Applied Digital easily beat estimates in Q1 with revenue up 69% QoQ with fit-out revenue contributing significantly in the quarter as the company prepares to roll out its first phase for CoreWeave. Barely two weeks after the earnings report, Applied signed a $5 billion, 15-year lease at Polaris Forge 2 with an unnamed, investment-grade hyperscaler, which helps de-risk its story from being tied solely to CoreWeave. Combined with the neocloud’s lease at PF1, the two deals for will generate $1.07 billion in average annual revenue over the lifetime of the contracts, or ~5X fiscal 2025 revenue, highlighting the attractiveness of pivoting to AI data centers.
More importantly, Applied also quietly disclosed that they have 4GW in the pipeline with additional capacity under review, doubling from our prior update Applied Digital: Bitcoin Miner Hinting at Rare, Hyperscaler Deal, and nearly 3X more than the 1.4GW disclosed two quarters ago. With only 600MW currently contracted at PF1 and PF2, this hints at substantial upside to contracted revenue and net operating income at full scale.
Applied Extends Active Pipeline to 4.3GW, the Highest Among Miners
Perhaps the most important update coming from Q1’s earnings call was that Applied has quickly and quietly expanded its active development pipeline – just two quarters ago, the company disclosed 1.4GW in the pipeline, yet now it has tripled this to 4.3GW this quarter across nine sites. This would rank Applied atop the leading miners, outpacing Galaxy’s 3.5GW and IREN’s 2.9 GW of grid-connected power.
For the 4.3 GW, CEO Wesley Cummins explained that these are projects “we feel could move into that construction box in the next 6 to 12 months, and some of those could be even sooner. So those are things we're actively working on with permitting, with power, with all of those pieces that we think in the next 6 to 12 months can move into the construction pipeline.” He added that there is demand for sites ranging from hundreds of MW to a multi-GW scale, with emphasis on sites “built in a single location so that you get the cost advantages of building a scale in a single location,” which Applied’s pipeline covers with sizes from 250MW to 1GW+.
This is especially important as Applied continues to reiterate its ability to shorten its construction timelines, from 24 months down to 12 to 14 months. Management is working to match the pace of building with power delivery, starting construction early to ensure buildings are prepped and ready once power is available. Essentially, Applied is hinting that with limited holdups from permitting, with smooth power delivery and necessary financing, it could bring its pipeline to power in as quickly as two and a half years.
This could make the company increasingly more attractive from hyperscalers as other miners are not targeting having even 1GW online by the end of 2027 – management also disclosed that they have “entered negotiations with 2 additional hyperscalers for 2 new locations,” with 100MW under negotiation.
Should this pipeline materialize to operational capacity, Applied’s revenue and NOI opportunities could be 6X its current contracted capacity of 600MW. Assuming deal terms similar to PF1 and PF2, the remaining 3.7 GW pipeline could be worth $6.1 billion to $6.7 billion in average annual revenue, compared to the $1.07 billion in average annual revenue it has currently contracted out.
$5 Billion Lease Secured at Polaris Forge 2
Applied broke ground on Polaris Forge 2 in September, with the facility having an initial 300MW capacity. Applied said in Q1’s call that it has secured financing for the project via Macquarie with an expected cost of $3 billion, or $10 million per MW.
On October 22, Applied announced that it had signed a $5 billion, 15-year deal with an unnamed hyperscaler for 200MW capacity at PF2. On the headline, this is a slight discount to CoreWeave’s lease at $1.67 million per MW per year on average versus $1.83 million per MW per year, with a slightly lower NOI margin of ~86% +/- 3% versus 88% for CoreWeave’s deal. Management explained that having the hyperscaler provides a lower cost of capital, thus the spread between capital cost and revenue is approximately equal.
Securing this second deal with a major hyperscaler is important as it helps de-risk the story from being linked to CoreWeave, whose financials are upside down and require creative ways to raise cash to finance lofty growth ambitions.
“Firmly” On Track to Reach $1B NOI Target in 5 Years
While the hyperscaler engagement at PF2 is certainly good news to hear, management provided a snapshot into long-term net operating income (NOI) targets, providing a clearer view of how the deals will translate into earnings.
Management stated that they believe they can reach an “annualized NOI run rate of approximately $500 million once Polaris Forge 1 is fully operational,” while the “tenant signing at our second campus should put us firmly on the path toward our $1 billion NOI target within the next five years.”
At full scale, the 600MW of contracted capacity would translate into approximately $932 million in NOI based on expected margins of 88% and 86% across its two deals. Looking further out to Applied’s current active pipeline, the remaining 3.7GW could generate around $5.5 billion in annual NOI on average at full scale at similar margins.
While not a true comparison to NOI, analysts currently project Applied’s EBITDA to rise more than 10X by 2028, from $60.7 million expected this fiscal year to $640.2 million as these two deals begin to ramp towards full capacity. This would represent an expansion of EBITDA margin from 20.4% to 66%, still below targeted NOI margins. Additionally, there is the potential for EBITDA to rise by another factor of 7-8X in the long run if Applied can successfully commercialize its entire active development pipeline.
Project Financing Deal with Macquarie Unlocks 5X More Capital
As we discussed in our prior analysis, financing partnerships and capital raises are central to funding Applied and its HPC buildout. In Q1, the company drew $112.5 million from its $5 billion preferred equity financing with Macquarie, which management says helped fund the completion of PF1.
Applied also secured $50 million from Macquarie Equipment Capital, to help fund the groundbreaking for PF2, while also adding that it intends to tap the $5 billion vehicle to help fund the subsequent buildout. Additionally, Applied noted that subsequent to the quarter, it raised an $200 million from an expanded offering of its Series G Preferred Stock, providing more capital to fund these buildouts.
In the earnings calls, management noted that they may have the ability to finance both PF1 and PF2 by themselves, but they would prefer to tap the project financing from Macquarie as it lets them unlock significant capacity growth:
“When you look from a capital perspective, what we're seeking to do there is we could finance the Ellendale campus Polaris Forge 1 by ourselves. We probably even finance Polaris Forge 2 by ourselves.
But what we're trying to put in place and what we have put in place now is the ability for us to scale much larger. We're looking more into the future and putting a mechanism in place that eliminates or minimizes the dilution at the public company for a set amount at the subsidiary for Macquarie. And this allows us to go forward. The Macquarie Capital, $5 billion of capital really unlocks $20 billion to $25 billion of total capital for us when you include project finance and that allows us to build a significant amount of capacity.”
Instead of being capital constrained with two builds, Applied believes the $5 billion line from Macquarie could allow them to build >2GW with the amount of capital it can unlock.
Brief Update on Polaris Forge 1, South Dakota Development
Applied Digital this week announced that the first 50MW phase for CoreWeave is now ready for service, with the remaining 350MW to be rolled out in phases through 2027.
The fit-out of PF1 contributed $26.3 million in revenue in the quarter, with this expected to ramp significantly in the first part of fiscal Q2 leading up to the start of service in late October. Now that the first 50MW phase is online, lease revenues will begin ramping in the latter half of fiscal Q2 ending November and ramp further in Q3 as the next 50MW comes online by year-end.
Applied also provided a brief update on progress in South Dakota, where it was reported back in May 2025 that the company was planning to construct a $16 billion, 430 MW data center. Management said that power would be available in South Dakota in 2026, though the one piece they say is the gating factor for development is a sales tax exemption for IT data center equipment.
Financials
Revenue Surges 69% QoQ, Driven by CoreWeave Fit-out
Applied’s revenue rose 69% QoQ and 84% YoY to $64.2 million, driven primarily by the fit-out of Polaris Forge 1, which contributed $26.3 million in tenant-fit out revenue. This was more than 41% ahead of estimates for $45.5 million in revenue.
For fiscal Q2, revenue is expected to be $82.2 million, up 28.7% YoY and 28% QoQ. Fiscal Q3 (ending Feb 2026) is currently projected to see $71.4 million in revenue, up 35% YoY but down (13.1%) QoQ as fit-out revenue shifts to lease revenue.
Fiscal 2026 revenue is expected to be $297.3 million for YoY growth of 106.2%, with fiscal 2027 (ending May 2027) currently projected at $553.0 million for 86% YoY growth.
Gross margins felt a pinch in Q1 due to the ramp in fit-out activity, though operating margins improved from Q4 yet remain a decent distance from GAAP profitability.
GAAP gross margin was 13.4% in Q1, down from 20.5% in Q4 and 27.6% a year ago due to increase in low margin fit-out revenue. Applied said the $26.3 million in fit-out revenue carried a cost of $25 million, implying barely a 5% gross margin. The ramp of fit-out in Q2 may further pressure gross margin though this should ease by Q3 as lease revenue arises.
GAAP operating margin was (34.7%) in Q1, up from (54.5%) in Q4; the 72.6% year-ago comp is not necessarily comparable due to a $24.8M gain on assets held for sale. Adjusted operating margin was (5.6%), improving from (8.1%) in Q4 but down from 6.2% in the year ago quarter.
GAAP net margin was (28.8%) in Q1, improving from (70%) in Q4 and not comparable to the 45.5% from the year ago quarter. Adjusted net margin was (11.8%), improving from (19.9%) in Q4 but down from (2.3%) a year ago.
EPS Beats, but Not Yet Profitable
Applied beat on EPS in the quarter, with adjusted EPS of ($0.03) coming in well ahead of the ($0.16) estimate. GAAP EPS also beat at ($0.07) versus the ($0.13) estimate.
Looking ahead to Q2, GAAP EPS is expected to dip slightly to ($0.11), likely driven by margin pressure related to the ramp in fit-out revenue, before rebounding slightly to ($0.09) in Q3. For fiscal 2026, GAAP EPS is projected at ($0.45), before improving to ($0.27) in fiscal 2027 and shifting to a profit of $0.86 in 2028.
Cash Flows Heavily Negative on High Capex
Cash flows were heavily negative, with FCF margin widening to (516%) in Q1 driven by a sharp increase in capex.
Operating cash flow was ($82.0 million) for (127.7%) margin, down from 18.0% in Q4 but improving from (217.8%) in the year ago quarter.
Free cash flow was ($331.4 million) for a (516.1%) margin, driven by $249 million PP&E purchases. This compared to a (503.5%) margin in Q4 and a (375%) margin in the year ago quarter.
Cash and equivalents totaled $114.1 million, not including Applied’s $362.5 million raise subsequent to quarter-end. Debt totaled $687.3 million.
Conclusion
Applied’s second deal with a hyperscaler customer at PF2 boosts confidence in its AI data center hosting story and de-risks it from CoreWeave, putting it firmly on track to reach its $1 billion net operating income target by 2030, up from $60.7 million expected this fiscal year. Additionally, Applied disclosed that they have an active pipeline of 4.3 GW but with only 700 MW of capacity under construction, highlighting that revenue and NOI opportunities at full scale could be up to 6X larger at similar terms.
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 own shares in APLD at the time of writing and may own stocks pictured in the charts.
The energy crisis brought about by AI is a strong thematic hat will surface many opportunities for stock investors, yet it’s those companies that supply power quickly that we are building our portfolio with.
As a Bitcoin miner, Applied Digital has repositioned itself as an AI data center energy stock, with campuses primarily in the Dakotas. With long-term lease agreements — including multi-hundred-megawatt deals with AI cloud providers like CoreWeave — Applied Digital is a front runner that offers lower power usage effectiveness (PUE), an important metric that measures how much electricity is consumed by a data center. This PUE is achieved through free cooling as the Dakota region is colder than other regions and has surplus power the State exports. The company also hinted they are in advanced, direct negotiations with a hyperscaler. If this is confirmed, it will be the first among the miners to sign a direct deal.
With that said, many Bitcoin miners offer financials that are deep in the red yet are expected to sharply rebound from the high-margin revenue that AI data center deals will provide. Layer-in that Bitcoin miners have a highly volatile trading history, and what you get is a high risk/high reward approach to participating in stocks that promises to solve AI’s largest bottleneck – which is that AI's power consumption is outpacing what the grid can supply.
