth in just two years.
Talen’s cash balance is extremely thin at $122 million versus its reported debt at $2.97 billion in Q2, though pro-forma debt is actually much higher at $6.56 billion, including $3.8 billion for the acquisitions of the Freedom and Guernsey natural gas plants to add >2.8GW capacity.
Valuation Valuation
Talen is trading at peak multiples on the top-line at 8.6x forward revenue, more than 2x its average multiple of 3.6x. On the bottom-line, Talen trades at 80.9x estimated FY25 EPS, at peak levels, but for FY26 EPS, Talen trades at a more reasonable 22.2x multiple, far below its peak 1-year forward multiple of nearly 41x. On an adjusted FCF basis, Talen trades at 18.8x FY26’s adjusted FCF per share guidance of $23.60 at midpoint, versus 43.1x FY25’s adjusted FCF of $10.30 at midpoint.
Notable Risks Notable Risks
Talen has a thin balance sheet with substantial debt, and has launched multiple financings via term loans, revolving credit facilities and senior notes to fund its recent acquisitions, though it still may need more cash over the next few quarters. The Amazon deal does de-risk the future growth story by locking in substantial revenue and free cash flow growth, but not for its immediate-term cash needs. Talen had emerged from Chapter 11 bankruptcy in 2023 with a significantly reduced balance sheet, though its cash balance has returned to extremely thin levels; the Amazon deal beckons a strong ramp in FCF that may be able to pad the balance sheet in the future.
Nebius: Microsoft Deal Worth up to $19.4 Billion Supports Hypergrowth Phase
Nebius is similar to CoreWeave, dubbing itself as AI native cloud infrastructure, which means the infrastructure was built specifically for AI workloads with architectures built on bare metal servers instead of hypervisor layers, for example. This approach along with a few other optimizations can result in significantly faster training.
Nebius offers an Nvidia-optimized cloud platform for teams that need to adjust compute resources, want high-performance storage and an easy-to-use AI environment yet do not want to manage the infrastructure or AI operations. The stock surged earlier this month off the announcement of a mega deal with Microsoft worth up to $19.4 billion. This deal signals just how supply constrained the market is, given Microsoft is willing to partner with a neocloud to scale quickly.
Given Nebius’ roots are from Yandex, one might question if Nebius is truly AI-native as some of the Finnish data center in Europe was built in 2014. However, that is also where one of Nebius’ strength lies, which is the company offers European data centers which helps a company like Microsoft to expand quickly overseas.
To serve the AI-native market, Nebius has been building out colocation sites to increase capacity and lower latency for the incoming inference market. Companies like Cloudflare and Shopify are early customers, both of which need to power inference at the edge. The company is also expanding beyond Europe with data center expansions in New Jersey and Kansas.
Overall Revenue Growth Overall Revenue Growth
Nebius reported a slight beat in Q2 with revenue of $105.1 million, up 625% YoY and 106% QoQ, versus estimates for $101.2 million. AI cloud infrastructure revenue rose more than 9x YoY in the quarter, fueled by strong demand for Hopper GPUs and almost peak GPU utilization.
Looking ahead, Q2 is marking an inflection point for revenue, with growth accelerating sequentially on a dollar basis. Q3 is estimated to see revenue up more than $52 million QoQ to $157.1 million, before rising another $106 million QoQ to $263.8 million in Q4. For FY25, Nebius has maintained its guidance for $450 million to $630 million in revenue, excluding $50 million to $70 million in revenue attributed to one of its subsidiaries, Toloka.
Nebius is one of the more unique names in the AI universe as one of the few, if not only, AI-focused stock to have multiple years of triple-digit revenue growth ahead. Nebius had already laid out expectations for a meaningful acceleration next year prior to the Microsoft deal, based on timing of capacity ramp in Q4 2025 and throughout 2026. The recent deal with Microsoft is extending this acceleration, with revenue revised more than $3 billion higher by 2029, or 24% to 65% above prior expectations.
Prior to the Microsoft deal, Nebius was expected to surpass $5 billion in annual revenue in 2029, doubling from $2.5 billion in 2027. After the deal, revenue estimates have been raised 24% to $3.1 billion in 2027, and 65% higher to $8.44 billion by 2029.
AI Revenue Growth AI Revenue Growth
In Q2, Nebius boosted its AI Cloud annualized run rate guidance by 14%, supported by its data center expansion, more power coming online in 2H and the rollout of more Blackwell and soon Blackwell Ultra GPUs. To note, the increased ARR guidance preceded the Microsoft deal, although it may be unlikely that ARR guidance will be raised much this year considering that capacity roll-out for Microsoft is more geared towards 2026.
Q2 ARR rose ~73% QoQ and 438% YoY to $439 million, marking a significant two-quarter expansion from $90 million in Q4. Nebius is targeting ARR to more than double over the next two quarters, raising its guidance to $900 million to $1.1 billion, up from its prior view for $750 million to $1 billion.
Nebius explained that the “increase in our ARR guidance reflects the strong demand we are seeing and the expected delivery of additional GPU capacity later this year, particularly the Blackwell Ultras. Because much of this capacity will come online by the end of the year, the impact will show up more in ARR than within the year’s revenue.” Much of the ARR growth is likely to be weighted towards Q4, considering the majority of GPU installations will take place that quarter, such as the ~4,000 Blackwell Ultras in the UK.
Earnings Earnings
Nebius is expected to report negative EPS through all of fiscal 2025 and 2026, with no clear indication yet of when the company could or will shift to profitability. Current estimates show Nebius losing ($1.47) in 2025 and minimal improvement to ($1.34) in 2026.
There are no analyst estimates beyond 2026, but given the capital intensity of greenfield data center development, continuous scaling of capacity, not to mention potential investments into the autonomous driving unit Avride, Nebius may face a long road to profitability.
Margins Margins
While gross margin is expanding rapidly, up from 26% in Q4 to over 70% in Q2, operating margins remain deeply negative as due to the high costs of aggressively expanding data center capacity and GPUs on hand.
Excluding Toloka’s contribution, its subsidiary that was deconsolidated in Q2, gross margin was 71.3%, up nearly 20 points from 51.5% in Q1 and improving nearly 25 points from 46.9% in the year ago quarter. Nebius is now slightly below CoreWeave’s Q2 gross margin of 74.2%, and it remains to be seen how much further upside there is to gross margin as new capacity comes online.
Q2 operating margin was (105.8%), a notable improvement from (236.4%) in Q1 and (773.8%) in the year-ago quarter. Nebius is exhibiting initial signs of operating leverage, with total operating expenses up just 71% YoY versus the 625% YoY growth in revenue; this was driven mainly by 560% YoY growth in depreciation from increased server capacity, as SG&A decreased (10%) YoY.
GAAP net margin was 556%, as Nebius benefited from a one-time $597.4 million gain from Toloka. Adjusted net margin offers a clearer view of Nebius’ trajectory, coming in at (87.1%) in Q2, up from (164.4%) in Q1 and (424.8%) in the year ago quarter.
Cash Cash
Aggressive capacity expansion plans and a 33% increase in FY25 capex expectations from $1.5 billion to $2 billion, or nearly 4x expected revenue for the year, mean Nebius is quickly burning through cash.
Operating cash flow was ($167.7) million in Q2, for a (159.6%) margin, improving from a (357.7%) margin in Q1. This represented just a $30 million sequential improvement from ($197.8) million, suggesting that the path to positive cash flows may be prolonged. Free cash flow was ($678.3) million in Q2 for a (645.4%) margin, versus ($741.8 million) for a (1341.4%) margin in Q1.
Capex was $510.6 million in Q2, or nearly 5x of revenue, down slightly from $544 million in Q1. 1H capex surpassed $1.05 billion, meaning $0.5 billion to $1 billion is on deck for 2H.
Cash and equivalents totaled $1.68 billion, up only slightly from $1.45 billion in Q1 as Nebius deployed much of June’s $1 billion raise to capex and operations. Including the recent $2.75 billion raise and $1 billion share sale, cash is likely around or above $5 billion. At current burn rates, Nebius would have nearly 10 quarters of cash, but it is likely that capex will accelerate (think of CoreWeave as a leading indicator) to support growth in power and capacity. Debt was $0.98 billion as of Q2, not including the recently priced $2.75 billion in notes.
Valuation Valuation
Nebius also trades at a substantial premium to CoreWeave despite lagging its competitor in terms of active power, contracted power and revenue scale. Based on FY26’s revenue estimate, Nebius trades at a forward 18.2x multiple, more than triple CoreWeave’s 5.9x multiple. However, given revenue growth is expected to be triple-digits through 2028, the market is pricing Nebius to quickly grow into these multiples.
Notable Risks Notable Risks
Nebius is even more complicated than CoreWeave given they also own a capital-intensive autonomous driving division, among other investments, and was formally the company Yandex. Nebius is high-risk given its success depends on how much capital the company can raise, and the likelihood it remains in CoreWeave’s shadow is high. Regardless of how Nebius competes with CoreWeave, it remains an AI bubble stock as the company has to hope the stock prices goes up to raise cash, which will dilute shareholders (or raise debt). It’s a vicious cycle as during any months/quarters that AI stocks are soft, Nebius stock will carry outsized execution risk.
Cloudflare: Multi-Faceted AI Positioning, Steady Growth
Thematic: 9/10
Fundamentals: 5/10
Valuation: 1/10
Brief Overview: Brief Overview:
There is a quiet strength in Cloudflare’s fundamentals and key metrics. For example, Cloudflare passed a $2B run rate for the first time, signed their first $100M deal, dollar-based net retention (DBNRR) seems to have bottomed along with a slight 1.3% acceleration in revenue. Regarding the bottom line, Cloudflare is certainly stronger than many cloud peers yet tends to walk a razor’s edge due to capex.
Cloudflare references its business units as “Acts” – Act 1, Act 2 and Act 3. The company defines Act 1 as application security, Act 2 as Zero Trust and Act 3 as the Workers Platform. For our purposes as stock investors, it’s Act 3 we are most interested in.
Regarding AI inference and the Workers Platform, management connected some important dots on the earnings call as to why agentic AI will drive forward the massive inference trend. The I/O Fund team recently dug up a stat inference is expected to account for 60% to 70% of AI workloads by 2030. In particular, Cloudflare emphasizes their position is what will help the company win this market: “The fact that we sit in front of so much of the web and that more than half of our dynamic traffic is already between APIs means that we are strategically positioned to deliver the agentic web of the future.”
Cloudflare also introduced Act 4 – a new product that will help AI search engines connect with (and potentially) pay publishers for using derivatives of their copyrighted works. Although the amount of demand for this and exactly how Cloudflare will monetize this new product is not clear, it is interesting management feels confident enough to call the new use case its fourth act.
Overall Revenue Growth Overall Revenue Growth
Cloudflare reported its largest revenue beat in the last six quarters at 2.1% above consensus, with Q2 revenue up 27.8% YoY to $512.3 million. This also marked a slight 1.3 point acceleration on the top-line from 26.5% growth in Q1.
For Q3, Cloudflare guided for revenue of $543.5 to $544.5 million, ahead of estimates at the time for $538.9 million. This corresponds to a slight deceleration to the mid-to-high 26% YoY growth range, where Cloudflare is expected to remain through Q4. This provides no clear indication yet that the company is able to drive a sustained revenue acceleration aided by AI. However, consensus estimates are now above the high-end of management’s guidance at $544.9 million, essentially already pricing in stronger momentum fueling a beat for Q3.
For the full-year, Cloudflare raised its outlook to $2,113.5 million to $2,115.5 million, for YoY growth of 26.7%. This is a $22.5 million increase at midpoint from Cloudflare’s prior outlook for $2,090 million to $2,094 million for growth of 25.3%.
Key AI Metric Key AI Metric
Although management has been optimistic about AI driving a re-acceleration on the top-line, Cloudflare has not broken out AI revenue or contribution to growth. Other key metrics have remained strong in Q2.
RPO increased 39% YoY and 6% QoQ to $1.98 billion. Current RPO accounted for 66% of total RPO, or ~$1.30 billion, increasing 33% YoY in Q2, a four point acceleration from 29% growth in Q1.This is also a notable uplift from 26% growth in the year ago quarter. Billings also increased 33% YoY to $559.2 million, a third straight quarter with growth above 30% YoY.
Paying customers increased 27.5% YoY to 267,929 in Q2, the second quarter in a row with 27%+ growth. This is a notable improvement from 17% and 21% growth in Q1 and Q2 2024. Cloudflare stated that it added a record number of customers YoY spending over $1M and over $5M.
Earnings Earnings
Cloudflare topped estimates in Q2 driven by the revenue beat and stronger adjusted margins, and boosted its FY25 adjusted EPS outlook as a result. GAAP EPS was ($0.15), missing estimates for ($0.08) as GAAP margins drifted lower. Adjusted EPS was $0.21, beating estimates for $0.18, fueled the outperformance in adjusted operating margin in the quarter.
For Q3, Cloudflare guided for $0.23 in adjusted EPS, a slight uptick sequentially, while for FY25, the company raised its forecast from $0.79-$0.80 to $0.85-$0.86. This corresponds to growth of ~14.5% YoY, up from the mid-6% range previously. Growth is expected to be much stronger in FY26 at ~30% YoY to $1.12.
Margins Margins
Gross margins drifted lower in Q2, driven by both an increase in depreciation expenses and in allocated costs from higher network traffic from paying customers. GAAP operating margins followed, moving further away from reaching break-even.
Cloudflare had an interesting comment on long-term margins, stating that it expects to remain comfortably in its 75% to 77% adjusted gross margin target despite passing on substantial savings to Workers’ customers. This suggests that upside to operating margins will be driven by expenditures, such as moderating higher sales & marketing spending, at 36% of revenue versus its target range of 27-29%, and high SBC at 24% of revenue.
GAAP gross margin was 74.9% in Q2, down nearly 3 points YoY and 1 point QoQ. Adjusted gross margin was 76.3%, down 2.7 points YoY and 0.8 points QoQ.
GAAP operating margin was (13.1%), down 4.4 points YoY and 2 points QoQ. Adjusted operating margin was 14.1%, approximately flat YoY and up 2.4 points QoQ; this was also ahead of guidance for 12.6%. For Q3, Cloudflare guided for adjusted operating income of $75-76 million, pointing to adjusted operating margin of 13.9%, down nearly 1 point YoY and moderating slightly QoQ.
GAAP net margin was (9.8%), down 6 points YoY and 1.8 points QoQ. Adjusted net margin was 14.7%, down 2.6 points YoY but up 2.5 points QoQ.
Cash Cash
Cash flow margins contracted sequentially, while Cloudflare significantly bolstered its cash pile after a large convertible note issuance.
Operating cash flow was $99.8 million for a 19% margin, flat YoY but down from a 30% margin in Q1. Free cash flow was $33.3 million for a 6% margin, down 4 points YoY and 5 points QoQ. Network capex was 11% of revenue in Q2, down from 17% of revenue in Q2. Cloudflare stuck to its guidance for network capex to be 12-13% of revenue for the year, suggesting slight moderation in 2H.
In June, Cloudflare raised $1.97 billion in new convertible notes due 2030, raising its cash on hand to $3.96 billion while convertible notes outstanding rose to $3.26 billion.
Valuation Valuation
Cloudflare is trading at peak multiples, and is the second-most expensive name in the software universe behind Palantir. Cloudflare trades at 36.2x forward revenue, slightly below its late September peak of 37.6x but nearly double its average 19.7x multiple.
On the bottom-line, Cloudflare is not yet GAAP profitable, but on an adjusted basis, the company trades at 256x forward earnings, again far above its 144.5x average multiple.
Notable Risks Notable Risks
Given Cloudflare’s valuation, the entry is probably the most important aspect of this stock right now – whereas in the medium to long-term the most important aspect is timing for the broader inference market. The market looks to already be pricing in a sustainable AI inference-aided reacceleration on the topline despite the fact that this has not appeared concretely, amplifying risk if this does not visibly pan out over the next couple of quarters.
Conclusion:
We are thrilled about our new tier Discovery as the results are able to deliver new ideas to enthusiastic AI investors, such as ourselves. We quickly spotted the limitations around running an active portfolio that does not dedicate a separate effort to new idea generation as the market moves fast with new winners emerging every year. As we look at Q4 and beyond, we believe this quarterly analysis combined with an actively managed Top 10 list will become a strong offering. Our cumulative record proves we are one of the strongest teams in the world on AI stocks. Moving forward, our goal is to use our proven methodology to deliver additional value add as we participate heavily in the once-in-a-lifetime trend of AI.
In the coming weeks, we expect things to shift rapidly as new information is published daily during earnings season. Please reference our Top 10 Watchlist spreadsheet and incoming analysis as critical tools for staying on top of the Must Know stocks in the space.
Our goal is to update this list weekly, so stay tuned for frequent updates!
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Every Thursday at 4:30 pm Eastern, the I/O Fund team holds a webinar for premium members to discuss how to navigate the broad market, as well as various stock and crypto entries and exits. Beth Kindig offers weekly deep dives including lesser-known cryptocurrencies and AI stocks, plus the team offers trade alerts. The I/O Fund team is one of the only audited portfolios available to individual investors. If you’d like to subscribe to the Advanced Market Signals plan, email us at premium@io-fund.compremium@io-fund.com.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Hyperscalers are spending hundreds of billions of dollars annually on AI data center capex, from physical data center space, GPUs and servers, hardware and networking. With these substantial sums flowing towards GPUs that are now being refreshed on an annual cadence, the impetus for hyperscalers, neoclouds and other cloud providers turns to how quickly these GPUs can be energized and deployed, to maximize the period of returns before the next generation comes online.
If a company like Microsoft buys tens of billions of Nvidia’s Blackwell GPUs, the longer the massive investment in GPUs waits for power, the more delayed that revenue and profits become. In turn, this plays into market share as competitors who can energize GPUs faster will have a critical head start over those that are waiting for power. This is simple in concept, yet the lack of power having vast consequences cannot be overstated if you combine the sheer size of investments being made in AI alongside fierce, heightened competition.
AI is a spending race, but this means it is at the core, a power race. It does not matter if a hyperscaler spends tens of billions more on capex if it cannot secure the power to stand up new data center infrastructure to then deploy those GPUs immediately. The AI market is officially moving from being compute constrained to being power constrained, and this shift is important for I/O Fund members to prepare for.
We were among the first research companies to cover this topic in June of 2024, many quarters before the problem became well-known. We furthered this by investing early in a Bitcoin miner and one of the year’s highest-performing AI energy stocks. When we say we work hard to be early to trends for the benefit of our Members, we mean exactly that.
Given that we are soon approaching the moment when AI becomes (painfully) power constrained, we want to revisit this trend by examining how hyperscalers and others are powering new data centers. Below, we also look at the methods that can provide power the quickest, along with insights into location and costs, and more. As you can imagine, this analysis will help inform additional stock ideas as we position for 2026 and beyond.
Why Power is Critical, and Why it Will Continue to Be
More than one year ago, we first discussed how quickly power consumption was increasing with new GPUs in the analysis AI Power Consumption: Rapidly Becoming Mission-Critical. This trend is set to continue with Nvidia pushing towards an ultimate goal of super-sized 1MW server racks, or 8x more than GB200 racks.
Nvidia’s Blackwell lineup already brings a significant increase in power consumption, nearly double the H200’s 70 kW at 120 kW for the GB200 NVL72 and 140 kW for the upcoming GB300 racks.
Beyond Blackwell, Nvidia’s future design lineup shows continual increases in power consumption. Its Rubin generation is expected to boost thermal design power (TDP) by 50% over Blackwell at up to 180 kW per rack, with the upgraded Vera Rubin then doubling this to 360 kW per rack by 2027. In its largest configuration, the Vera Rubin NVL576, dubbed the ‘Kyber’ rack, could draw as much as 600 kW (0.6 MW), or 5x that of the GB200 NVL72 in just a two-year design timeframe. These figures do not include networking, interconnects, cooling and other hardware, which will further boost power draw per rack.