For investors, the challenge in participating in this trend is two-fold – how to identify opportunities in the vast energy sector and how to manage volatility given the margins and cash in this sector is lumpy (to put it kindly). We recently covered Bitcoin miners in our analysis “Bitcoin Miners Addressing AI’s Near-term Time to Power Bottleneck with up to $50 Billion in Commitments"Bitcoin Miners Addressing AI’s Near-term Time to Power Bottleneck with up to $50 Billion in Commitments"
As stated in the analysis: “Instead of having to worry about the prolonged process of site selection, permitting, planning, and more before final power delivery, these neoclouds instead have near immediate access to the powered shell. While retrofitting for liquid cooling, networking and connectivity may still be necessary and pose some challenges, [Bitcoin] miners offer a fast time to operation and relatively cheaper capex costs for a hyperscale-size data center outfitted with tens to hundreds of thousands of GPUs.
Bitcoin miners can only meet a fraction of this growth, likely around several gigawatts in total. Yet their innate ability to deliver this power over the next 12 to 24 months, supporting up to hundreds of thousands of high-end GPUs in larger-scale facilities, is why miners are prime targets to meet hyperscalers and neoclouds’ immediate power needs.”
Below we look at Applied Digital in more detail, the first in a Series of Bitcoin Miners that we have planned for our premium members over the coming weeks.
Note, this is a momentum stock and we plan to adhere strictly to risk management.
Applied Digital Located in the Dakotas with PUE of 1.18
Applied Digital made a strong argument in a recent white paper that offering a more strategic location in the Dakotas can result in savings of up to $60 to $90 million per year by offering free cooling.
We’ve covered in the past the importance of liquid cooling, as AI/ML require massive amounts of data processing, and as future generations of CPUs and GPUs are released, these systems will exceed air cooling capacity. Liquid cooling solves throttling, which occurs when CPUs and GPUs overheat and are throttled back to avoid damage to the chip. In the case of high-performance computing, liquid cooling reduces total cost of ownership as air cooling requires air conditioning and server fans to run constantly.
Cooling data center servers is responsible for 40% of the data center energy consumption. According to Dell, enclosed DLC solutions can save up to 23% of energy compared to traditional air-cooled racks. McKinsey places this number at 27% savings when there is 75% liquid cooled and 25% air cooled servers.
What Applied Digital is proposing is that mechanical liquid cooling is not needed as much in cooler regions such as the Dakotas, stating their mining operation “uses ambient air to cool liquid directly – no mechanical refrigeration needed for significant portions of the year.”
This represents a PUE improvement from the industry average of 1.58 to Applied Digital’s stated 1.18, translating into estimated annual savings of $60–90 million. In the example below, the calculation highlights $84 million in savings, driven by a significantly lower cost per kilowatt-hour.
Source: Applied Digital white paper
When comparing the Dakotas to other regions, such as North Virginia and Texas, the Dakotas have surplus energy with 33% exported compared to these other power-constrained regions. The region also offers 220 days of free cooling, which drives the power savings described above.
Management discussed these benefits in the recent earnings call, stating: “This design seeks to achieve a projected PUE of 1.18 and near 0 water consumption intended to ensure exceptional efficiency and sustainability. We like this location for its abundant low-cost synergy, some of which is generated from stranded power with over 200 days of free natural cooling. We have calculated that 100-megawatt data center customer could save up to $2.7 billion over a 30-year period as compared to the current industry data centers in other regions.”
CoreWeave is Primary Customer; Potential Incoming Hyperscaler(s)
As it stands, CoreWeave is the primary customer for Applied Digital’s capacity. Last quarter, CRWV signed a 15-year lease agreement for 250 MW and an extension for 150 MW for the Polaris Forge 1 data center site located in Ellendale, North Dakota.
The roll-out is expected to look like this:
First 100-megawatt facility scheduled to be operational in Q4 of 2025
Second 150-megawatt facility coming online in mid-2026
Third 150-megawatt facility planned for 2027
There are two paths to revenue for this deal. The first is “fit-out” revenue where Applied Digital is paid to prepare the site. This revenue is expected to kick-in for the current quarter and next quarter, leading to a “significant sequential” increase in revenue. From there, the lease deal is worth $11 billion over a 15-year time frame, or about $733M per year averaged out.
Here is what was stated: “We expect revenue to increase significantly sequentially, beginning in the quarter ending for August 2025 due to the technical fit out of our first Polaris Forge 1 building. Note, our customer pays the cost of this fit-out with a small margin to the company. This fit-out revenue will largely be recognized in both the current fiscal quarter and as well as the quarter ending November 30, 2025. Now this is before the actual lease revenue for the facility begins to be recognized.”
CoreWeave has a few strategic advantages over hyperscalers when it comes to pursuing energy deals with hyperscalers. The first is the speed in which a smaller company like CoreWeave can negotiate, the second is the company’s willingness to acquire power that is located away from major metro hubs, and third, CoreWeave’s core business is supply fast and AI-optimized access to GPUs, whereas Big Tech has many priorities across their core segments.
However, with that said, the market will favor any miner that can sign a hyperscaler given CoreWeave’s financials are upside down and require creative ways to raise cash to finance their ambitions. When it comes to a steady, stable flow of cash for a Bitcoin miner (who they, themselves have high debt leverage), there is no better customer than a hyperscaler. You can think of CoreWeave signing lease deals with Bitcoin miners as somewhat of a house of cards – it works well when stocks are up, it does not work well if we go through a longer period of AI softness.
This is where Applied Digital stands out – the company has hinted they are in negotiations with two hyperscalers with one of the hyperscalers being in “advanced negotiations.” Management made sure to point out how rare it is for a Bitcoin miner to have a hyperscaler as a potential customer.
The crux of the issue for Bitcoin miners is the quality and longevity of these stocks as they will have to diversify beyond cash-strapped neoclouds to raise their credit ratings and offer investors visibility in terms of funding the construction of these massive multi-billion dollar projects. Therefore, I’m quoting in full the discussion on the high likelihood that Applied Digital not only secures a hyperscaler deal – but do so soon:
“Besides CoreWeave, we have completed the diligence and onboarding process with 2 other investment- grade North American hyperscalers […] We also expect to benefit from this competitive advantage as new entrants to the market confront time, money and effort it takes to overcome these industry syncretic barriers to entry for other players. We also, for us, we feel we are now in a position to do business with these companies in the future with a much shorter negotiating and contracting completion process. In fact, we are currently in various stages of negotiation with several investment-grade hyperscalers for large capacity campuses other than our Polaris Forge 1 campus with 1 of those negotiations being in an advanced stage.”
How Quickly can Applied Digital Add Capacity (and How Much):
Applied Digital has a few paths for announcing more deals. The first is that Polaris 1 offers 1GW of capacity and yet CoreWeave has contracted 400 MW of the available 1 GW.
Secondly, Applied Digital broke ground on a new $3 billion campus this month called Polaris 2 in Harwood, North Dakota. This new campus is expected to “scale beyond scheduled operations in 2026 and full capacity in early 2027.”
In terms of size for Polaris 2 and stated delivery for full capacity by 2027, the following was shared in the call indicating it could be at least 1 GW and will be completed in 12-14 months time: “Building on the momentum from these leases and the surging demand for AI infrastructure, we're actively marketing our multi- gigawatt pipeline to a diverse group of customers. […] As a result, we believe we've reduced our projected build times from 24 months to 12 to 14 months, allowing us to deliver on large-scale commitments faster and more efficiently than before.”
Note, the exact size and timeline was not officially provided yet the comment above helps to frame initial expectations.
The Value of 2 GW
If we assume Applied Digital will lease 2G by early 2027, and if we figure that each 400 MW is worth $73 million based on the CoreWeave deal, then a rough estimate of what Applied Digital's AI factories will be worth is $18.25M per 100MW. This means the total 2GW is roughly worth $365 million. This does not take into account the “fit-out” revenue, which we will get a better idea of in the upcoming quarter.
Given the 2027 estimates are for $501 million in revenue, taken at face value, one could argue Polaris 1 and Polaris 2 are priced in. This does not take into account additional sites, and if a hyperscaler pays more than what CoreWeave is paying. According to commentary on the call, APLD may be breaking ground on a third site: “We do expect to break ground and work has already started for that on 1 additional campus and potentially 2 before the end of this year.”
Project Costs and Financing
When we look at capex costs of $11-$13 million per 100 MW, the cash flow for Applied Digital should pay for the project costs. Financing partnerships and capital raises are central to funding APLD and its HPC buildout. For example, Macquarie has committed up to $5 billion, with $900 million already allocated to Ellendale and $4.1 billion available for future development. In addition, Applied raised roughly ~$875 million in FY25 though equity and debt financing, with another $268.9 million post-quarter.
Cash / Debt Position: At the end of Q4 FY25, cash stood at $120.9 million and debt at $688.2 million, yielding a Cash / Debt ratio of .18x. Including the post quarter raise, the ratio improves to 0.57x, providing significantly more liquidity runway. Still, this lags peers such as IREN (0.59x) and RIOT (0.38x), while being broadly in line with WULF (0.18x).
As you can see from the tables below, Applied’s Cash / Debt ratio has been extremely volatile. FY24 cash levels hovered around $30-40 million with minimal debt, spiked to $314 million in Q2 FY25 due to financing inflows, and declined again as capex surged in Q3 and Q4. This underscores the company’s reliance on external financing, followed by rapid deployment into HPC infrastructure.
Looking deeper into the Company’s investor base, another strong signal comes from Nvidia’s involvement. In September 2024, NVDA participated in APLD’s $160 million private placement financing, landing NVDA a ~3% stake in the Company valued around $63.7 million at the time. Importantly, Applied was already recognized as a Preferred NVIDIA Cloud Partner prior to this deal, meaning the investment represents more than just capital, it reinforces strategic alignment. For Applied, having Nvidia on the cap table provides credibility, potential preferential access to GPU’s, and validates the Company’s pivot into high-performance compute hosting. Relative to peers, few digital infrastructure firms can point to this level of backing, which helps APLD as a trusted partner for hyperscalers and AI-native cloud firms.
While historical financials illustrate the growing pains of a transition year, investors should focus on the Company’s financing position and go-forward revenue estimates, which accelerate meaningfully over the next four quarters. The $7 billion contracted HPC backlog implies a structural step-up in revenue visibility, while the post-quarter raise improves the cash/debt ratio from .18x to .57x, extending liquidity runway. These forward-looking metrics are more indicative of Applied’s trajectory than backward-looking margins which are distorted by non-cash charges and build-out spend. Ultimately, Applied’s ability to balance its liquidity runway with execution on HPC energization will determine whether the $7 billion in contracted backlog translates into the sharp revenue acceleration that analysts expect.
Revenue
Revenue optics are noisy due to Cloud Services fluctuations, but the underlying Hosting growth is much healthier than the FY headline suggests. The Q4 exit run-rate and $7 billion plus in HPC lease backlog indicate a more favorable forward profile.
Q4 Revenue came in at $38.0 million, representing 41% growth YoY and 4% growth QoQ. YoY growth reflects Ellendale operating at full capacity (vs. partial outages in Q4’24) and the transition away from related-party contracts, leaving a more third-party-driven base. QoQ growth of $1.5 million shows the strength of Hosting revenues amid Cloud Services decline, which was reclassified as “held for sale”.
FY2025 Revenue of $144.2 million represents 6% YoY growth or $7.6 million compared to $136.6 million in FY24. The growth this year looks weak compared to FY24, mainly due to the termination of legacy contracts in Q1‘25 and inconsistent Cloud Services revenue across the fiscal year.
While FY25 revenue printed slightly below consensus ($144.2 million vs. $148.0 million), the shortfall was isolated to Cloud Services, which has since been discontinued. Hosting was broadly in line, underscoring the durability of the core business. As you can see in the chart below, analysts expect top-line growth to accelerate sharply moving forward, with estimates calling for $250 million in FY26 and $502 million in FY2027. These strong forward estimates help explain why shares have traded less on historical misses and more on execution / timing of HPC energization.
Shifting our focus toward the segment breakdown, Hosting is the durable core business with stable recurring revenues between $35 – $38 million per quarter. As mentioned above, Q4 marked the first time both Jamestown (106 MW) and Ellendale (180 MW) facilities ran at stable full capacity.