This rapid increase in power consumption per GPU generation is critical, as existing infrastructure is simply unable to meet these escalating power demands. For example, Applied Digital pointed out that nearly 70% of current data centers “contain racks requiring between four and nine kW of power, and less than two percent of data centers have racks with greater than 50kW.” For comparison, Super Micro’s GB200 NVL72 SuperCluster requires 132kW, while the upcoming Kyber rack could more than quadruple that to 600kW. Because Blackwell-based servers are now 15x to 30x the power density, cooling and power delivery strategies have to be redesigned, as liquid cooling now becomes a necessity.
This sharp rise in power density means current infrastructure may be unable to transition from 4-9kW racks to >130kW racks without incurring significant retrofitting costs, while building new infrastructure bypasses that hurdle and allows for optimization for high-powered racks.
The push to 600kW racks over the next few years means this is not a transient problem, rather it is one that the industry will continue to face, meaning continuous new construction may be needed to handle surging power demand. For example, Vantage’s upcoming 1.4 GW campus in Texas for Oracle is designed for ultra-high density racks up to 250kW, yet this will not be enough power to able to host NVL576 racks in just two to three years’ time. Additionally, a former Microsoft Azure AI executive reportedly said he estimated that the “requirements in terms of power for the data center would probably at least double every three years and maybe exponentially so over a period of time,” further reinforcing this.
How Much Does 1 GW Cost?
At this point, we can reasonably estimate costs to build 1 GW of new AI data center capacity, though ultimately, this will depend upon location (given some of the disparities in construction costs regionally), as well as hardware, as newest gen-GPUs will require more advanced ancillary equipment and cooling tech than older, less power generations (such as Hopper).
For example, GPUs and necessary ancillary hardware including InfiniBand/Ethernet, cooling and other equipment is likely in the range of $18 million to $24 million per MW for high-end GPUs such as Nvidia’s B200s.
IREN disclosed that it purchased ~4,200 B200s and ancillary equipment for $193 million, which calculates out to the mid ~$21 million per MW range based on total power draw of 1.93kW and a 1.1 PUE. For AMD’s Instinct MI350X GPU, which consumes ~1kw before ancillary equipment, costs could be towards the lower end of range, given its pricing at ~$25,000 per GPU, versus ~$30,000 to $40,000 for the B200. Pricing and power needs for custom silicon are much more opaque, though it is likely that these chips come at a greater discount compared to Nvidia and AMD’s leading GPUs.
Translating this to GW-scale, IREN’s 50MW facility can host >20K B200 GPUs, which, based on its purchase pricing, translates out to $21.6 billion per GW. For its planned 2GW Sweetwater campus, IREN claims it can support ~600K GB300 GPUs, which, at ~$80,000 per GPU, would cost $48 billion, or $24 billion per GW.
Cushman & Wakefield estimates new data center construction costs per MW in the range of $9-15 million across key markets, averaging $11-13 million. This aligns with estimates from CBRE for $10-14 million per MW, though costs can reach $16-20 million per MW in certain cases (or higher). This is simply for the ‘powered shell’, or the core building that has grid connection and has power in place, but has not been outfitted with racks or servers per tenant specifications.
So, assuming construction costs of ~$12 to $14 million per MW, total costs per MW for new facilities, GPUs and ancillary equipment are estimated between $30 to $38 million per MW. In total, this projects to roughly $30 to $38 billion per GW to build a data center from the ground up.In total, this projects to roughly $30 to $38 billion per GW to build a data center from the ground up. This is backed by Microsoft’s early 2025 announcement for ~$80 billion in spending for AI data centers in fiscal 2025 corresponding to >2 GW of new capacity additions.
Putting This in Terms of Capex
While capex is ultimately one of the more important figures for AI investors to track, it’s necessary to put in perspective how much capacity this capex correlates to, considering the tight grip power has over the industry.
Consider that the largest tech firms – Microsoft, Amazon, Alphabet, Meta and Oracle – are on track to likely spend upwards of $380 billion this year, up more than 50% YoY, and closer to $500 billion in 2026, with the majority going towards data centers. This puts total spending in 2025 and 2026 potentially as high as $880 billion, not even including CoreWeave, Nebius, xAI and others building out capacity.
Running off this calculation that each GW of new capacity could cost between $30 to $38 billion from the ground up, for the powered shell, GPUs and related hardware, we can reasonably estimate how many GW that Big Tech could bring online based on capex spend.
Assuming ~70% of capex goes directly towards data center capacity, given that hyperscalers and neoclouds alike continue to talk about supply constraints and strength of demand, this projects Big Tech could bring ~7-9 GW online with 2025 capex of $380 billion. For 2026, this would project ~9-12 GW of new capacity coming online, or cumulative total of ~16-21 GW.
Now, assuming closer to ~85% of capex goes towards new data center capacity, as Microsoft, Amazon and Alphabet now have a new cloud contender to deal with in Oracle (who also must build capacity rapidly to meet nearly half a trillion in RPO), this projects much larger buildouts. For 2025, this would estimate ~8.5-11 GW of new capacity and another 11-14 GW in 2026, or cumulative needs of nearly 20-25 GW. At midpoint, this would be ~4 GW higher than the prior assumption.
Forecasts All Point to Surging Capacity & Demand Growth
As we had covered in our free newsletter, Nuclear Power Emerging as a Clean AI Data Center Energy Source, data center power demand is expected to grow at an accelerated clip through the end of the decade and beyond, with more powerful GPUs and surging growth in inference two main drivers.
For example, Boston Consulting Group forecasts 45 GW of growth in global data center power demand in just three years, from 82 GW in 2025 to 127 GW by 2028. This represents an acceleration from a 12% CAGR from 2020 to 2023, to a 16% CAGR from 2023 to 2028.
On the other hand, McKinsey projects data center capacity will rise ~2.5x to 219 GW by 2030, up from 82 GW in 2025, with AI contributing 70% of that demand. This corresponds to total capacity growth of 137 GW over the next five years, with 112 GW coming from AI.
Goldman Sachs estimated global data center power usage at 55 GW in early 2025, far below BCG’s 82 GW figure. However, GS projects power usage to reach 84 GW in 2027 and increase further to 122 GW by 2030, corresponding to total growth of just 67 GW. However, considering 2025 and 2026 capex spend could support >20 GW of new capacity, this forecast may understate the pace of capacity growth.
This is quite the wide range of projected capacity growth over the next three to five years. But, more importantly, what level of capex does this require? Given the prior calculations for each GW to cost between $30 to $38 billion from the ground up (not accounting for future generation chips), building out 67 to 112 GW by 2030 could necessitate anywhere between $2 trillion to $4.3 trillion in capex over the next five years. McKinsey estimates that spending could reach as much as $6.7 trillion through 2030, which may support up to ~176 GW of new capacity.
At current projections, Big Tech is expected to spend nearly $1.2 trillion on capex from 2024 through 2026, meaning that these projections are well in the realm of possibility based on current spending trends.
Data Center Spending Up 30% YoY
The data center market in the US remains heavily constrained as high levels of demand are outstripping surging supply, even as data center construction reached $40 billion annualized in June, up 30% YoY and a new record. Primary market supply, including key regions such as Northern Virginia, Atlanta and Dallas-Fort Worth, rose 17.6% from H2 2024 and 43.4% YoY to a record 8.16 GW in 1H 2025, per CBRE.
Also, we see the highest amount of new data center construction in markets offering the lowest build costs. This suggests hyperscalers and other providers are seeking the cheapest paths for new capacity while aiming to bypass long connection wait times with behind-the-meter or off-grid sources, such as on-site gas turbines.
CBRE noted that Northern Virginia, Atlanta, San Antonio and Dallas-Forth Worth were the four markets with the highest amount new construction in the first half of 2025. The four combined for 4.86 GW of total capacity under construction, or nearly 82% of total new construction activity across primary and secondary markets. Interestingly, Seattle, which has one of the longer times to power at ~48 months, versus 36 months for Dallas, has just 9.5 MW under construction, while Chicago, with some of the highest construction costs, only has 244 MW under construction.
More on Location – Latency & Climate are Factors to Consider
There are more nuances about location beyond time to power and construction costs that factor into site selection, attractiveness and ultimately where hyperscalers, neoclouds or even miners choose to build. Two primary factors include latency and climate.
The map below shows data center presence by city, with major markets in dark blue, emerging markets with more abundant power in blue, and secondary markets in gray. Many of the recent headline-grabbing builds can easily be placed into either primary or emerging market locations.
The reason that these new, larger builds are often located in primary and emerging markets is not simply because of strong existing infrastructure or more power availability, but also for proximity to key cities with low-latency. For example, TeraWulf’s New York site offers sub-7ms latency to New York and <8ms to Boston, while Galaxy’s Helios data center in Texas offers <15ms latency to Dallas. Research from Applied Digital found that Stargate’s Abilene site and Northern Virginia both have <80ms latency to major cities across the US, with 100ms feeling ‘instantaneous’ to users.
However, data center vacancies dropped to a record low 1.6%, signaling strong demand from hyperscaler and AI customers, who continue to lock up supply quickly. CBRE added that nearly three-quarters of under-construction capacity of 5.25 GW was already pre-leased by hyperscalers and other providers aiming to secure capacity amid land and power constraints.
Industry Executives See Power as a Primary Constraint
Commentary from executives at hyperscalers, neoclouds, Bitcoin miners, colocation providers and commercial real estate firms all point to power as a key constraint (and consideration) facing the market this year and next:
CBRE said in its H1 2025 Data Center Report that “power availability and infrastructure delivery timelines remained the most decisive factors shaping site selection, leasing activity and pricing across all major U.S. markets.”
Equinix executives stated that “the amount of power we need isn't sitting around on the grid. And so we are planning, and I think most people in the room that are doing data center development are ensuring you have clear line of sight to that power before you take down any land or plan any data center capacity.” Execs also noted that the “reality of it is there [are] constraints in the marketplace, whether that's power availability in the key metros where we're looking to operate…”
A survey by Bloom Energy of 44 hyperscaler and colocation developers found that availability of power was the number one consideration for new site selection, with 84% of respondents placing that in the top 3 with an average rating of 7.8 out of 10.
TeraWulf CEO Paul Prager said he believes there is “a good argument that the market might even be tighter in 2026 than in 2025 given ongoing power constraints and rising hyperscaler CapEx.”
Amazon CEO Andy Jassy said that “you see some of the constraints and they kind of exist in multiple places, [but] the single biggest constraint is power.” Microsoft CEO Satya Nadella said Microsoft needs “power in specific places so that we can either lease or build at the pace at which we want.”
Google Cloud’s Thomas Kurian explained that “as you get these more powerful chips, they also take a lot more power. And power is, in many cases, a short resource.” Arm’s CEO Rene Haas has said that without improvements in efficiency, "by the end of the decade, AI data centers could consume as much as 20% to 25% of U.S. power requirements. Today that’s probably 4% or less."
Types of Data Center Builds, and Where Hyperscalers are Currently Going for Power
There are four common types of hyperscale/AI data center builds that are prevalent in the market: greenfield, build-to-suit, colocation, and brownfield. Each offers a different set of pros and cons as it relates to time to power, customization ability, and cost.
Greenfield: Refers to a hyperscaler or CSP owning the land, power and building the data center facility and infrastructure from the ground up. Greenfield builds offer the highest degree of customization ability over every aspect of the facility from power delivery to rack placement, though it comes at a much higher cost and often with the longest timelines to completion due to permitting, site selection, and grid connection.
Build-to-suit: Refers to a developer owning the land and securing the power for the facility, and constructing the data center tailored to the needs of the hyperscaler or CSP buying the capacity, typically via long-term leases. Build-to-suit data centers offer hyperscalers design flexibility without taking on the higher capex needs of a greenfield build.
Colocation: Refers to when a hyperscaler or CSP rents capacity (racks, power and cooling infrastructure) from a provider, offering a path to meet quick capacity needs though with no input on facility design.
Brownfield: Refers to retrofitting existing infrastructure to meet hyperscaler/AI needs, such as what Bitcoin miners are pursuing with existing mining infrastructure. Brownfield builds are often cheaper and faster than greenfield, but can be limited in terms of power and space by what is present with the existing infrastructure.
Meta’s 5GW Hyperion Campus
Meta is undertaking some of the industry's largest data center projects to support its AI superintelligence quest, with its greenfield Prometheus data center in Ohio expected to be the first 1GW campus to come online in 2026. This is followed by its Hyperion campus in Louisiana, expected to have an initial capacity of 2GW before scaling to 5GW over several years.
Meta’s Hyperion campus is expected to cost $50 billion (~$10 million/MW at full scale), with the social media giant securing $29 billion in financing from PIMCO and Blue Owl to fund the project. The facility’s power requirements are immense, equivalent to approximately 4 million homes.
To power this, Entergy is constructing three new combined-cycle gas turbines coming online in late 2028 to provide an initial 2.3 GW of power, while building new substations and installing new transmission lines. Entergy also may add an additional 2 GW of solar power to support the expansion of the campus towards 5 GW. Meta is said to be covering the $3.2 billion cost of the turbines for the first 15 years, while also pledging to bring 1.5GW of solar and battery power to the grid.
Amazon Strikes $18B Nuclear Deal with Talen
Amazon made a splash with the largest ever nuclear power PPA in history, purchasing 1.92 GW of nuclear power from Talen Energy to support colocated AWS data centers in Pennsylvania. Under the deal, Talen will ramp to full volume no later than 2032, with the deal extending through 2042 with options for further extension.
The two had initially attempted to go with a ‘behind-the-meter’ deal where Amazon would purchase power directly from Talen and bypass the grid, though FERC had blocked this in late 2024 on concerns about grid reliability and upwards pressure on consumer rates. Under the new $18 billion contract, the colocated power agreement “will transition to a ‘front-of-the-meter’ arrangement after the completion of transmission reconfigurations expected in the spring of 2026,” after which the plant will provide power to PJM’s grid with Talen acting as the supplier to Amazon and PPL responsible for transmission and delivery.
More broadly speaking, Amazon has signed utility-scale solar and wind deals globally, while also supplementing data center sites with on-site solar to augment grid power. Amazon is also said to be exploring fuel cell and gas turbine use at facilities to have more direct control over power.
Microsoft Adds More than 2GW of New Capacity
At the start of 2025, Microsoft disclosed that it was planning to spend roughly $80 billion through the end of its fiscal year in June $80 billion on AI-enabled data centers, to help ease capacity constraints and meet strong demand. In July, CEO Satya Nadella announced that Microsoft “ stood up more than 2 gigawatts of new capacity over the past 12 months alone,” marking a rather aggressive capacity expansion considering the company was said to have ~5GW at its disposal in early 2024.
Microsoft is continuing to build out its data center footprint, announcing a $4 billion additional investment in Wisconsin to house “hundreds of thousands” of Nvidia’s GPUs in a new facility, joining a $3.3 billion data center announced last year. Aligning with Nadella’s comments, Microsoft is also committing to leasing new capacity, with a mega build-to-suit deal with Nebius in New Jersey and a $6.2 billion colocation deal with Nscale and Aker in Norway.
Nebius’ new data center in New Jersey is being constructed by DataOne, who said in March that it would deliver the first phase of the data center in 20 weeks via a behind-the-meter solution. In Norway, Aker says that the new five-year deployment beginning in 2026 is powered by secured grid capacity and 100% renewable energy.
Alphabet Procuring Clean Energy to Support Data Centers
Alphabet is progressing towards its 24/7 Carbon-Free Energy by 2030 target, where each data center is backed 24/7 by clean energy. To support this, Alphabet said in its 2025 Sustainability Report that it procured 8 GW of clean energy primarily via long-term PPAs, that, once operational, “could generate nearly four times more electricity than our incremental load growth from 2023 to 2024.”
These include solar, wind and battery storage, as well as future investments for advanced geothermal or small nuclear reactors. The company said that in 2024, these PPAs brought 2.5 GW of clean energy to the grid to support its data centers.
Oracle Signs Deal with Bloom Energy for On-Site Power, Backs 1.4GW Natural Gas Data Center
Oracle has made a handful of different moves on the power side, signing a deal with Bloom Energy for near-immediate fuel cell deployment while also backing Vantage’s new 1.4 GW gas-powered West Texas data center.
Bloom is working to deploy its fuel cell tech at select Oracle Cloud Infrastructure (OCI) data centers in the US, with deployments expected to occur through late July to late October 2025. However, neither Oracle nor Bloom confirmed the scope, size or value of these deployments for on-site power generation.
Oracle is backing Vantage’s upcoming 1.4 GW data center, which is expected to see the first of ten buildings go live in the second half of 2026, built to handle next-gen ultra-high-density racks up to 250kW (versus ~130kW for Blackwell). Per Bloomberg, Oracle is set to spend more than $1 billion annually to power the campus with gas generators rather than waiting for a utility connection.
Crusoe Adds Natural Gas Turbines to Power Data Centers
Crusoe, developer of Stargate’s Abilene data center, has partnered with investment firm Engine No.1 to access 4.5 GW of power from seven of GE Vernova’s natural gas turbines that Engine No. 1 and Chevron’s joint venture purchased earlier this year.
These turbines are expected to bypass the grid and provide power directly to Crusoe’s data center campuses, with energy supply likely in place by 2027. However, Crusoe did not disclose whether this power would be directed to Stargate’s Abilene data center, as it is reportedly in discussions with multiple hyperscalers about where this power may be deployed.
xAI Tapping Gas Turbines for Colossus
xAI is powering its Colossus supercomputer via gas turbines, having more than doubled its number of turbines from 15 to 35 in April this year. The gas turbines have a combined capacity of ~422MW, per the Southern Environmental Law Center (SELC), though the group alleges that only 15 of these turbines are permitted. In May, the SELC also noted that xAI was aiming to add between 40 and 90 more turbines for its second Colossus data center in Memphis, raising concerns about pollution and health risks to nearby civilians.
Meeting Future Hyperscaler Power Needs
The most pressing question is, where does the industry go from here for data center power? Future hyperscaler needs continue to grow, with Amazon, Microsoft, Alphabet and Oracle combining for more than $1 trillion in RPO that will (hopefully) convert to revenue, while the broader industry could see anywhere between 67 to 112 GW (or more) of growth through 2030.
Utilities Expect Power Delivery Far Behind Hyperscaler Expectations
There exists a significant disconnect between when hyperscale and colocation developers expect to have site power, and when utilities expect to be able to deliver said power, according to research from Bloom Energy from April. Therefore, connecting new data centers to the grid in quick fashion may not be the most feasible option for hyperscalers looking to deploy gigawatts of capacity quickly, and instead, alternative power sources may be in higher demand.
For example, across the board, developers are expecting to have power delivered by 2027 on average, with most regions seeing expectations as early as late 2025. This is likely driven by consistent strong demand for AI infrastructure services, as new capacity will allow hyperscalers to meet more demand and drive more revenue.
Yet, utilities do not expect to deliver power in most of these primary and secondary markets until 2028, at the earliest, with Austin/San Antonio seeing one of the longest timelines at mid-2029.
This is supported by research from TD Cowen regarding grid connection timelines for new data centers, which span anywhere from 36 months to 48 months in these markets.
TD estimates connection timelines in Chicago at ~36 months and San Antonio at ~42 months, aligning with responses from Bloom’s survey. There has also been discussion regarding even longer timelines; in 2024, Bloomberg reported that utility Dominion Energy said >100MW data centers in Virginia were facing up to seven year wait times for new connection hookups.
Primary Market Grids at Risk of Shortfalls
Many primary markets like Northern Virginia, Texas, and Chicago are not necessarily the best equipped to handle surging data center demand, as the power grid in these regions is at elevated risk of supply shortfalls during extreme conditions.
For example, PJM’s grid will be at elevated risk from 2026 onwards, along with ERCOT in Texas, whereas the upper Midwest (MISO) is already at elevated risk. For example, ERCOT projects peak net loads may outpace generation capacity as soon as 2026. Thus, interconnection delays for its grid (and MISO) could stretch to up to 5 years to allow for more generation capacity to come online to avoid further stress.