Cloud Services drove growth early in FY25 during Q1 and Q2, shrank in Q3, and has now been re-classified as held for sale. These fluctuations are largely driven by GPU contract structure changes and management’s decision to wind down the unit.
HPC / AI Leases are not currently contributing to FY25 revenue but these are critical for the Company’s trajectory and should be viewed as the growth driver. Management highlighted that signed leases with Coreweave represent $7 billion in contracted revenue to be recognized over 15 years. For added context, this contract alone is roughly 49x FY25 revenue of $144.2 million. Suffice to say, the Company is undergoing transition among its key segments but the strong backlog implies a change in trajectory.
Margins: Underlying gross margin strength offset by OpEx, SG&A caution signal
Gross margin trend is positive but material swings in Opex and financing costs mask any improvement. Q4 spike in SG&A is a red flag on cost control that should be monitored going forward. The key takeaway here is that unit economics are improving at the gross level but cost discipline and financing overhands remain key risks.
Q4 gross profit of $7.8 million, down $7.6 million or 49% compared to Q3, but up $3.7 million or 90% compared to Q4’24. Gross margins of 20.5% vs. 15.2% in Q4’24 reflect 530 basis points of improvement, driven largely by the increase in revenue, along with better facility uptime and efficiency, plus mix shift away from lower-margin Cloud Services.
Q4 operating income of ($20.7) million, $1.7 million lower than Q3’25 and $18.0 million lower than Q4’24. Operating margin of (54.5%) is down substantially from (9.9%) in Q4’24 and heavily pressured by a 115% increase YoY in SG&A spend as the company expanded headcount, increased stock comp, and ramped corporate overhead associated with the HPC buildout.
Q4 Net income of ($26.6) million reflects sequential improvement of $30.1 million against Q3’25 and annual improvement of $8.7 million compared to Q4’24. Net Income Margin of (70.0%) is up from (131.3%) reported in Q4’24, driven by higher revenue and better gross margins, slightly offset by increase in opex mentioned above.
FY25 gross profit $42.7 million is up $12.8 million compared to FY24. This represents a gross margin of 29.6%, 770 bps of incremental progress compared to 21.9% seen in FY24. The full year improvement is largely driven by increased scale and less power-related disruptions.
FY25 operating loss of ($16.8) million reflects improvement of $16.1 million compared to ($32.9) million operating loss in FY24. This represents an Operating margin of (11.7%) compared to (24.0%) in FY24. Despite the heavy expenditures noted in Q4, margins improved on a full-year basis, indicating operating leverage.
FY25 Net Loss of ($161.0) million continues to expand versus ($74.0) million loss reported in FY24. This represents Net Margins of (112%), down compared to (54.2%) in FY24. Despite the top line growth and improving unit economics, net margins for the full year worsened due to $119 million in non-cash losses tied to debt conversions, warrants, and derivative liabilities.
EPS
Adjusted figures provide a more accurate view of operating performance, showing narrowing core losses, while GAAP EPS was heavily distorted by financing and derivative accounting. Investors should focus on Adjusted EPS, which shows sequential and YoY improvement, while GAAP EPS remains noisy until the financing structure stabilizes.
Q4 GAAP EPS of ($0.12) improved from ($0.28) in Q4’24, reflecting stronger revenue and gross profit at both Jamestown and Ellendale as uptime normalized. On an adjusted basis, EPS narrowed to ($.03), close to breakeven, as higher gross margin partially offset heavier SG&A tied to HPC build-out.
FY25 GAAP EPS of ($0.80) worsened from ($0.65) in FY24 due to $119 million in non-cash charges (debt conversions, warrant issuances, and derivative liabilities). Adjusted EP of ($.06) versus ($0.11) in FY24 highlights meaningful progress in narrowing core losses despite heavy investment and Cloud services volatility.
Cash Flow and Balance Sheet
FY25 was defined by record capital deployment into HPC infrastructure. While Q4 cash burn moderated sequentially, full year OCF and FCF deterioration underscore the scale of investment and reliance on external financing. Liquidity was extended through large capital raises but leverage ballooned. Sustained improvement in OCF and eventual FCF stabilization will hinge on timely energization and monetization of HPC leases.
Q4 Operating Cash Flow of ($15.7) million, up from ($37.2) million in Q3’24, as working capital normalized. Operating Cash Flow margin improved to (44.3%) versus (101.9%) in Q3.
Q4 Free Cash Flow of ($79.2) million, improved from ($261.5) million reported in Q3, with FCF margin narrowing to (208.4%) from (716.4%). Sequentially, lower capex spend drove the improvement.
FY25 Operating Cash Flow of ($115.4) million down significantly from $13.8 million in FY24, with OCF margin falling to (80.0%), down from 10.1% reported in FY24.
FY25 Free Cash Flow of ($797.0) million, down significantly from ($128.0) million in FY24. FCF Margin of (553.0%), down from (93.7%) in FY24, reflecting the heavy investment cycle for Polaris Forge and HPC facilities.
Cash of $120.9 million, up from $31.7 million in Q4’24 and up $46.7 million sequentially, was boosted by roughly $875 million in equity and debt financing. As noted by Management, this “does not include the additional $268.9 million in proceeds from our ATM and Series G preferred stock offering that occurred post quarter.”
Debt of Q4 was $688.2 million, up from $653.3 million in Q3’25 and a significant step up compared to $79.5 million as of Q4’24 (nearly 9x YoY increase).
Conclusion:
Applied Digital asserts they are unique due to being located in the Dakotas where free cooling is available for more than half the year. The company also points toward this region’s surplus of power as a strategic advantage compared to power constrained regions like Texas and North Virginia.
Most importantly, should Applied Digital announce a deal with a hyperscaler, then it will be the first among the miners. Should the deal materialize, it would indicate their operations are attractive on a competitive basis given hyperscalers have a slow, thorough process for site review.
We hold a small allocation in Applied Digital and plan to actively risk manage the position.
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 APLD at the time of writing and may own stocks pictured in the charts.
Lumentum has been on our radar for some time as the company supplies components for datacom transceivers and optical interconnects. Lumentum is a small-cap company with differentiated technology that has caught the attention of heavyweight NVIDIA. We’ve been closely monitoring Lumentum for roughly a year, waiting patiently for their EML lasers for 200G to ship, enabling 800G and 1.6T bandwidths. Any progress here should continue into 2026-2027 for 400G data lanes and 3.2T bandwidths.
Optical interconnects help data centers accelerate data throughput between data centers and inside the data center between servers or racks, while reducing latency and power consumption. AI is driving cloud demand higher from the hyperscalers, leading to more data being created and processed, thus helping drive a need for these interconnects to meet demand for high-speed, low power data transmission in data centers.
The company supplies three types of lasers for datacom transceivers: VSCELs, CW lasers for silicon photonics and EML-based lasers. Per our previous analysis, the 200G EMLs are what is expected to drive an inflection in 2025.
In the most recent quarter, Lumentum reported their Cloud and Networking segment grew 18.3% YoY and 20.2% QoQ to $339.2 million. This caused total revenue to grow 9.7% yet total revenue is expected to accelerate to 46.6% and 43% in the second half of the year.
More on Lumentum’s EML Lasers:
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.
In the December quarter, Lumentum stated they began shipping its 200G lane speed externally-modulated lasers (EMLs) to multiple customers. The Company stated they would increase its EML capacity by 40% YoY from June 2024 to June 2025, followed by another 40% increase by the end of calendar 2025 as its Thailand production accelerates – representing compounding growth in an 18-month time frame:
“So, we're just overall the demand is outstripping even that 40% compounded twice in an 18 month period of time. We do have additional capacity expansion beyond the end of calendar '25 obviously and those are investments that we've made over the past several quarters that come online in calendar '25, which will give us increased capacity as well.”
Lumentum’s indium phosphide 100G EMLs (Externally-Modulated Lasers) are being shipped and used in leading single-mode 400G and 800G optical transceivers. These customers are now designing the company’s 200G EMLs into their next generation of transceivers. Already being a lead supplier for 100G EML transceivers, the company is set up to be in pole position for the 200G EML transceivers.
Per the most recent earnings call: “We achieved another record for EML unit shipments in Q2 and began delivering 200G lane speed EMLs to multiple customers.
Based on the breadth of our 200G EML design wins, we expect to gain additional laser transmitter market share in the upcoming wave of 800G and 1.6T transceivers, utilizing the more efficient 200G EMLs for AI applications. Complementing our EMLs are our new 200G lens integrated photodetector arrays, which adds to our content opportunity in next generation 800G and 1.6T transceivers as well as strengthens our vertical integration strategy for our own cloud modules.”
Specifically, the company’s experience in InP long haul transceivers is being tapped as AI servers scale, especially since InP reduces power consumption compared to silicon. Lumentum is also working on higher speed optical links, including 400G per lane. This means more data, fewer lanes if you compare it to 200G per lane. This is important especially as data centers will eventually move not only to 1.6 terabits per second (1.6T) (4X400G or 8X200G) but also 3.2T (8X400G).
Management also stated the new 200G lens integrated photodetector arrays “adds to our content opportunity in next generation 800G and 1.6T transceivers.”
Lumentum’s Co-Packaged Optics Opportunity for High Power Lasers:
Silicon photonics are the only viable choice for rack-to-rack and across the data center due to the need for high bandwidth and lower power at high speeds. There is also low-loss over long distances with optical fiber, which refers to preserving the original signal, whereas copper sees signal degradation over longer distances.
Where there is a debate and an important shift occurring is in the networking between chips and inside the rack. We’ve recently covered the benefits of co-packaged optics as a replacement for pluggable transceivers when networking between chips. At 800G and 1.6T speeds, the electrical signals over power circuit boards (PCBs) run too hot, are power hungry and create loss for the signal. The overall goal is to move the optics closer in proximity to overcome scaling issues when increasing electrical speed.
Nvidia stated that by replacing pluggable optics with silicon photonics on the package, it can “deliver 3.5x more power efficiency, 63x greater signal integrity, 10x better network resiliency at scale and 1.3x faster deployment compared with traditional methods.”
Nvidia VP Ian Buck stated at GTC that the CPO switches help reduce power consumption by eliminating the need for external lasers and pluggable transceivers to achieve a significant reduction in power from 39 watts to 9 watts. Buck explained that this “gives you that benefit from going from 39 watts of power down to only 9 watts of power for the same number of ports, and that's huge. It doesn't sound like 39 sounds a lot. But if you get 400,000 GPUs in an AI supercomputer, there's like 24 megawatts of lasers like so that's a lot of laser light that could be optimized and made more efficient.”
While Lumentum may see some decline in its pluggable optics transceiver business as the industry shifts, it is also well-positioned to benefit from the transition. The company is a key supplier of high-power lasers required for co-packaged optics (CPO), which could offset near-term losses. Taken together, the impact should be at the onset net neutral — with (potentially significant) upside over time from growing demand for EML lasers, the transition to 200G and 400G per lane, plus optical switching (see below).
Regarding CPOs, here is what was stated on the most recent call: “But I'd also highlight for Lumentum a little bit of a unique situation where we're more modest share on transceivers today and share gaining, if you will, over time. So even if there is some cannibalization in the mid to long-term, we don't think it impacts our transceiver opportunity and creates an expanded opportunity for high power lasers.”
Nearly a year ago, Lumentum announced an ultra-high output power 1310 nm DFB laser in beta, stating it was designed to reduce the “required number of lasers” while boosting efficiency and reliability in large scale AI/ML infrastructure.
It was also shared that Lumentum is shipping preproduction volumes now related to CPOs:
‘As part of one collaboration, we began shipping preproduction volumes of our unique ultra-high power lasers to an AI infrastructure customer for a proprietary interconnect solution in Q2 and have received follow-on orders as well as excellent feedback on the product's performance. This is a very exciting opportunity.”
It was later teased out on the earnings call that 2026 could be a big year for Lumentum’s high power laser: “I would say preproduction volume with forecasts for more meaningful volume throughout calendar '25 and then real extremely high volume in calendar '26. So, feedback is very positive. The performance is very unique and the line width that we produce with the laser is superior to anything in the market.”