This means that building out gigawatts of new capacity in at risk regions may place more emphasis on behind-the-meter deals, on-site generation to minimize strain on the grid, or adding back-up power sources to allow for shifting off the grid when needed.
For example, Oracle is said to be paying ~$1 billion annually for gas generators to power Vantage’s upcoming 1.4 GW data center in West Texas instead of waiting for the grid to be ready, or an extra 2.5% of its operating expenses each year for a single site. Microsoft likely selected Nebius in New Jersey for its ability to deliver hundreds of MW of capacity in ~12 months by going behind-the-meter. xAI stood up its Colossus cluster with 100K GPUs in just 4 months with gas turbines.
Where Does the Power Come From?
There are multiple different ways that hyperscalers, neoclouds and developers can get power to data centers to meet upcoming demand growth over the next few years, each offering its own benefits and drawbacks.
Grid interconnection: This is when data centers connect to the power grid under standard service, providing access to flexible power needs with no additional capex and a wide range of power generation options, including renewables. However, grid interconnection requests are often the longest time to power, ranging from three to seven years for hyperscale data centers in most key markets.
Behind-the-meter: BTM refers to when data centers connect directly to the power source and bypass the grid (meter), which can offer significant time advantage with stand-up times often in the range of several months to a year, along with cost savings from buying power direct versus at retail price. It also gives data centers more control over the power as well as a lower risk for disruption from grid outages. BTM deals can be sourced from multiple different power sources, such as solar, wind or nuclear.
On-site power generation: With on-site power, data centers will install their own power source within the facility grounds, also offering a relatively quicker time to power of a few months to over a year. On-site power can come in many forms, such as Bloom’s fuel cells, natural gas turbines or generators such as those from GE Vernova or Caterpillar, and in the 2030s and beyond, potentially small modular nuclear reactors. Bloom Energy’s survey found that 38% of data centers expect some form of on-site power by 2030, up from 13% last year.
Natural gas turbines/generators: NG is a widely available fuel source with a broad pipeline in the US, offering continuous power to data centers. Turbines can come in a range of sizes and be easily deployed, such as Caterpillar subsidiary Solar’s SMT-130 turbines that xAI is using, or GE Vernova’s LM2500XPRESS that Crusoe is using, scaling up to 1GW capacity. Notably, NG turbines could help meet substantial future demand, as GE Vernova is expanding manufacturing in South Carolina to be able to ship 20 GW worth in 2027. Large (>225MW) turbines are reportedly sold out over the next three years.
Fuel cells: Similar to NG, fuel cells can be quickly deployed (in as little as three months per Bloom and Oracle’s deal), and provide continuous power for operations. Due to be a relatively newer tech, FCs can come at a higher cost than NG, but without the related emissions. Bloom is planning to double its FC manufacturing capacity to 2GW in 2026 to meet rising on-site power demand.
Small modular reactors: SMRs are drawing more interest for future demand needs, as commercialization at scale is not likely until 2030 or beyond. Google is working with Kairos to bring 0.5 GW of SMR capacity online from 2030 through 2035, while Oklo and NuScale are progressing with commercialization plans and a long-term combined ~20 GW backlog.
Retrofitting existing infrastructure, ie. Bitcoin mining: This leverages existing infrastructure with secured power to the building, offering quick delivery times as short as a few weeks to a year, depending on cooling, flooring or other upgrades needed. While this method can offer quick time to power for >100MW sizes with low latency, low electricity costs and cooling expertise, miners are rather capital constrained and may be unable to build out capacity beyond what is currently in their pipelines. Miners have been attracting substantial deal activity, primarily from neoclouds, from an ability to deliver larger chunks of power quickly, with capex costs well below greenfield builds.
Where Hyperscalers May Go for Power Needs
Power is becoming one of the largest constraints for hyperscalers, neoclouds and developers, as surging power consumption with each GPU generation is necessitating new infrastructure to handle these increasingly power dense racks. The industry is racing to deploy hundreds of billions of dollars’ worth of AI servers and related hardware before the next refresh cycle to drive growth and maximize ROI.
As a reminder, Big Tech capex implies potentially more than 20 GW of new capacity will come online this year and next, while industry forecasts suggest global demand could rise between 67 to 112 GW by 2030.Traditional grid interconnections face several years’ worth of delays and cannot keep pace with how quickly hyperscalers want to stand up new data centers, putting the emphasis on alternative strategies to secure power.
Gas generators and turbines are emerging as a popular choice and likely will remain popular with tens of GW of manufacturing capacity coming online in 18 months along with readily available fuel. Fuel cells can also help meet near-immediate needs with rapid deployment timelines, though capacity is limited to only a portion of expected demand growth over the long run. Bitcoin miners have also found a role in meeting near-term demand, yet availability capacity is thinning out quickly following multiple long-term deals.
The I/O Fund is conducting deep-dive research on the energy sector as part of our ongoing focus on AI infrastructure. Our team is evaluating leading energy stocks that could play a pivotal role in powering the next wave of data center and AI growth. The results will be featured in our Top 10 New Ideas report, which will be delivered exclusively to Discovery Members by mid-October. Learn more here. Top 10 New Ideas report, which will be delivered exclusively to Discovery Members by mid-October. Learn more here.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Credo easily had one of the best earnings reports across the tech sector this quarter. The company reported growth of 274% YoY and 31% QoQ for revenue of $223.1 million. This beat estimates on the top line by 17% with management raising full-year revenue growth outlook by 35 points, from 85% YoY to 120% YoY. This corresponds to revenue of at least $960 million, and considering Credo expects two new hyperscaler customers to ramp in mid to late FY26, the company could well be on a trajectory to reach $1 billion in revenue this fiscal year. This would be an entire year ahead of analyst expectations, which projected $1 billion in revenue in FY27 heading into the report.
The bottom line also shined with adjusted EPS beating estimates by 44.4%. This represents growth of 1,200% YoY from a thin $0.04 in the prior-year quarter. Triple-digit growth of 425% on the bottom line is expected to follow although flat QoQ.
To be prudent, Credo’s Q2’s guide could be perceived as soft, at ~5.4% QoQ versus the 31% QoQ reported in Q1. Additionally, analysts were concerned about lumpiness in lead customers, although to be fair, Credo has done quite well with few customers up to this point by tackling a large total addressable market (TAM) with reliable yet cost-effective solutions. Credo’s proprietary serializer/deserialzer (Ser/Des) technology, active electric cables and digital signal processing (DSPs) are the cornerstone of its IP portfolio, giving the company a significant competitive advantage as it enables the power-efficient connectivity and reasonable pricing that the company is known for.
In a nutshell, this is why Credo is reporting surging growth in a highly competitive market: “Reliability and power efficiency [leads] to choosing AECs over optical solutions as they are up to 1,000x more reliable and consume half the power. AECs virtually eliminate link fabs, which are intermittent losses of connection, boosting cluster reliability and productivity while reducing power consumption.”
More information on how Credo plans to sustain growth into 2026 is noted below in the Q&A section, along with more details from tonight’s standout earnings report.
Revenue Growth Accelerates Nearly 100 Points
Credo smashed its own guidance by more than $33 million as it reported $223.1 million in revenue in Q1, riding strong demand for its power-efficient high-speed connectivity solutions. Growth accelerated nearly 100 points sequentially to 274% YoY, well ahead of estimates for 219%. On a sequential basis, revenue rose 31% QoQ, a slight acceleration from 26% QoQ in fiscal Q4.
Coming into the report, the bar was already set high as estimates called for a 150-point revenue acceleration over three quarters to >200%YoYgrowth.
For fiscal Q2, Credo guided for $230 to $240 million, or 226% YoY at midpoint, decelerating nearly 50 points from Q1. Sequentially, the guide feels more conservative at ~5.4% QoQ considering the product momentum Credo has behind it with AECs, optics and DSPs.
As of Q4, management had provided initial FY26 revenue guidance of >$800 million representing 85% YoY growth. Now, this has been updated to growth of 120% YoY, or north of $960 million for the year.
Customer Concentration:
In Q4’25, Credo disclosed its largest customer accounted for 61% of revenue (widely believed to be Microsoft). Management said they expected up to five >10% customers in FY26, up from three in FY25. Two additional hyperscalers were highlighted as ramping in mid-and late-FY26, with potential to become > 10% customers.
In its Q1’26, Credo provided some additional color around progress on these fronts. Management confirmed that the three >10% customers noted in Q4’25 were consistent in Q1’26, coming in at 35%, 33% and 20% of total revenue respectively.
The largest customer in FY25 continues to be the largest customer so far in FY26, signaling deepening relationships on-top of these new customer adds. Management also noted that Q1 included its first material revenue contribution from a 4th hyperscaler, with the 4th customer expected to surpass the >10% threshold by year end. Management expects continued progress around diversification through FY26.
Key Segments
Product Revenue: came in $217.1M, up 279% YoY and 31% QoQ. This is just shy of +303% YoY growth in Q4FY25 but still represents hypergrowth scale. Credo’s core engine remains AECs, which continue to benefit from rack-scale AI deployments). The fact that product revenue sustained triple-digit YoY growth while already running at a $200M+ quarterly pace suggest demand remains well ahead of consensus estimates.
IP License Revenue of $6.0M vs $4.2M in Q4 represented 44% QoQ. Still a smaller slice of the “revenue” pie but showing sequential growth. While licensing is still <3% of revenue but provides margin-accretive diversification. Growth here reinforces the stickiness of Credo’s SerDes IP, but the story remains dominated by physical product sales.
Engineering Services not broken out in this release, likely immaterial versus Product Sales.
Key Takeaways:
AEC remains the primary growth driver – scaling with hyperscaler AI deployments. Product revenue is highly concentrated.
Optics / DSP not broken out numerically this quarter, but management previously guided for 100% growth in FY26. Given the revenue beat, optics may already by contributing to incremental upside.
Retimers / PCIe 6 are still in early stage, but momentum in design wins should show up later FY26- FY27.
GAAP Gross Margin was 67.4%, up from 67.2% last quarter and up from 62.5% in prior-year quarter. Adjusted gross Margin was 67.5%, down from 68.0% last quarter but up from 62.9% in prior-year quarter. Coming into the report, Credo had guided for: GAAP gross margin of 63.4 – 65.4% and adjusted gross margin of 64 – 66%. This shows continued expansion despite hypergrowth, a rare feat at this scale. Gross margins are expanding as volumes surge – evidence that AEC and optics pricing power is intact, and scale is not being bought at the expense of margin.
GAAP Operating Margin was 27.2%, up from 19.9% last quarter and up from (24.2%) in prior-year quarter. Adjusted Operating Margin was 43.1%, up from 36.8% last quarter and up from 3.7% in prior-year quarter. This is the standout – massive operating leverage as opex grew only ~11% QoQ vs. the 31% pick up in revenue. Non-GAAP operating margin of 40% signals that Credo is already functioning with elite efficiency, while still in hypergrowth mode.
GAAP Net Margin was 28.4%, up from 21.5% last quarter and up from (15.9%) in prior-year quarter. Adjusted Net Margin was 44.1%, up from 38.4% last quarter and up from 11.7% in prior-year quarter. Net Margins largely mirror the operating leverage story – Credo has become a profit machinefar earlier in its lifecycle than most hardware names. The non-GAAP margin profile rivals leading semiconductor companies, while GAAP remains strong despite rising stock comp.
Adjusted EPS up 1,200% YoY
Credo reported GAAP EPS of $0.34, up from $0.20 last quarter and up from ($0.06) in prior-year quarter.
Q1’s adjusted EPS was $0.52, up 1,200% YoY from a thin $0.04 in the prior-year quarter and beating estimates by more than 44%. This also was a ~50% sequential improvement from $0.35 in Q4 driven by strong margin expansion down the line.
Credo is currently expected to report 435% YoY growth in adjusted EPS in Q2 to $0.37 and 58.4% growth in Q3 to $0.40, though given the margin strengths and sizable Q1 beat, these figures could move higher in the coming days.
Solid Cash Flow Generation in Q1
Credo’s balance sheet and cash flow position remain a core strength, underpinned by solid profitability, disciplined working capital management, and a debt-free structure. The company exited Q1 FY26 with $480 million in cash and investments, up from $431 million last quarter which should provide ample liquidity to support on-going product ramps and elevated R&D spend. Operating and free cash flow remained robust despite a sequential moderation from Q4’s unusually strong collections, with improvements in receivables management (DSO down to 73 from 86) helping offset continued investment in inventory. Inventory days held steady even as dollar levels rose, suggesting stocking is keeping pace with sales growth rather than accelerating further. Payables contracted notably, with DPO falling to 68 days from 91, reflecting less supplier financing and contributing to a softer cash conversion cycle. Overall, Credo continues to generate healthy cash margins, maintain a fortress balance sheet, and reinvestment modestly in capacity through capex.
Though cash flows moderated slightly from Q4, both operating and free cash flow margins remained >20%. This represented substantial YoY expansion in OCF and FCF margins of 35 to 45 points.
GAAP Operating Cash Flow of $54.2 million, down slightly from $57.8 million last quarter and up from ($7.2 million) in the prior-year quarter. This represents an OCF Margin of 24.3%, down from 34.3% last quarter but up significantly from (12.1%) in the prior-year quarter.
Free Cash Flow of $53.1 million, down from $54.2 million last quarter and up from ($13.1 million) in the prior-year quarter. This represents an FCF Margin of 23.8%, down from 31.9% last quarter but up more than 45 points from (21.9%) in the prior-year quarter.
Cash and Cash Equivalents (including short-term investments) of $479.6 million, up from $431.3 million last quarter. Credo remains debt-free.
Accounts receivable of $181.2M up from $162M in Q4’25 and $71.8M in Q1’25. Q1’26 implies DSO of 73 days compared to 86 days as of Q4’25, reflecting a meaningful improvement in collections. Credo is monetizing sales faster despite rapid growth which is supportive for cash flow sustainability and offsets some of the working capital drag from inventory.
Inventory of $116.6M, up from $90.0M last quarter and up from $31.5M in prior-year quarter. Q1’26 DIO (Days Inventory Outstanding) of 144 days is largely flat compared to Q4’25 DIO of 145 days. Inventory levels are holding steady relative to COGS, despite inventories increasing in dollar terms. This suggests the big build last quarter may have been a step-function while Q1 was more consistent stocking in line with higher sales volume.
Accounts payable of $54.9M, down from $56.2M in prior quarter and $38.47 in prior year. DPO (Days Payable Outstanding) in Q1’26 is down to 68 days compared to 91 days in Q4’25. This reflects a notable decline of 23 days in payables days and could be due to earlier supplier payments, changes in terms, or timing effects.
Capex was $2.8M, driven mostly by production equipment. This figure is down from $3.7M in prior quarter and down from $22.0M in prior year quarter. production equipment
Earnings Call Q&A Highlights
Commentary on Customer Concentration:
There was a moment during the call when the price action was more muted +5% versus +12% now. I believe it happened when management disclosed the lead customer represented 35% of revenue in Q1 compared to 61% last quarter. However, the market is being fickle if so, as customer diversification should be seen as a strength.
Here is what the CFO stated:
“Vivek Arya BofA Securities:
First set of questions is on the AEC market. If you could quantify how large each of your 10% customers were if they were the same as you had in the prior quarter?
And then Bill, if I zoom out, the market for you is now run rating closer to $1 billion or so. How large do you think this market is over time? And do you think this market is cannibalizing traditional copper? Or do you think at some point, it is even cannibalizing optical transceivers?
Daniel Fleming, CFO:
Vivek, this is Dan. Let me address the 10% customer question that you had. So as we mentioned in our prepared remarks, we had three 10% customers in Q1. They were the same three customers that were 10% customers in Q4. The mix was a little bit different, though. Our largest customer was 35% of revenue. Second largest was 33% and the third largest was 20%. So we're quite pleased with that kind of the customer diversity that we demonstrated within Q1.
But having said that, we expect continued diversification, as Bill highlighted, throughout fiscal '26. We do see two additional hyperscalers ramping, one of which, which was our fourth hyperscaler, we mentioned should reach to be a 10% customer for the full year of fiscal '26.
And then the last thing I'll mention on customer our largest customer for fiscal '25 is the largest driver of our growth in fiscal '26 as we stand right now and look forward. So that's an important factor to bear in mind as well as you look at how our year will progress..”
Later, an analyst asked for more clarity as to why the lead customer was down yet management stated the largest customer from last year will remain their largest customer for FY26 “by far, actually.”
Copper AECs versus Optical:
I think it’s worth repeating why Credo is seeing outsized demand in an otherwise crowded market – and one that can change rapidly in terms of which suppliers are confirmed, and if direct active cables are used (DACs), active electric cables (AECs) or optional solutions.
The question was: “And do you think this market is cannibalizing traditional copper? Or do you think at some point, it is even cannibalizing optical transceivers?”
This question is important as it’s asking why can Credo take market share with AECs from both traditional copper and optical.
The CEO stated: “But I think with the emphasis on reliability as it relates to clusters that we see customers really considering using AECs. It's really driven by reliability. I mentioned on the prepared comments that we're up to 1,000x more reliable, effectively reducing length laps and having the uptime of the cluster being much more.
Again, reiterating that if you have got a single link flap in, say, 10,000 or 100,000 or 1 million GPU cluster, it brings the entire cluster down because there's no redundancy from that NIC to tour connection. And so we're actually seeing the TAM expanding. And I think for the first time in history that you're seeing copper replacing optical connections. So we're quite bullish on the market generally.”
Specifically, where Credo has done well is with row-scale connectivity, which links racks in a single row. However, there are additional ongoing opportunities for Credo in rack-scale networking with NiC and Tor (top of rack) switches, as these both represent distances ideal for the reliability and power efficiency of AECs compared to optical. NiC-to-Tor would be a new opportunity for Credo: “Again, to reiterate, we see a huge opportunity from NIC to TOR applications as well as switch rack applications. And that is for front end and really emphasized much more strongly in back end, both in scale out and scale up.”
Overall management felt confident that the advantages AECs offer will continue to see increased adoption by hyperscalers with scale-out being the main driver now for Credo, yet scale-up being a future driver as well that is “an order of magnitude larger”
“When we talk about rack-to-rack, this is really the expanding part of the TAM really in both markets from the standpoint of AI back-end networks. as well as switch racks. Switch racks are starting to go to multiple rack architectures. And so as we talk about the near-term opportunity for rack-to-rack, it's really represented by the scale-out network. But long term, we see that the scale-up network also represents a really large growth in TAM, given the fact that we expect the volumes for scale-up to be potentially in order of magnitude larger than the scale-up connections.
So I think on several fronts, we can make the argument that we're still in the early stages. And I don't think there's any doubt in the market about if you can use copper, you will use copper given the advantages for reliability, power and cost.”
This was also stated in terms of how the market will only grow from here for Credo:
“Yes, I wouldn't say it's across the board. I would say that the first step typically is intra-rack, so 3-meter or less connections within the same rack. And this is just recent over the last 6 to 9 months that we've seen traction as our customers start to realize the opportunity to deliver much better cluster reliability and also secondarily better power. And so I would say that we're at the early stages still of having the market expand into rack-to-rack types of solutions. But I do think there's going to be an acceleration in the way that our customers view and use AECs.”
Increasing TAM from Tighter GPU Clusters, PCIe Retimers, LROs and More:
In the opening remarks, the CEO emphasized that Credo is expanding its total addressable market in a few key areas. That message is especially relevant for investors in a hypergrowth stock like this one, where analyst models currently forecast a sharp slowdown — from 274% YoY growth to just 26% over the next three quarters. Analysts have consistently set the bar too low, as shown by Credo’s roughly 30% beats for three straight quarters. Still, the key question remains whether the company can continue sustaining this pace of growth.
Management pointed toward the following:
Packing more GPUs into clusters is a catalyst for AECs:“We also see the trend towards GPU and cluster densification to continue to be a catalyst for an expanding AEC TAM. Over the past year, we've seen customer interest for AECs expand from intra-rack solutions to rack-to-rack solutions.”