CPO (with Switch ASIC) + Optical Switch + GPU Clusters = The Future of Scale Up AI Systems
Optical switches are a new kind of switch for AI clusters that handles 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.
There are many competitors within optical switching, with heavyweights Broadcom and Arista coming to mind, yet Lumentum believes their MEMS-Based technology can set them apart. Although the discussion around MEMS can get quite technical, the idea is that Lumentum is a smaller, (potentially) key supplier for customers putting optical circuit switches into their data centers. Another use case for using Lumentum is to rearchitect or write software to enable the optical circuit switching.
According to a previous earnings call, Lumentum has “already shipped evaluation units to customers who have provided overwhelmingly positive feedback on our performance.” It was also stated that “more meaningful growth will probably be in calendar 2026” for the optical switching circuit products.
Partner for Nvidia’s Silicon Photonics & Marvell 400G per Lane
Nvidia recently announced 1.6 terabits per second port switches at GTC “to deliver 3.5x energy savings and 10x resilience in AI factories.” Lumentum was named as a partner, among others in the industry.
This analysis has covered some of Lumentum’s unique advantages and why they would be chosen as a partner, such as:
At the forefront of increasing bandwidths, from data rates of 400G, 800G, and the upcoming 1.6T
At the forefront of faster data pipelines, from 100G per lane (today), 200G per lane (ramping now and into H2 2025), and 400G per lane (into 2026 and beyond)
Indium Phosphide lasers are more expensive due to volumes at the component level compared to silicon — yet is made up for in data centers as InP reduces power consumption. InP also offers higher signal integrity.
1310nm DFB lasers are high power lasers that enable the transition to co-packaged optics.
Last month, Lumentum announced an InP DFB-MZI optical transmitter in partnership with Marvell to demonstrate 400G-per-lane PAM4. The long acronym refers to a laser-modulator combo — specifically, a DFB laser integrated with a compact Mach-Zehnder (MZI) modulator. This was paired with Marvell’s 400G PAM4 DSP, which operates at 225 Gbaud and enables the high-speed signaling needed to reach 400G per lane.
The result is a high-performance, power-efficient optical transmitter that outperforms traditional silicon photonics — “particularly in applications where signal integrity and efficiency are critical,” per the press release. The InP DFB-MZI platform sets the stage for architectures expanding to 1.6T and 3.2T bandwidth, which is in the near future, up from the 800G modules in production today.
Lumentum is Supply Constrained; Building Capacity
Lumentum’s demand far exceeds supply, including indium phosphide supply and components such as CW lasers.
The company is expanding its transceiver manufacturing capacity through the construction of a three-story facility and cleanroom in Thailand, which complements existing production lines. The first floor is completed and ready for tool installation.
As stated, Lumentum began shipping its 200G lane speed externally-modulated lasers (EMLs) to multiple customers in Q2 F2025. The Company aims to increase its EML capacity by 40% YoY from June 2024 to June 2025, followed by another 40% increase by the end of 2025 as its Thailand production accelerates.
In Q2, the Company invested $65 million in capex to expand cleanroom capacity and increase equipment capacity for InP wafer production to support EML chip manufacturing at its Thailand site. Lumentum has also been invested in expanding its indium phosphide (InP) wafer fabrication facilities, which are critical for producing high-speed lasers like the EMLs used in transceivers.
CEO Lowe reiterated this in the Q2 F2025 conference call:
“We're still on track to what we've been saying, which was 40% or higher growth from the June quarter of calendar '24 to the June quarter of calendar '25 and then another 40% by the end of calendar '25. So, that's for all 100 gig and 200 gig. I think we were in our prepared remarks, we talked about being overall supply constrained, not 200 gig because we can start a 200 gig wafer or a 100 gig wafer just the same. So, we're just overall the demand is outstripping even that 40% compounded twice in an 18-month period of time… Our wafer fab expansion plans to enable higher volumes of EMLs and other indium phosphide lasers and photodetectors continues to be on track. We still anticipate that demand for our EML chips will continue to exceed supply, at least into calendar year 2026. We are experiencing strengthening demand for our DCI products as well as long haul transmission and transport solutions.”
A Note on China Exposure
Interestingly, Lumentum does not have as much exposure to China whereas most supply chain troubles right now are due to sourcing in this problematic geo-political zone
This was stated on the last call:
“We have some production that happens in China and then those components are integrated into bigger components, bigger products at our Thailand facility. And then most of the shipments happen from Thailand, even if they are to the US or to Mexico as well. And then much of the growth that we're seeing is shipments that are coming from Japan as well as from Caswell for the transmission products that we have. So because of that, at least in the short to mid-term, unless policy changes, happen at a government level, we're not expecting much of an impact.”
This was also stated in the introduction:
“Second, we are scaling capacity for our highly differentiated laser transmitter chips in our indium phosphide wafer fabs and optical circuit switch and transceiver production capacity in our proven factories outside of China to meet the rising demand.”
Quite a lot has changed in terms of policy at the government level since the last earnings call. The overall commentary regarding China reliance being minimal on the supply chain side may be true — but tariffs could upend enough of the supply chain to affect Lumentum’s customers (and overall demand).
Also, despite not relying on manufacturing in China, Lumentum reported Hong Kong as a major customer in the 10-Q at 19% of revenue for LITE, second to United States. Geo-political tensions should cause China/HK to source domestically either voluntarily or through blacklists.
Telecom Problematic End Market yet DCIs are a Bright Spot
There has been a steep inventory correction in telecom that has led to substantial revenue decline and significant margin erosion, presenting a major fundamental headwind for data center growth to overcome. However, data center interconnects (DCIs) are helping to drive the turnaround in this otherwise problematic end market.
Lumentum offers tunable laser and coherent pluggable transceivers for data center interconnects (DCIs). These long-distance data transmissions are traditionally used for telecom purposes and can range up to hundreds of kilometers yet are now seeing demand for data center buildouts.
In April 2025, Lumentum announced the sampling of new 400/800G ZR+ L-band pluggable transceivers and the general availability of its 800G ZR+ C-band module. The L-band modules effectively double the usable wavelength range and available fiber capacity. By expanding into both C-band and L-band spectrums, Lumentum’s transceivers enable increased fiber capacity, and are able to serve AI and cloud-based applications. These pluggable modules reduce overall system complexity and cost.
Regarding DCIs, per the previous earnings call: “we're seeing dramatic strength in anything ZR, anything to connect data centers as data centers are being built out, and that can take the form of ZR modules. But given our share of tunable lasers that go into ZRs, that's where we're going to see a dramatic pickup in the telecom side.”
The most recent update from management is that demand remains very, very strong: “And so the data center interconnect and the networks that are interconnecting these data centers is showing very, very strong demand. And so DCIs and the components as well as ROADMs amplifiers and longer haul coherent transmission connecting further apart data centers is very strong.”
Revenue Growth Poised to Accelerate
Q2 FY25 revenue grew by 9.65% YoY and 19.37% QoQ to $402.2 million, driven by strength in its Cloud and Networking segment, beating estimates by 2.87%.
Management guided Q3 revenue between $410 million to $425 million, with a midpoint of $417.5 million for 13.9% growth.
Analysts expect revenue to grow 46.6% in the June quarter to $452 million and 43% YoY in the September quarter to $481.9 million in Q1 F2026.
Management also reaffirmed its commitment to reaching quarterly revenue of $500 million by the end of calendar year 2025, driven by improving trends with its networking equipment manufacturing customers.
Segments:
Cloud and Networking Drives Revenue and Margins
In the most recent quarter, Cloud and Networking grew 18.3% YoY and 20.2% QoQ to $339.2 million with management stating they saw “sequential increases in nearly all of our Cloud and Networking product lines.” Cloud and Networking segment profit grew 16.2% for an increase of 330 basis points sequentially and an increase of 610 basis points year-on-year on higher revenue. This segment includes optical transceivers, the datacom chips/lasers that go into optical modules (EMLs, VSCELs, CW lasers, etc) and telecom/data center interconnects.
Previously, in the September quarter, the segment saw a meaningful inflection with growth of 23% YoY and 11% QoQ. Next quarter, management stated they expect the segment to increase by $25 million QoQ, which would represent growth of 7.3% QoQ – not quite as high as the previous quarter yet it’s been clearly noted in our previous analysis that H2 would be the bigger ramp. Our analysis on delayed Nvidia suppliers also connects some dots on this particular timing for Lumentum.
According to management commentary, the company shipped record EML units with datacom transceivers shipping to their largest hyperscaler customer and volume production shipments to a new customer.
The company stated they expect to increase their market share:
“Based on the breadth of our 200G EML design wins, we expect to gain additional laser transmitter market share in the upcoming wave of 800G and 1.6T transceivers, utilizing the more efficient 200G EMLs for AI applications. Complementing our EMLs are our new 200G lens integrated photodetector arrays, which adds to our content opportunity in next generation 800G and 1.6T transceivers as well as strengthens our vertical integration strategy for our own cloud modules.”
Industrial Tech Continues to Contract
Q2 F2024 Industrial Tech revenue fell (21.4%) YoY and grew 15.4% QoQ to $63 million. The sequential increase was driven by higher industrial laser shipments, partially offset by lower 3D sensing shipments.
Next quarter, management expects revenues to decline sequentially by $10 million, driven by declines in both commercial lasers and 3D sensing.
Adj. Margins Bottomed Out and Accelerating
GAAP Gross margin was 24.7% with adjusted gross margin at 32.3%. Per management, company gross margins will “sequentially increase as manufacturing utilization improves as well as an increase in Datacom laser shipments.”
GAAP operating margin was (12.8%) for operating losses of ($51.6) million.
Adjusted operating margins rose to 7.9% compared to 3% last quarter. It’s expected to further expand to 10% at the midpoint this quarter.
GAAP net margin was (15.1%) for net losses of $61 million
GAAP adjusted net margin was 7.46% for adjusted profits of $30 million
Adj. EPS Returns to Growth and Accelerates Quickly
Q2 FY25 adj. EPS improved to $0.42, beating consensus estimates by 16.97%. Management guided Q3 F2025 adj. EPS between $0.47 to $0.53, with a midpoint of $0.53. There is outsized growth in EPS from the rebound on small numbers:
Analysts expect Q4 adj. EPS to grow 957.38% to $0.63
Q1 is expected to see 332.24% to $0.78
The difference between GAAP and non-GAAP operating margin is due to the stock-based compensation expenses and amortization of acquired intangibles.
Elevated Capex Spend in Thailand Manufacturing Site Pressuring Free Cash Flow
Q2 FY25 operating cash flow was $24.3 million or 6.1% of revenue. The Company spent $74 million capex in Q1 FY25 and $64 million in Q2, resulting in sequential negative free cash flow.
Free cash flow in Q2 was ($15.9 million) or -4% of revenue. The Company closed the quarter with $896.7 million in cash and cash equivalents and $2.47 billion in debt. This puts the debt-to-equity ratio at 2.75 due to high capex spending — which is high and not ideal in this environment. However, if Lumentum can prove the capex will quickly be converted to revenue, it may become a non-issue by this time next year.
Lumentum spent $64 million in capex for expanding cleanroom capacity at the Thailand manufacturing site and increasing equipment capacity for indium phosphate wafer production to support EML chip manufacturing.
Valuation
Lumentum trades at a forward price-earnings (P/E) of 31.3 and a current PE ratio of 143 (although this looks drastic given the weak bottom line the company is rebounding from).
The price/sales (P/S) ratio is 2.6 and forward P/S is 2.3. The five-year average P/S ratio is 3.2
These valuations do not reflect an AI story should Lumentum catch the AI bid (in a bull market) the valuation could be 5-6. This may not be in the near-term given tariff related concerns, weak semiconductor sector performance and Nasdaq entering a bear market officially following the (20%) decline. However, looking into the second half of the year, should conditions improve, Lumentum is capable of trading higher.
Q&A from Earnings Call
Yield/Supply Issues are Limiting Growth
According to the earnings call, the Cloud and Networking segment is expected to increase by $25 million yet the company has “demand that far surpasses that.” Management went on to explain: “we could have probably seen a double-digit increase sequentially quarter-over-quarter had we not had some of those supply chain shortages.”
It was later more specifically called out as yield issues on the transceiver side, with this comment being in context of the lower margin: “So, yes, so during the quarter, we had some yield issues related to new product ramps within our Transceiver business. That probably was a headwind of anywhere from 100 to 150 basis points.”
This isn’t exactly surprising given the clear commentary around capex and the need to increase capacity. It was also later stated the supply issues should ease by the June quarter:
“To your second question, we're gating our module customers’ ability to grow their Transceiver business by the lack of worldwide indium phosphide for EMLs and CW lasers, quite frankly, we're having challenges actually getting enough CW lasers for our own transceivers. So that is actually impacting us on the short-term this quarter, hope to have that resolved in the June quarter. “
With even further questioning, it was revealed that hermetic packaging is also creating supply shortages, which refers to sealed enclosures that protect optical components.
“But the worldwide shortage of things like hermetic packages is creating a challenge for the kinds of volumes that our customers are looking for especially in coherent components and narrow line with lasers. And so, if we have those packages and we could get them more readily, we could grow, as Wajid said, double-digits quarter-on-quarter. It is going to hamper us in both the March quarter as well as the June quarter, and we're working diligently to minimize that impact in the June quarter, but the March quarter is what it is because we need those deliveries now in order to turn products for the quarter and we have a limited supply and ability to get that for example.”
Hyperscaler Customers
In the most recent quarter, the 10-Q states that three customers accounted for 16%, 14% and 11% of total revenue with two customers accounting for gross accounts receivable.
In the prior quarter, two customers accounted for 15% and 12% of revenue, respectively.
When comparing to the quarter a year ago, there were three customers at 19%, 13% and 11% of total revenue.
Management provided the following color in regard to how the ramp is going with the three customers:
“In Q2, Datacom transceiver revenue grew sequentially as expected, driven by an increase in shipments to our largest cloud hyperscale customer and the start of volume production shipments to one of our new customers we highlighted on prior calls.
We continue qualification work with the other new customers and expect to start initial volume production during the fourth quarter continuing to ramp through the first half of fiscal '26. Transceiver manufacturing capacity expansion is also progressing as planned.”
During the call, there was a Q&A exchange that drilled deeper into a potential fourth customer:
“Ananda Baruah:
“On the new ramping customer, can you give us some sense of time frame when you think that it hits run rate and or any context around what run rate I know this is probably kind of somewhat project based, but when you think it hits run rate, what time frame? And then the second one is sounds like this year, calendar year '25, you're not going to run into sort of congestion between your EML chips and your own transceivers since they're largely SiPho (silicon photonics) based right now.”
Lowe:
“As far as the new customer reaching run rate, I'd say that's going to take some time. So don't be raising your projections for that customer. But as I said earlier, we're qualifying a second product there that should come on by the end of the calendar year. So, I'd say around the end of the calendar year, we should be in full motion with that customer. And then as we talked about in the script, the third customer start production in fiscal Q4 and really hit run rate, I'd say, by the fall time. So, September, October. That product is scheduled to ramp significantly faster. So that's the ramp color.”
Conclusion:
Lumentum is a small-cap company with differentiated technology that has caught the attention of heavyweight NVIDIA. We’ve been closely monitoring Lumentum for roughly a year, waiting patiently for their EML lasers for 200G to ship, enabling 800G and 1.6T bandwidths. Any progress here should continue into 2026-2027 for 400G data lanes and 3.2T bandwidths.
Their indium phosphide (InP) laser technology offers significant power efficiency advantages over traditional silicon photonics. This is increasingly important as power consumption becomes a central concern in scaling AI data centers. Lumentum’s collaboration with NVIDIA, integrating their high-efficiency lasers into NVIDIA’s Spectrum-X and Quantum-X photonics networking switches, is a nod towards the EML lasers and their ultra high power lasers for CPOs playing a critical role in future architectures.
With that said, Lumentum is in the high risk bucket due to being a small cap. The stock requires speculation as to when a shift in fundamentals will occur. To date, we have seen an inflection for one quarter, yet we need more evidence before an inflection becomes a meaningful trend. The sharp acceleration provided for in analyst consensus in H2 could wane if supply chain troubles trickle down and result in slower sourcing for AI systems.
Ultimately, it’s well worth our time to earmark companies like Lumentum and identify potential entry targets.
The I/O Fund recently launched our new Discovery tier which surfaces new ideas the I/O Fund does not own yet at a pace of 30-40 new stocks per year. Coverage will include AI hardware, AI software, crypto and more, from a leading tech portfolio.
Sample research we published in March and April:
Key supplier to TSMC’s new high-growth platform called Compact Universal Photonic Engine (COUPE)
Nuclear and natural gas supplier for AI data centers
A breakdown of the risks and opportunities for the biggest IPO in the AI sector
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 has been on our radar for some time as the company supplies components for datacom transceivers and optical interconnects. Lumentum is a small-cap company with differentiated technology that has caught the attention of heavyweight NVIDIA. We’ve been closely monitoring Lumentum for roughly a year, waiting patiently for their EML lasers for 200G to ship, enabling 800G and 1.6T bandwidths. Any progress here should continue into 2026-2027 for 400G data lanes and 3.2T bandwidths.
Optical interconnects help data centers accelerate data throughput between data centers and inside the data center between servers or racks, while reducing latency and power consumption. AI is driving cloud demand higher from the hyperscalers, leading to more data being created and processed, thus helping drive a need for these interconnects to meet demand for high-speed, low power data transmission in data centers.
The company supplies three types of lasers for datacom transceivers: VSCELs, CW lasers for silicon photonics and EML-based lasers. Per our previous analysis, the 200G EMLs are what is expected to drive an inflection in 2025.
In the most recent quarter, Lumentum reported their Cloud and Networking segment grew 18.3% YoY and 20.2% QoQ to $339.2 million. This caused total revenue to grow 9.7% yet total revenue is expected to accelerate to 46.6% and 43% in the second half of the year.
More on Lumentum’s EML Lasers:
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.
In the December quarter, Lumentum stated they began shipping its 200G lane speed externally-modulated lasers (EMLs) to multiple customers. The Company stated they would increase its EML capacity by 40% YoY from June 2024 to June 2025, followed by another 40% increase by the end of calendar 2025 as its Thailand production accelerates – representing compounding growth in an 18-month time frame:
“So, we're just overall the demand is outstripping even that 40% compounded twice in an 18 month period of time. We do have additional capacity expansion beyond the end of calendar '25 obviously and those are investments that we've made over the past several quarters that come online in calendar '25, which will give us increased capacity as well.”
Lumentum’s indium phosphide 100G EMLs (Externally-Modulated Lasers) are being shipped and used in leading single-mode 400G and 800G optical transceivers. These customers are now designing the company’s 200G EMLs into their next generation of transceivers. Already being a lead supplier for 100G EML transceivers, the company is set up to be in pole position for the 200G EML transceivers.
Per the most recent earnings call: “We achieved another record for EML unit shipments in Q2 and began delivering 200G lane speed EMLs to multiple customers.
Based on the breadth of our 200G EML design wins, we expect to gain additional laser transmitter market share in the upcoming wave of 800G and 1.6T transceivers, utilizing the more efficient 200G EMLs for AI applications. Complementing our EMLs are our new 200G lens integrated photodetector arrays, which adds to our content opportunity in next generation 800G and 1.6T transceivers as well as strengthens our vertical integration strategy for our own cloud modules.”
Specifically, the company’s experience in InP long haul transceivers is being tapped as AI servers scale, especially since InP reduces power consumption compared to silicon. Lumentum is also working on higher speed optical links, including 400G per lane. This means more data, fewer lanes if you compare it to 200G per lane. This is important especially as data centers will eventually move not only to 1.6 terabits per second (1.6T) (4X400G or 8X200G) but also 3.2T (8X400G).
Management also stated the new 200G lens integrated photodetector arrays “adds to our content opportunity in next generation 800G and 1.6T transceivers.”
Lumentum’s Co-Packaged Optics Opportunity for High Power Lasers:
Silicon photonics are the only viable choice for rack-to-rack and across the data center due to the need for high bandwidth and lower power at high speeds. There is also low-loss over long distances with optical fiber, which refers to preserving the original signal, whereas copper sees signal degradation over longer distances.
Where there is a debate and an important shift occurring is in the networking between chips and inside the rack. We’ve recently covered the benefits of co-packaged optics as a replacement for pluggable transceivers when networking between chips. At 800G and 1.6T speeds, the electrical signals over power circuit boards (PCBs) run too hot, are power hungry and create loss for the signal. The overall goal is to move the optics closer in proximity to overcome scaling issues when increasing electrical speed.
Nvidia stated that by replacing pluggable optics with silicon photonics on the package, it can “deliver 3.5x more power efficiency, 63x greater signal integrity, 10x better network resiliency at scale and 1.3x faster deployment compared with traditional methods.”
Nvidia VP Ian Buck stated at GTC that the CPO switches help reduce power consumption by eliminating the need for external lasers and pluggable transceivers to achieve a significant reduction in power from 39 watts to 9 watts. Buck explained that this “gives you that benefit from going from 39 watts of power down to only 9 watts of power for the same number of ports, and that's huge. It doesn't sound like 39 sounds a lot. But if you get 400,000 GPUs in an AI supercomputer, there's like 24 megawatts of lasers like so that's a lot of laser light that could be optimized and made more efficient.”
While Lumentum may see some decline in its pluggable optics transceiver business as the industry shifts, it is also well-positioned to benefit from the transition. The company is a key supplier of high-power lasers required for co-packaged optics (CPO), which could offset near-term losses. Taken together, the impact should be at the onset net neutral — with (potentially significant) upside over time from growing demand for EML lasers, the transition to 200G and 400G per lane, plus optical switching (see below).
Regarding CPOs, here is what was stated on the most recent call: “But I'd also highlight for Lumentum a little bit of a unique situation where we're more modest share on transceivers today and share gaining, if you will, over time. So even if there is some cannibalization in the mid to long-term, we don't think it impacts our transceiver opportunity and creates an expanded opportunity for high power lasers.”
Nearly a year ago, Lumentum announced an ultra-high output power 1310 nm DFB laser in beta, stating it was designed to reduce the “required number of lasers” while boosting efficiency and reliability in large scale AI/ML infrastructure.
It was also shared that Lumentum is shipping preproduction volumes now related to CPOs:
‘As part of one collaboration, we began shipping preproduction volumes of our unique ultra-high power lasers to an AI infrastructure customer for a proprietary interconnect solution in Q2 and have received follow-on orders as well as excellent feedback on the product's performance. This is a very exciting opportunity.”
It was later teased out on the earnings call that 2026 could be a big year for Lumentum’s high power laser: “I would say preproduction volume with forecasts for more meaningful volume throughout calendar '25 and then real extremely high volume in calendar '26. So, feedback is very positive. The performance is very unique and the line width that we produce with the laser is superior to anything in the market.”
CPO (with Switch ASIC) + Optical Switch + GPU Clusters = The Future of Scale Up AI Systems
Optical switches are a new kind of switch for AI clusters that handles 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.
There are many competitors within optical switching, with heavyweights Broadcom and Arista coming to mind, yet Lumentum believes their MEMS-Based technology can set them apart. Although the discussion around MEMS can get quite technical, the idea is that Lumentum is a smaller, (potentially) key supplier for customers putting optical circuit switches into their data centers. Another use case for using Lumentum is to rearchitect or write software to enable the optical circuit switching.
According to a previous earnings call, Lumentum has “already shipped evaluation units to customers who have provided overwhelmingly positive feedback on our performance.” It was also stated that “more meaningful growth will probably be in calendar 2026” for the optical switching circuit products.
Partner for Nvidia’s Silicon Photonics & Marvell 400G per Lane
Nvidia recently announced 1.6 terabits per second port switches at GTC “to deliver 3.5x energy savings and 10x resilience in AI factories.” Lumentum was named as a partner, among others in the industry.