Management also pointed toward scale-up as a significant growth driver, which makes sense and is what our thesis is formed on, yet it’s good to see the CEO emphasize this: “And so I think that for us, we'd just like to see the market go faster sooner because the scale-up opportunity represents a significant increase in TAM really over the next two to five years.”
Optical DSPs and LROs: Credo foresees expanding their TAM beyond copper with the goal of doubling optical revenue in FY26. Management hinted they plan to release more products for optical networking at the system-level and 800G LROs. There was also mention of improving the connection between GPUs and memory as a greenfield they plan to go after.
Ethernet Retimers and PCIe Retimers: Credo has recently expanded into PCIe solutions for AI networking which they stated significantly broadens TAM ahead of the shift to 200G per lane scale-up architectures. These products are called Toucan and Magpie.
During the Q&A, the CEO stated the PCIe scale-up opportunity was a bigger opportunity than Ethernet scale-up: “Yes. I would say that the near-term opportunity for us to scale up is really with the PCIe protocol as we see the market moving from PCIe Gen 5 to Gen 6. We do see that AECs will represent a really nice opportunity, both for intra-rack as well as rack-to-rack as scale-up goes row scale”
Conclusion:
We had stated in our Nvidia earnings write-up that the QoQ growth in Nvidia's networking segment should spell good things for I/O Fund members who hold Credo and Astera Labs. So far, so good in terms of the read-through.
What is unique about Credo is not only the hypergrowth that flows effortlessly to the bottom line, but that by my estimation, we are still very early to this trend. I recently said in an interview that networking is what defines the current generation of GPUs, and certainly Credo’s report supports this takeaway.
Typically, I’d be concerned a company like Credo is hitting peak growth, and there could be lumpy quarters; however, management spent a good deal of time on the earnings call going over why Credo is doing so well and why Credo will continue to do well from both a product differentiation standpoint (AECs are in high demand over traditional copper and optical solutions) but also how they plan to expand to meet the fluid needs of intra-rack and rack-to-rack architectures.
This marks the final week of a busy earnings season for the I/O Fund, and Credo saved the best for last. The company continues to hold a prime spot at the top of our aggressive buy list.
Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund own shares in CRDO at the time of writing and may own stocks pictured in the charts.
Coherent has all of the right products to potentially become a sizable player in AI networking. Primarily, Coherent’s growth story centers around supplying Nvidia with pluggable optical transceivers (400G, 800G, 1.6T) including EML lasers, VSCEL lasers and CW lasers, and emerging CPO technologies for next-generation switches and interconnects.
Transceiver speed has been growing with the highest data rates ranging from 100G to 200G to 400G. AI servers are driving a market for 800G data rates, which are shipping in production now, and 1.6T rates, which are initially shipping yet expected to grow in volume come 2026.
Coherent’s transceivers work with both Ethernet or InfiniBand, as well as proprietary protocols such as Nvidia’s NVLink and Nvidia’s interconnect chips NVSwitch. The company has stated that their 100ZR pluggable transceivers can upgrade old 10GBps Ethernet links with 100 GBps at the optical network edge, representing a 10X upgrade.
Coherent designs and manufactures the components, such as lasers, detectors and passive optics. Manufacturing the components (as opposed to buying them) is a strength as the company has supply chain resiliency by controlling the end-to-end process, and the manufacturing takes place primarily in the United States with some in Europe.
What COHR must navigate is that other companies also offer optical interconnects, and thus, competition is heightened in the datacom transceivers, silicon photonics, and CW/EML/VSCEL laser space. For example, Chinese companies such as InnoLight and Eoptolink compete with Coherent, larger companies like Broadcom, and then smaller companies with agile engineering like Lumentum also directly compete. In fact, Lumentum’s report was stronger than Coherent’s report this quarter in a a few key areas which likely drove COHR’s weak price action.
In terms of fundamentals, Coherent was weak in terms of the company guiding for lower sequential growth in its datacom segment than the prior two quarters. The company’s margins are also coming under pressure, which is the most important line item for a hardware company in terms of investor appetite.
With that said, I don’t think Coherent is out of the game by any means. There are some important things in the pipeline – all discussed for you below.
Quick note on Coherent Divesting the Aerospace and Defense Business:
This quarter, Coherent divested its underperforming segment Aerospace and Defense; selling the unit for $400 million. This will help the company streamline its operations and provide more cash to help pay down elevated debt. The sale will result in Coherent exiting from 10 sites and reducing employee count by 550 employees, and thus, will be accretive to EPS.
According to analyst notes, the expected headwind to revenue will be $170 million, thus growth will appear lower in the forward growth estimates provided below although the company will be growing organically at a higher rate.
Revenue Growth Decelerating to Single-Digits
Coherent reported a slight 1.1% beat to revenue estimates, reporting Q4 revenue of $1.53 billion, up 16.4% YoY. Revenue growth has decelerated more than 10 points since the start of the fiscal year.
For Q1, Coherent offered guidance for $1.46 to $1.60 billion in revenue, or $1.53 billion at the midpoint, excluding ~$0.02 billion in revenue related to its Aerospace & Defense unit that it expects to occur after the sale is closed.
This compares to consensus estimates for $1.55 billion. On a YoY basis, this points to YoY growth of 13.3% at midpoint, with growth expected to decelerate to the mid single-digit levels in both Q2 and Q3.
For fiscal 2025, Coherent reported 23.4% YoY growth in revenue to $5.81 billion, driven by a 51% increase in data center and communications revenue, offset by a (2%) decline in industrial and other revenue. This is discussed further in detail below.
For fiscal 2026, Coherent is expected to report just 8.4% revenue growth to $6.3 billion, down from 10.0% growth to $6.37 billion at the end of July; however, the decline in growth estimates looks to stem from the Aero & Defense unit divestment. BofA had estimated the sale would create a ~$170 million headwind (implying ~$6.2 billion), yet estimates are only ~$70 million lower, suggesting other segments are offsetting some of the impact.
May 2025 Investor’s Day:
A few months back, Coherent provided updated growth targets for revenue and earnings. It’s notable that Coherent is guiding for lower top line growth in the coming years yet helps to illustrate the primary growth will be on the bottom line.
Revenue growth of 23% in FY25 versus target model of 10% to 15%
Yet, operating margin expansion expected of 6.2+ points to 24% operating margin
Key Segments: Networking Growth Decelerates to 39%
Networking remains the primary driver of Coherent’s growth, with the company announcing that it commenced revenue shipments of its first 1.6T transceiver products and its first differentiated liquid-crystal optical circuit switch (OCS) platform.
Networking revenue rose 39% YoY and 5% QoQ to $945.2 million in Q4, with its share of revenue growing by two points sequentially to 62%. For the full year, Networking revenue rose 49% YoY to $3.42 billion, accounting for 59% of revenue.
Growth has decelerated steadily throughout the fiscal year, from 61% in Q1 to 45% growth in Q3 and now 39% in Q4. However, the primary concern is that the segment was showing a rather sharp deceleration on a QoQ basis, to the lowest sequential growth since early FY24. Coherent reported just 5% QoQ growth in networking in Q4, versus 10% in Q3.
Lasers revenue declined (2%) YoY and (4%) QoQ to $348 million, the segment’s first YoY decline in five quarters. Lasers accounted for 23% of revenue in Q4, down from 24% in Q3. For the full year, Lasers revenue grew just 3% to $1.44 billion.
Materials revenue declined (15%) YoY and was flat QoQ at $236.2 million, its sharpest decline since Q1. Of the three segments, Materials was the only to see revenue decline in FY25, down more than (6%) YoY to $953.8 million.
End Markets: Data Center Reporting Low 3% QoQ Growth
Turning to end markets, Coherent has now reorganized its reporting into just two end markets: Data center and Communications, and Industrial (which includes its previously reported Instrumentation and Electronics end markets).
The main concern is that Data center growth has slowed dramatically on a QoQ basis, from 11% last quarter to 3% in Q4, despite signs of accelerating AI systems demand (capex spend increasing from Big Tech, Blackwell shipping, Blackwell Ultra seeing initial shipments, etc)
Data center and Communications revenue rose 39% YoY and 5% QoQ in Q4 to $942 million. For the full-year, revenue increased 51% to $3.44 billion.
Within this, Data center revenue increased 38% YoY but just 3% QoQ. For FY25, Data center revenue rose 61% YoY. Similar to Networking, the major concern here is that Data center growth has slowed dramatically on a QoQ basis, from 11% last quarter to 3% in Q4, despite those signs of accelerating AI systems demand.
Management did state that “sequential growth rates can fluctuate quarter-to-quarter based on lumpiness of demand from our customers or supply or capacity related things,” but they offered little to soothe fears of rising competitive pressure within Nvidia’s supply chain. Aside from saying the see strong demand ahead, management dodged the question about when Data center’s QoQ growth would accelerate.
In a brief update on the Data center product roadmap, Coherent said it expects 1.6T transceiver volume to ramp throughout calendar 2025 with more meaningful contribution in calendar 2026, while demand continued to grow in Q4 for <1.6T data rates.
For Communications, Coherent said Q4 saw accelerated growth, up 11% QoQ and 42% YoY. For the year, Communications revenue grew 23%. Management stated that the 100G ZR product family is ramping rapidly, and they expected increasing revenue contribution through FY26 from 100G, 400G and 800G ZR/ZR+ transceivers.
Industrial revenue declined (8%) YoY and (2%) QoQ to $587 million, while revenue for the full-year declined (2%) YoY to $2.37 billion. Coherent said that above-market growth in industrial lasers and services was offset by a decline in silicon carbide, consistent with softer end market demand from autos. Management said that silicon carbide has stabilized and is not expected to be a headwind in FY26.
Margins Expand YoY in FY25
Despite gross margin expanding in Q4, GAAP operating margin shrunk, pressuring GAAP EPS; however, adjusted operating margin met management’s guidance for the quarter. For the full-year, Coherent delivered expansion for gross and operating margins.