This analysis has covered some of Lumentum’s unique advantages and why they would be chosen as a partner, such as:
At the forefront of increasing bandwidths, from data rates of 400G, 800G, and the upcoming 1.6T
At the forefront of faster data pipelines, from 100G per lane (today), 200G per lane (ramping now and into H2 2025), and 400G per lane (into 2026 and beyond)
Indium Phosphide lasers are more expensive due to volumes at the component level compared to silicon — yet is made up for in data centers as InP reduces power consumption. InP also offers higher signal integrity.
1310nm DFB lasers are high power lasers that enable the transition to co-packaged optics.
Last month, Lumentum announced an InP DFB-MZI optical transmitter in partnership with Marvell to demonstrate 400G-per-lane PAM4. The long acronym refers to a laser-modulator combo — specifically, a DFB laser integrated with a compact Mach-Zehnder (MZI) modulator. This was paired with Marvell’s 400G PAM4 DSP, which operates at 225 Gbaud and enables the high-speed signaling needed to reach 400G per lane.
The result is a high-performance, power-efficient optical transmitter that outperforms traditional silicon photonics — “particularly in applications where signal integrity and efficiency are critical,” per the press release. The InP DFB-MZI platform sets the stage for architectures expanding to 1.6T and 3.2T bandwidth, which is in the near future, up from the 800G modules in production today.
Lumentum is Supply Constrained; Building Capacity
Lumentum’s demand far exceeds supply, including indium phosphide supply and components such as CW lasers.
The company is expanding its transceiver manufacturing capacity through the construction of a three-story facility and cleanroom in Thailand, which complements existing production lines. The first floor is completed and ready for tool installation.
As stated, Lumentum began shipping its 200G lane speed externally-modulated lasers (EMLs) to multiple customers in Q2 F2025. The Company aims to increase its EML capacity by 40% YoY from June 2024 to June 2025, followed by another 40% increase by the end of 2025 as its Thailand production accelerates.
In Q2, the Company invested $65 million in capex to expand cleanroom capacity and increase equipment capacity for InP wafer production to support EML chip manufacturing at its Thailand site. Lumentum has also been invested in expanding its indium phosphide (InP) wafer fabrication facilities, which are critical for producing high-speed lasers like the EMLs used in transceivers.
CEO Lowe reiterated this in the Q2 F2025 conference call:
“We're still on track to what we've been saying, which was 40% or higher growth from the June quarter of calendar '24 to the June quarter of calendar '25 and then another 40% by the end of calendar '25. So, that's for all 100 gig and 200 gig. I think we were in our prepared remarks, we talked about being overall supply constrained, not 200 gig because we can start a 200 gig wafer or a 100 gig wafer just the same. So, we're just overall the demand is outstripping even that 40% compounded twice in an 18-month period of time… Our wafer fab expansion plans to enable higher volumes of EMLs and other indium phosphide lasers and photodetectors continues to be on track. We still anticipate that demand for our EML chips will continue to exceed supply, at least into calendar year 2026. We are experiencing strengthening demand for our DCI products as well as long haul transmission and transport solutions.”
A Note on China Exposure
Interestingly, Lumentum does not have as much exposure to China whereas most supply chain troubles right now are due to sourcing in this problematic geo-political zone
This was stated on the last call:
“We have some production that happens in China and then those components are integrated into bigger components, bigger products at our Thailand facility. And then most of the shipments happen from Thailand, even if they are to the US or to Mexico as well. And then much of the growth that we're seeing is shipments that are coming from Japan as well as from Caswell for the transmission products that we have. So because of that, at least in the short to mid-term, unless policy changes, happen at a government level, we're not expecting much of an impact.”
This was also stated in the introduction:
“Second, we are scaling capacity for our highly differentiated laser transmitter chips in our indium phosphide wafer fabs and optical circuit switch and transceiver production capacity in our proven factories outside of China to meet the rising demand.”
Quite a lot has changed in terms of policy at the government level since the last earnings call. The overall commentary regarding China reliance being minimal on the supply chain side may be true — but tariffs could upend enough of the supply chain to affect Lumentum’s customers (and overall demand).
Also, despite not relying on manufacturing in China, Lumentum reported Hong Kong as a major customer in the 10-Q at 19% of revenue for LITE, second to United States. Geo-political tensions should cause China/HK to source domestically either voluntarily or through blacklists.
Telecom Problematic End Market yet DCIs are a Bright Spot
There has been a steep inventory correction in telecom that has led to substantial revenue decline and significant margin erosion, presenting a major fundamental headwind for data center growth to overcome. However, data center interconnects (DCIs) are helping to drive the turnaround in this otherwise problematic end market.
Lumentum offers tunable laser and coherent pluggable transceivers for data center interconnects (DCIs). These long-distance data transmissions are traditionally used for telecom purposes and can range up to hundreds of kilometers yet are now seeing demand for data center buildouts.
In April 2025, Lumentum announced the sampling of new 400/800G ZR+ L-band pluggable transceivers and the general availability of its 800G ZR+ C-band module. The L-band modules effectively double the usable wavelength range and available fiber capacity. By expanding into both C-band and L-band spectrums, Lumentum’s transceivers enable increased fiber capacity, and are able to serve AI and cloud-based applications. These pluggable modules reduce overall system complexity and cost.
Regarding DCIs, per the previous earnings call: “we're seeing dramatic strength in anything ZR, anything to connect data centers as data centers are being built out, and that can take the form of ZR modules. But given our share of tunable lasers that go into ZRs, that's where we're going to see a dramatic pickup in the telecom side.”
The most recent update from management is that demand remains very, very strong: “And so the data center interconnect and the networks that are interconnecting these data centers is showing very, very strong demand. And so DCIs and the components as well as ROADMs amplifiers and longer haul coherent transmission connecting further apart data centers is very strong.”
Revenue Growth Poised to Accelerate
Q2 FY25 revenue grew by 9.65% YoY and 19.37% QoQ to $402.2 million, driven by strength in its Cloud and Networking segment, beating estimates by 2.87%.
Management guided Q3 revenue between $410 million to $425 million, with a midpoint of $417.5 million for 13.9% growth.
Analysts expect revenue to grow 46.6% in the June quarter to $452 million and 43% YoY in the September quarter to $481.9 million in Q1 F2026.
Management also reaffirmed its commitment to reaching quarterly revenue of $500 million by the end of calendar year 2025, driven by improving trends with its networking equipment manufacturing customers.
Segments:
Cloud and Networking Drives Revenue and Margins
In the most recent quarter, Cloud and Networking grew 18.3% YoY and 20.2% QoQ to $339.2 million with management stating they saw “sequential increases in nearly all of our Cloud and Networking product lines.” Cloud and Networking segment profit grew 16.2% for an increase of 330 basis points sequentially and an increase of 610 basis points year-on-year on higher revenue. This segment includes optical transceivers, the datacom chips/lasers that go into optical modules (EMLs, VSCELs, CW lasers, etc) and telecom/data center interconnects.
Previously, in the September quarter, the segment saw a meaningful inflection with growth of 23% YoY and 11% QoQ. Next quarter, management stated they expect the segment to increase by $25 million QoQ, which would represent growth of 7.3% QoQ – not quite as high as the previous quarter yet it’s been clearly noted in our previous analysis that H2 would be the bigger ramp. Our analysis on delayed Nvidia suppliers also connects some dots on this particular timing for Lumentum.
According to management commentary, the company shipped record EML units with datacom transceivers shipping to their largest hyperscaler customer and volume production shipments to a new customer.
The company stated they expect to increase their market share:
“Based on the breadth of our 200G EML design wins, we expect to gain additional laser transmitter market share in the upcoming wave of 800G and 1.6T transceivers, utilizing the more efficient 200G EMLs for AI applications. Complementing our EMLs are our new 200G lens integrated photodetector arrays, which adds to our content opportunity in next generation 800G and 1.6T transceivers as well as strengthens our vertical integration strategy for our own cloud modules.”
Industrial Tech Continues to Contract
Q2 F2024 Industrial Tech revenue fell (21.4%) YoY and grew 15.4% QoQ to $63 million. The sequential increase was driven by higher industrial laser shipments, partially offset by lower 3D sensing shipments.
Next quarter, management expects revenues to decline sequentially by $10 million, driven by declines in both commercial lasers and 3D sensing.
Adj. Margins Bottomed Out and Accelerating
GAAP Gross margin was 24.7% with adjusted gross margin at 32.3%. Per management, company gross margins will “sequentially increase as manufacturing utilization improves as well as an increase in Datacom laser shipments.”
GAAP operating margin was (12.8%) for operating losses of ($51.6) million.
Adjusted operating margins rose to 7.9% compared to 3% last quarter. It’s expected to further expand to 10% at the midpoint this quarter.
GAAP net margin was (15.1%) for net losses of $61 million
GAAP adjusted net margin was 7.46% for adjusted profits of $30 million
Adj. EPS Returns to Growth and Accelerates Quickly
Q2 FY25 adj. EPS improved to $0.42, beating consensus estimates by 16.97%. Management guided Q3 F2025 adj. EPS between $0.47 to $0.53, with a midpoint of $0.53. There is outsized growth in EPS from the rebound on small numbers:
Analysts expect Q4 adj. EPS to grow 957.38% to $0.63
Q1 is expected to see 332.24% to $0.78
The difference between GAAP and non-GAAP operating margin is due to the stock-based compensation expenses and amortization of acquired intangibles.
Elevated Capex Spend in Thailand Manufacturing Site Pressuring Free Cash Flow
Q2 FY25 operating cash flow was $24.3 million or 6.1% of revenue. The Company spent $74 million capex in Q1 FY25 and $64 million in Q2, resulting in sequential negative free cash flow.
Free cash flow in Q2 was ($15.9 million) or -4% of revenue. The Company closed the quarter with $896.7 million in cash and cash equivalents and $2.47 billion in debt. This puts the debt-to-equity ratio at 2.75 due to high capex spending — which is high and not ideal in this environment. However, if Lumentum can prove the capex will quickly be converted to revenue, it may become a non-issue by this time next year.
Lumentum spent $64 million in capex for expanding cleanroom capacity at the Thailand manufacturing site and increasing equipment capacity for indium phosphate wafer production to support EML chip manufacturing.
Valuation
Lumentum trades at a forward price-earnings (P/E) of 31.3 and a current PE ratio of 143 (although this looks drastic given the weak bottom line the company is rebounding from).
The price/sales (P/S) ratio is 2.6 and forward P/S is 2.3. The five-year average P/S ratio is 3.2
These valuations do not reflect an AI story should Lumentum catch the AI bid (in a bull market) the valuation could be 5-6. This may not be in the near-term given tariff related concerns, weak semiconductor sector performance and Nasdaq entering a bear market officially following the (20%) decline. However, looking into the second half of the year, should conditions improve, Lumentum is capable of trading higher.
Q&A from Earnings Call
Yield/Supply Issues are Limiting Growth
According to the earnings call, the Cloud and Networking segment is expected to increase by $25 million yet the company has “demand that far surpasses that.” Management went on to explain: “we could have probably seen a double-digit increase sequentially quarter-over-quarter had we not had some of those supply chain shortages.”
It was later more specifically called out as yield issues on the transceiver side, with this comment being in context of the lower margin: “So, yes, so during the quarter, we had some yield issues related to new product ramps within our Transceiver business. That probably was a headwind of anywhere from 100 to 150 basis points.”
This isn’t exactly surprising given the clear commentary around capex and the need to increase capacity. It was also later stated the supply issues should ease by the June quarter:
“To your second question, we're gating our module customers’ ability to grow their Transceiver business by the lack of worldwide indium phosphide for EMLs and CW lasers, quite frankly, we're having challenges actually getting enough CW lasers for our own transceivers. So that is actually impacting us on the short-term this quarter, hope to have that resolved in the June quarter. “
With even further questioning, it was revealed that hermetic packaging is also creating supply shortages, which refers to sealed enclosures that protect optical components.