Q4 GAAP gross margin was 35.7%, up 2.8 points YoY and half a point QoQ. Adjusted gross margin was 38.1%, slightly above the midpoint of guidance for 37-39%, up 2.3 points YoY but down 0.4 points QoQ.
Q4 GAAP operating margin was 0.4%, down 4.4 points YoY and QoQ; as a result of the pending divestment, Coherent recorded $85 million in asset impairment charges, impacting the margin. Adjusted operating margin was 18%, up 2.6 points YoY but down 0.6 points QoQ.
Q4 GAAP net margin was (6.3%), down 2.6 points YoY and down 7.3 points QoQ. Adjusted net margin was 12.6%, up more than 4 points YoY and nearly 1 point QoQ.
For Q1 FY26, management offered guidance for adjusted gross margin and adjusted operating expenses:
Adjusted gross margin was guided between 37.5% to 39.5%, up 1.8 points YoY and 0.4 points QoQ at midpoint.
Adjusted operating expenses were guided between $290-310 million, implying adjusted operating margin at 18.9%, up 2.8 points YoY and 0.9 points QoQ.
For FY25:
GAAP gross margin expanded 4.3 points YoY to 35.2%, while adjusted operating margin expanded 3.6 points YoY to 37.9%.
GAAP operating margin increased 3 points YoY to 5.0%, while adjusted operating margin increased 4.7 points YoY to 17.8%.
GAAP net margin expanded 4.1 points YoY to 0.8%, while adjusted net margin expanded 3.8 points to 11.9%.
EPS Beat in Q4, Guidance In-Line Q1
Coherent reported an 8.7% adjusted EPS beat in Q4, though offered guidance for Q1 in line with consensus estimates.
Q4 adjusted EPS was $1.00, ahead of estimates for $0.92 and representing YoY growth of nearly 64%. For Q1, Coherent guided for $0.93 to $1.13 in adjusted EPS, in line with the $1.03 estimate at midpoint and representing a deceleration to ~39% YoY growth. The deceleration is stemming from minimal expansion in adjusted margins in recent quarters combined with the top-line deceleration.
For FY25, Coherent reported 192% YoY growth to $3.53 in adjusted EPS. FY26 is estimated to see growth moderate to 30% YoY to $4.59, while management noted that the Aero & Defense divestment is expected to be accretive to EPS.
Cash and Balance Sheet
Cash flows moderated and cash flow margins shrunk to the lowest levels in the past six quarters. Coherent also noted that proceeds from the divestment will be used to pay down debt.
Operating cash flow was $130.3 million in Q4, down approximately (20%) YoY and QoQ. OCF margin was 8.5%, down nearly 4 points YoY and the first single-digit margin in the last five quarters.
For FY25, operating cash flow rose 16% YoY to $633.6 million, for a 10.9% margin, down 0.7 points YoY due to the softer Q4.
Free cash flow was ($1 million) in Q4, for a (0.1%) margin, down nearly 5 points YoY. It also was the first quarter with negative FCF since Q2 FY24.
For FY25, free cash flow was $192.8 million, down (3%) YoY. FCF margin was 3.3%, down nearly 1 point YoY.
Cash and equivalents were $909.2 million, while debt was $3.69 billion.
Inventories rose 3.6% QoQ to $1.44 billion.
Earnings Call Q&A:
800G and 1.6T Shipping, yet Data Center reporting declining QoQ growth
As discussed in our previous writeup on Coherent, the company supplies EML Lasers, VSCEL Lasers and CW Lasers for silicon photonics. While 100G per lane for 400G and 800G optical transceivers is what is supporting the growth now, it’s expected that 200G per lane and even 400G per lane for 1.6T optical transceivers is what will drive growth in the coming quarters.
Management offered the following update in terms of 800G ramping now and 1.6T ramping i the coming quarters:
“Okay. On the first part of the discussion, the way — I think the way to think about it is if you start with the 800-gig ramp, the 800-gig is obviously growing this calendar year versus prior. We expect 800-gig to grow again next calendar year, and that's ramping very quickly. And then on top of that, 1.6T, we believe, starts to ramp on top of that 800-gig ramp. We saw initial revenue in the prior quarter. We expect that revenue to grow over the coming quarters.”
In the May investor’s day, the company provided the following timeline for the ramp of the new data rates, showing Coherent has a healthy pipeline over the next few years.
However, despite these updates – the market is nervous that Coherent is not able to compete given data center sequential growth is declining QoQ. There was a pointed question on the call on the unusual QoQ decline Coherent is expecting (lumpiness). Although the answer from the CEO does not directly address the timing issues, the concern is significant enough to quote the exchange in full:
Vivek Arya, BofA:
So the first one, Jim, I realized this is a little bit more short-term oriented. But when I look at your data center and communications segment, sequential growth rate has gone from 9% in March to 5% in June, and I think your September implied is probably at or somewhat below this number, even though you're starting to ramp 1.6T and OCS. So what is the right way to interpret, right, this kind of somewhat slowdown because when I look at the deployment of AI clusters, they seem to be accelerating in the back half and one of your closest peers guided to double-digit sequential growth. So how would you address that pushback and do you think the sequential growth rates can start to reaccelerate at some point?
James Robert Anderson, CEO
Yes. Thanks, Vivek. On a quarter-to-quarter basis, the sequential growth rates can fluctuate quarter-to-quarter based on lumpiness of demand from our customers or supply or capacity related things, so. But I think if you look over the full year of fiscal '25, I'm quite pleased with our growth in data center for full year. We saw over 60% growth and even faster growth in the higher speed data rates. And we believe over that fiscal '25, we gained share over that fiscal '25. We feel good about fiscal '25 results. And as I said, looking forward into the current fiscal year, again, we see very strong demand ahead of us, and a number of different growth vectors, 800-gig, 1.6T. We talked about OCS as well — as well as seeing very strong demand in our DCI segment. So we feel good about the growth ahead of us. And certainly, making sure that we've got all the capacity in place to meet that — those demand signals.
InP Capacity Tripled plus 6-inch production line offers additional capacity
EMLs were traditionally used by telecom customers, yet became attractive for AI servers due to meeting the 200G per second speeds necessary for 1.6T optical modules to support AI models. These are called single mode optics, made of Indium Phosphide, which has been used instead of silicon for long-haul networking due to being a superior choice for optical functions, such as enabling the laser, modulator, photodetector and amplifier.
InP is more expensive at the component level as four EMLs are needed compared to two lower-cost CW lasers for silicon photonics modules, yet this difference at the component level can be made up for in data centers as InP reduces power consumption.
Setting aside the data center segment lumpiness, one reason Coherent may not be down for the count (time will tell) is they have recently tripled their InP capacity and are rolling-out a 6-inch InP production line for yet another significant increase in capacity.
According to Coherent, this will be the world’s first 6-inch wafers for InP. Inevitably, there were questions on the call about how quickly Coherent could produce more EML and CW lasers plus Co-Packaged Optics (CPOs) with this new production line:
“This is a big benefit to us in 2 ways, both capacity, obviously, on a larger wafer size we get a significant increase in capacity. But also, we expect a significant cost structure advantage. And so as we fully ramp that capacity both the ability to just drive higher volume, but lower cost structure is a big advantage for that. So we're really excited about that and looking forward to that ramp.”
Apple Partnership on VSCEL Lasers:
Apple and Coherent have partnered in the past, yet there was a new partnership announced recently for VSCEL lasers to be used in iPhones and iPads. Per the opening remarks:
“As mentioned in a recent Apple announcement regarding their American manufacturing program, we've entered into a new multiyear agreement with Apple for a new generation of VCSEL products that support Apple's iPhone and iPad products. We expect revenue from this expanded partnership with Apple to begin in the second half of calendar '26.”
Analysts are penciling in 2027 as the year where Coherent could see a more meaningful boost in revenue although despite this takeaway, management did repeat impact would be seen H2 CY2026:
“Yes, we feel really good about that expansion. I would describe it as an expansion of the partnership. It's a new generation of VCSELs that go into Apple iPad and iPhones and we do see that as an increase in revenue that will start to have impact to our revenue in the second half of next calendar year, so second half of '26”
Additional Key Points:
There are many moving pieces with Coherent and the following are also notable points from the earnings call:
Optical circuit switching will increase their TAM by $2 billion. OCS was covered in the past here. Coherent has a more advanced approach to OCS through liquid crystal technology versus the more mechanical MEMS technology that competitors offer.
Communications revenue is particularly resilient with 11% QoQ growth (outpacing DC when you separate the two). Driving this QoQ growth is DCIs (data center interconnects). Although recognized outside of the data center segment as part of telecom, the demand is driven by AI as the long-distance data transmissions were traditionally used for telecom purposes, yet and can range up to hundreds of kilometers and are now seeing demand for data center buildouts.
USA footprint: Coherent has 20 U.S. manufacturing locations in 13 states, an advantage should trade wars heat up.
Conclusion:
Coherent is in all the right place, but they are certainly not alone. The market and BofA analyst are correct to question why Coherent’s data center growth is declining QoQ especially given their competitor Lumentum had a solid earnings report. Lumentum saw 16% QoQ growth in Cloud & Networking in its Q4 and guided for ~10% QoQ in Q1.
Ultimately, we had trimmed Coherent quite a bit prior to the report simply because we felt Credo and Astera Lab deserved higher allocations. After the report, I feel the same – which is that Coherent has some work to do to not only secure a spot in our portfolio but to compete with the strong AI networking stocks we already own.
On our Discovery tier, we discuss if stocks we don't own such as Lumentum are a buy or not, plus other Nvidia Blackwell beneficiaries and AI data center energy plays. Portfolio Manager, Knox Ridley, also recently held a Discovery webinar discussing setups for new stocks. Sign up now to read more of our cutting-edge research and to view the 1-hour webinar.
Current Pro and Advanced Members: To subscribe to Discovery with 30% off, please click here to email us or email premium@io-fund.com and mention code DISCOVERY30.Current Pro and Advanced Members: To subscribe to Discovery with 30% off, please click here to email usclick here to email us or email premium@io-fund.com and mention code DISCOVERY30.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund do not own shares in COHR at the time of writing and may own stocks pictured in the charts.