“But the worldwide shortage of things like hermetic packages is creating a challenge for the kinds of volumes that our customers are looking for especially in coherent components and narrow line with lasers. And so, if we have those packages and we could get them more readily, we could grow, as Wajid said, double-digits quarter-on-quarter. It is going to hamper us in both the March quarter as well as the June quarter, and we're working diligently to minimize that impact in the June quarter, but the March quarter is what it is because we need those deliveries now in order to turn products for the quarter and we have a limited supply and ability to get that for example.”
Hyperscaler Customers
In the most recent quarter, the 10-Q states that three customers accounted for 16%, 14% and 11% of total revenue with two customers accounting for gross accounts receivable.
In the prior quarter, two customers accounted for 15% and 12% of revenue, respectively.
When comparing to the quarter a year ago, there were three customers at 19%, 13% and 11% of total revenue.
Management provided the following color in regard to how the ramp is going with the three customers:
“In Q2, Datacom transceiver revenue grew sequentially as expected, driven by an increase in shipments to our largest cloud hyperscale customer and the start of volume production shipments to one of our new customers we highlighted on prior calls.
We continue qualification work with the other new customers and expect to start initial volume production during the fourth quarter continuing to ramp through the first half of fiscal '26. Transceiver manufacturing capacity expansion is also progressing as planned.”
During the call, there was a Q&A exchange that drilled deeper into a potential fourth customer:
“Ananda Baruah:
“On the new ramping customer, can you give us some sense of time frame when you think that it hits run rate and or any context around what run rate I know this is probably kind of somewhat project based, but when you think it hits run rate, what time frame? And then the second one is sounds like this year, calendar year '25, you're not going to run into sort of congestion between your EML chips and your own transceivers since they're largely SiPho (silicon photonics) based right now.”
Lowe:
“As far as the new customer reaching run rate, I'd say that's going to take some time. So don't be raising your projections for that customer. But as I said earlier, we're qualifying a second product there that should come on by the end of the calendar year. So, I'd say around the end of the calendar year, we should be in full motion with that customer. And then as we talked about in the script, the third customer start production in fiscal Q4 and really hit run rate, I'd say, by the fall time. So, September, October. That product is scheduled to ramp significantly faster. So that's the ramp color.”
Conclusion:
Lumentum is a small-cap company with differentiated technology that has caught the attention of heavyweight NVIDIA. We’ve been closely monitoring Lumentum for roughly a year, waiting patiently for their EML lasers for 200G to ship, enabling 800G and 1.6T bandwidths. Any progress here should continue into 2026-2027 for 400G data lanes and 3.2T bandwidths.
Their indium phosphide (InP) laser technology offers significant power efficiency advantages over traditional silicon photonics. This is increasingly important as power consumption becomes a central concern in scaling AI data centers. Lumentum’s collaboration with NVIDIA, integrating their high-efficiency lasers into NVIDIA’s Spectrum-X and Quantum-X photonics networking switches, is a nod towards the EML lasers and their ultra high power lasers for CPOs playing a critical role in future architectures.
With that said, Lumentum is in the high risk bucket due to being a small cap. The stock requires speculation as to when a shift in fundamentals will occur. To date, we have seen an inflection for one quarter, yet we need more evidence before an inflection becomes a meaningful trend. The sharp acceleration provided for in analyst consensus in H2 could wane if supply chain troubles trickle down and result in slower sourcing for AI systems.
Ultimately, it’s well worth our time to earmark companies like Lumentum and identify potential entry targets.
The I/O Fund recently launched our new Discovery tier which surfaces new ideas the I/O Fund does not own yet at a pace of 30-40 new stocks per year. Coverage will include AI hardware, AI software, crypto and more, from a leading tech portfolio. new Discovery tier which surfaces new ideas the I/O Fund does not own yet at a pace of 30-40 new stocks per year. Coverage will include AI hardware, AI software, crypto and more, from a leading tech portfolio.
Sample research we published in March and April:
Key supplier to TSMC’s new high-growth platform called Compact Universal Photonic Engine (COUPE)
Nuclear and natural gas supplier for AI data centers
A breakdown of the risks and opportunities for the biggest IPO in the AI sector
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 has been on our radar for some time as the company supplies components for datacom transceivers and optical interconnects. Lumentum is a small-cap company with differentiated technology that has caught the attention of heavyweight NVIDIA. We’ve been closely monitoring Lumentum for roughly a year, waiting patiently for their EML lasers for 200G to ship, enabling 800G and 1.6T bandwidths. Any progress here should continue into 2026-2027 for 400G data lanes and 3.2T bandwidths.
Optical interconnects help data centers accelerate data throughput between data centers and inside the data center between servers or racks, while reducing latency and power consumption. AI is driving cloud demand higher from the hyperscalers, leading to more data being created and processed, thus helping drive a need for these interconnects to meet demand for high-speed, low power data transmission in data centers.
The company supplies three types of lasers for datacom transceivers: VSCELs, CW lasers for silicon photonics and EML-based lasers. Per our previous analysis, the 200G EMLs are what is expected to drive an inflection in 2025.
In the most recent quarter, Lumentum reported their Cloud and Networking segment grew 18.3% YoY and 20.2% QoQ to $339.2 million. This caused total revenue to grow 9.7% yet total revenue is expected to accelerate to 46.6% and 43% in the second half of the year.
More on Lumentum’s EML Lasers:
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.
In the December quarter, Lumentum stated they began shipping its 200G lane speed externally-modulated lasers (EMLs) to multiple customers. The Company stated they would increase its EML capacity by 40% YoY from June 2024 to June 2025, followed by another 40% increase by the end of calendar 2025 as its Thailand production accelerates – representing compounding growth in an 18-month time frame:
“So, we're just overall the demand is outstripping even that 40% compounded twice in an 18 month period of time. We do have additional capacity expansion beyond the end of calendar '25 obviously and those are investments that we've made over the past several quarters that come online in calendar '25, which will give us increased capacity as well.”
Lumentum’s indium phosphide 100G EMLs (Externally-Modulated Lasers) are being shipped and used in leading single-mode 400G and 800G optical transceivers. These customers are now designing the company’s 200G EMLs into their next generation of transceivers. Already being a lead supplier for 100G EML transceivers, the company is set up to be in pole position for the 200G EML transceivers.
Per the most recent earnings call: “We achieved another record for EML unit shipments in Q2 and began delivering 200G lane speed EMLs to multiple customers.
Based on the breadth of our 200G EML design wins, we expect to gain additional laser transmitter market share in the upcoming wave of 800G and 1.6T transceivers, utilizing the more efficient 200G EMLs for AI applications. Complementing our EMLs are our new 200G lens integrated photodetector arrays, which adds to our content opportunity in next generation 800G and 1.6T transceivers as well as strengthens our vertical integration strategy for our own cloud modules.”
Specifically, the company’s experience in InP long haul transceivers is being tapped as AI servers scale, especially since InP reduces power consumption compared to silicon. Lumentum is also working on higher speed optical links, including 400G per lane. This means more data, fewer lanes if you compare it to 200G per lane. This is important especially as data centers will eventually move not only to 1.6 terabits per second (1.6T) (4X400G or 8X200G) but also 3.2T (8X400G).
Management also stated the new 200G lens integrated photodetector arrays “adds to our content opportunity in next generation 800G and 1.6T transceivers.”
Lumentum’s Co-Packaged Optics Opportunity for High Power Lasers:
Silicon photonics are the only viable choice for rack-to-rack and across the data center due to the need for high bandwidth and lower power at high speeds. There is also low-loss over long distances with optical fiber, which refers to preserving the original signal, whereas copper sees signal degradation over longer distances.
Where there is a debate and an important shift occurring is in the networking between chips and inside the rack. We’ve recently covered the benefits of co-packaged optics as a replacement for pluggable transceivers when networking between chips. At 800G and 1.6T speeds, the electrical signals over power circuit boards (PCBs) run too hot, are power hungry and create loss for the signal. The overall goal is to move the optics closer in proximity to overcome scaling issues when increasing electrical speed.
Nvidia stated that by replacing pluggable optics with silicon photonics on the package, it can “deliver 3.5x more power efficiency, 63x greater signal integrity, 10x better network resiliency at scale and 1.3x faster deployment compared with traditional methods.”
Nvidia VP Ian Buck stated at GTC that the CPO switches help reduce power consumption by eliminating the need for external lasers and pluggable transceivers to achieve a significant reduction in power from 39 watts to 9 watts. Buck explained that this “gives you that benefit from going from 39 watts of power down to only 9 watts of power for the same number of ports, and that's huge. It doesn't sound like 39 sounds a lot. But if you get 400,000 GPUs in an AI supercomputer, there's like 24 megawatts of lasers like so that's a lot of laser light that could be optimized and made more efficient.”
While Lumentum may see some decline in its pluggable optics transceiver business as the industry shifts, it is also well-positioned to benefit from the transition. The company is a key supplier of high-power lasers required for co-packaged optics (CPO), which could offset near-term losses. Taken together, the impact should be at the onset net neutral — with (potentially significant) upside over time from growing demand for EML lasers, the transition to 200G and 400G per lane, plus optical switching (see below).
Regarding CPOs, here is what was stated on the most recent call: “But I'd also highlight for Lumentum a little bit of a unique situation where we're more modest share on transceivers today and share gaining, if you will, over time. So even if there is some cannibalization in the mid to long-term, we don't think it impacts our transceiver opportunity and creates an expanded opportunity for high power lasers.”
Nearly a year ago, Lumentum announced an ultra-high output power 1310 nm DFB laser in beta, stating it was designed to reduce the “required number of lasers” while boosting efficiency and reliability in large scale AI/ML infrastructure.
It was also shared that Lumentum is shipping preproduction volumes now related to CPOs:
‘As part of one collaboration, we began shipping preproduction volumes of our unique ultra-high power lasers to an AI infrastructure customer for a proprietary interconnect solution in Q2 and have received follow-on orders as well as excellent feedback on the product's performance. This is a very exciting opportunity.”
It was later teased out on the earnings call that 2026 could be a big year for Lumentum’s high power laser: “I would say preproduction volume with forecasts for more meaningful volume throughout calendar '25 and then real extremely high volume in calendar '26. So, feedback is very positive. The performance is very unique and the line width that we produce with the laser is superior to anything in the market.”
CPO (with Switch ASIC) + Optical Switch + GPU Clusters = The Future of Scale Up AI Systems
Optical switches are a new kind of switch for AI clusters that handles 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.
There are many competitors within optical switching, with heavyweights Broadcom and Arista coming to mind, yet Lumentum believes their MEMS-Based technology can set them apart. Although the discussion around MEMS can get quite technical, the idea is that Lumentum is a smaller, (potentially) key supplier for customers putting optical circuit switches into their data centers. Another use case for using Lumentum is to rearchitect or write software to enable the optical circuit switching.
According to a previous earnings call, Lumentum has “already shipped evaluation units to customers who have provided overwhelmingly positive feedback on our performance.” It was also stated that “more meaningful growth will probably be in calendar 2026” for the optical switching circuit products.
Partner for Nvidia’s Silicon Photonics & Marvell 400G per Lane
Nvidia recently announced 1.6 terabits per second port switches at GTC “to deliver 3.5x energy savings and 10x resilience in AI factories.” Lumentum was named as a partner, among others in the industry.
This analysis has covered some of Lumentum’s unique advantages and why they would be chosen as a partner, such as:
At the forefront of increasing bandwidths, from data rates of 400G, 800G, and the upcoming 1.6T
At the forefront of faster data pipelines, from 100G per lane (today), 200G per lane (ramping now and into H2 2025), and 400G per lane (into 2026 and beyond)
Indium Phosphide lasers are more expensive due to volumes at the component level compared to silicon — yet is made up for in data centers as InP reduces power consumption. InP also offers higher signal integrity.
1310nm DFB lasers are high power lasers that enable the transition to co-packaged optics.
Last month, Lumentum announced an InP DFB-MZI optical transmitter in partnership with Marvell to demonstrate 400G-per-lane PAM4. The long acronym refers to a laser-modulator combo — specifically, a DFB laser integrated with a compact Mach-Zehnder (MZI) modulator. This was paired with Marvell’s 400G PAM4 DSP, which operates at 225 Gbaud and enables the high-speed signaling needed to reach 400G per lane.
The result is a high-performance, power-efficient optical transmitter that outperforms traditional silicon photonics — “particularly in applications where signal integrity and efficiency are critical,” per the press release. The InP DFB-MZI platform sets the stage for architectures expanding to 1.6T and 3.2T bandwidth, which is in the near future, up from the 800G modules in production today.
Lumentum is Supply Constrained; Building Capacity
Lumentum’s demand far exceeds supply, including indium phosphide supply and components such as CW lasers.
The company is expanding its transceiver manufacturing capacity through the construction of a three-story facility and cleanroom in Thailand, which complements existing production lines. The first floor is completed and ready for tool installation.
As stated, Lumentum began shipping its 200G lane speed externally-modulated lasers (EMLs) to multiple customers in Q2 F2025. The Company aims to increase its EML capacity by 40% YoY from June 2024 to June 2025, followed by another 40% increase by the end of 2025 as its Thailand production accelerates.
In Q2, the Company invested $65 million in capex to expand cleanroom capacity and increase equipment capacity for InP wafer production to support EML chip manufacturing at its Thailand site. Lumentum has also been invested in expanding its indium phosphide (InP) wafer fabrication facilities, which are critical for producing high-speed lasers like the EMLs used in transceivers.
CEO Lowe reiterated this in the Q2 F2025 conference call:
“We're still on track to what we've been saying, which was 40% or higher growth from the June quarter of calendar '24 to the June quarter of calendar '25 and then another 40% by the end of calendar '25. So, that's for all 100 gig and 200 gig. I think we were in our prepared remarks, we talked about being overall supply constrained, not 200 gig because we can start a 200 gig wafer or a 100 gig wafer just the same. So, we're just overall the demand is outstripping even that 40% compounded twice in an 18-month period of time… Our wafer fab expansion plans to enable higher volumes of EMLs and other indium phosphide lasers and photodetectors continues to be on track. We still anticipate that demand for our EML chips will continue to exceed supply, at least into calendar year 2026. We are experiencing strengthening demand for our DCI products as well as long haul transmission and transport solutions.”
A Note on China Exposure
Interestingly, Lumentum does not have as much exposure to China whereas most supply chain troubles right now are due to sourcing in this problematic geo-political zone
This was stated on the last call:
“We have some production that happens in China and then those components are integrated into bigger components, bigger products at our Thailand facility. And then most of the shipments happen from Thailand, even if they are to the US or to Mexico as well. And then much of the growth that we're seeing is shipments that are coming from Japan as well as from Caswell for the transmission products that we have. So because of that, at least in the short to mid-term, unless policy changes, happen at a government level, we're not expecting much of an impact.”
This was also stated in the introduction:
“Second, we are scaling capacity for our highly differentiated laser transmitter chips in our indium phosphide wafer fabs and optical circuit switch and transceiver production capacity in our proven factories outside of China to meet the rising demand.”
Quite a lot has changed in terms of policy at the government level since the last earnings call. The overall commentary regarding China reliance being minimal on the supply chain side may be true — but tariffs could upend enough of the supply chain to affect Lumentum’s customers (and overall demand).
Also, despite not relying on manufacturing in China, Lumentum reported Hong Kong as a major customer in the 10-Q at 19% of revenue for LITE, second to United States. Geo-political tensions should cause China/HK to source domestically either voluntarily or through blacklists.
Telecom Problematic End Market yet DCIs are a Bright Spot
There has been a steep inventory correction in telecom that has led to substantial revenue decline and significant margin erosion, presenting a major fundamental headwind for data center growth to overcome. However, data center interconnects (DCIs) are helping to drive the turnaround in this otherwise problematic end market.
Lumentum offers tunable laser and coherent pluggable transceivers for data center interconnects (DCIs). These long-distance data transmissions are traditionally used for telecom purposes and can range up to hundreds of kilometers yet are now seeing demand for data center buildouts.
In April 2025, Lumentum announced the sampling of new 400/800G ZR+ L-band pluggable transceivers and the general availability of its 800G ZR+ C-band module. The L-band modules effectively double the usable wavelength range and available fiber capacity. By expanding into both C-band and L-band spectrums, Lumentum’s transceivers enable increased fiber capacity, and are able to serve AI and cloud-based applications. These pluggable modules reduce overall system complexity and cost.
Regarding DCIs, per the previous earnings call: “we're seeing dramatic strength in anything ZR, anything to connect data centers as data centers are being built out, and that can take the form of ZR modules. But given our share of tunable lasers that go into ZRs, that's where we're going to see a dramatic pickup in the telecom side.”
The most recent update from management is that demand remains very, very strong: “And so the data center interconnect and the networks that are interconnecting these data centers is showing very, very strong demand. And so DCIs and the components as well as ROADMs amplifiers and longer haul coherent transmission connecting further apart data centers is very strong.”
Revenue Growth Poised to Accelerate
Q2 FY25 revenue grew by 9.65% YoY and 19.37% QoQ to $402.2 million, driven by strength in its Cloud and Networking segment, beating estimates by 2.87%.
Management guided Q3 revenue between $410 million to $425 million, with a midpoint of $417.5 million for 13.9% growth.
Analysts expect revenue to grow 46.6% in the June quarter to $452 million and 43% YoY in the September quarter to $481.9 million in Q1 F2026.
Management also reaffirmed its commitment to reaching quarterly revenue of $500 million by the end of calendar year 2025, driven by improving trends with its networking equipment manufacturing customers.
Segments:
Cloud and Networking Drives Revenue and Margins
In the most recent quarter, Cloud and Networking grew 18.3% YoY and 20.2% QoQ to $339.2 million with management stating they saw “sequential increases in nearly all of our Cloud and Networking product lines.” Cloud and Networking segment profit grew 16.2% for an increase of 330 basis points sequentially and an increase of 610 basis points year-on-year on higher revenue. This segment includes optical transceivers, the datacom chips/lasers that go into optical modules (EMLs, VSCELs, CW lasers, etc) and telecom/data center interconnects.
Previously, in the September quarter, the segment saw a meaningful inflection with growth of 23% YoY and 11% QoQ. Next quarter, management stated they expect the segment to increase by $25 million QoQ, which would represent growth of 7.3% QoQ – not quite as high as the previous quarter yet it’s been clearly noted in our previous analysis that H2 would be the bigger ramp. Our analysis on delayed Nvidia suppliers also connects some dots on this particular timing for Lumentum.
According to management commentary, the company shipped record EML units with datacom transceivers shipping to their largest hyperscaler customer and volume production shipments to a new customer.
The company stated they expect to increase their market share:
“Based on the breadth of our 200G EML design wins, we expect to gain additional laser transmitter market share in the upcoming wave of 800G and 1.6T transceivers, utilizing the more efficient 200G EMLs for AI applications. Complementing our EMLs are our new 200G lens integrated photodetector arrays, which adds to our content opportunity in next generation 800G and 1.6T transceivers as well as strengthens our vertical integration strategy for our own cloud modules.”
Industrial Tech Continues to Contract
Q2 F2024 Industrial Tech revenue fell (21.4%) YoY and grew 15.4% QoQ to $63 million. The sequential increase was driven by higher industrial laser shipments, partially offset by lower 3D sensing shipments.
Next quarter, management expects revenues to decline sequentially by $10 million, driven by declines in both commercial lasers and 3D sensing.
Adj. Margins Bottomed Out and Accelerating
GAAP Gross margin was 24.7% with adjusted gross margin at 32.3%. Per management, company gross margins will “sequentially increase as manufacturing utilization improves as well as an increase in Datacom laser shipments.”
GAAP operating margin was (12.8%) for operating losses of ($51.6) million.
Adjusted operating margins rose to 7.9% compared to 3% last quarter. It’s expected to further expand to 10% at the midpoint this quarter.
GAAP net margin was (15.1%) for net losses of $61 million
GAAP adjusted net margin was 7.46% for adjusted profits of $30 million
Adj. EPS Returns to Growth and Accelerates Quickly
Q2 FY25 adj. EPS improved to $0.42, beating consensus estimates by 16.97%. Management guided Q3 F2025 adj. EPS between $0.47 to $0.53, with a midpoint of $0.53. There is outsized growth in EPS from the rebound on small numbers:
Analysts expect Q4 adj. EPS to grow 957.38% to $0.63
Q1 is expected to see 332.24% to $0.78
The difference between GAAP and non-GAAP operating margin is due to the stock-based compensation expenses and amortization of acquired intangibles.
Elevated Capex Spend in Thailand Manufacturing Site Pressuring Free Cash Flow
Q2 FY25 operating cash flow was $24.3 million or 6.1% of revenue. The Company spent $74 million capex in Q1 FY25 and $64 million in Q2, resulting in sequential negative free cash flow.
Free cash flow in Q2 was ($15.9 million) or -4% of revenue. The Company closed the quarter with $896.7 million in cash and cash equivalents and $2.47 billion in debt. This puts the debt-to-equity ratio at 2.75 due to high capex spending — which is high and not ideal in this environment. However, if Lumentum can prove the capex will quickly be converted to revenue, it may become a non-issue by this time next year.
Lumentum spent $64 million in capex for expanding cleanroom capacity at the Thailand manufacturing site and increasing equipment capacity for indium phosphate wafer production to support EML chip manufacturing.
Valuation
Lumentum trades at a forward price-earnings (P/E) of 31.3 and a current PE ratio of 143 (although this looks drastic given the weak bottom line the company is rebounding from).
The price/sales (P/S) ratio is 2.6 and forward P/S is 2.3. The five-year average P/S ratio is 3.2
These valuations do not reflect an AI story should Lumentum catch the AI bid (in a bull market) the valuation could be 5-6. This may not be in the near-term given tariff related concerns, weak semiconductor sector performance and Nasdaq entering a bear market officially following the (20%) decline. However, looking into the second half of the year, should conditions improve, Lumentum is capable of trading higher.
Q&A from Earnings Call
Yield/Supply Issues are Limiting Growth
According to the earnings call, the Cloud and Networking segment is expected to increase by $25 million yet the company has “demand that far surpasses that.” Management went on to explain: “we could have probably seen a double-digit increase sequentially quarter-over-quarter had we not had some of those supply chain shortages.”
It was later more specifically called out as yield issues on the transceiver side, with this comment being in context of the lower margin: “So, yes, so during the quarter, we had some yield issues related to new product ramps within our Transceiver business. That probably was a headwind of anywhere from 100 to 150 basis points.”
This isn’t exactly surprising given the clear commentary around capex and the need to increase capacity. It was also later stated the supply issues should ease by the June quarter:
“To your second question, we're gating our module customers’ ability to grow their Transceiver business by the lack of worldwide indium phosphide for EMLs and CW lasers, quite frankly, we're having challenges actually getting enough CW lasers for our own transceivers. So that is actually impacting us on the short-term this quarter, hope to have that resolved in the June quarter. “
With even further questioning, it was revealed that hermetic packaging is also creating supply shortages, which refers to sealed enclosures that protect optical components.
“But the worldwide shortage of things like hermetic packages is creating a challenge for the kinds of volumes that our customers are looking for especially in coherent components and narrow line with lasers. And so, if we have those packages and we could get them more readily, we could grow, as Wajid said, double-digits quarter-on-quarter. It is going to hamper us in both the March quarter as well as the June quarter, and we're working diligently to minimize that impact in the June quarter, but the March quarter is what it is because we need those deliveries now in order to turn products for the quarter and we have a limited supply and ability to get that for example.”
Hyperscaler Customers
In the most recent quarter, the 10-Q states that three customers accounted for 16%, 14% and 11% of total revenue with two customers accounting for gross accounts receivable.
In the prior quarter, two customers accounted for 15% and 12% of revenue, respectively.
When comparing to the quarter a year ago, there were three customers at 19%, 13% and 11% of total revenue.
Management provided the following color in regard to how the ramp is going with the three customers:
“In Q2, Datacom transceiver revenue grew sequentially as expected, driven by an increase in shipments to our largest cloud hyperscale customer and the start of volume production shipments to one of our new customers we highlighted on prior calls.
We continue qualification work with the other new customers and expect to start initial volume production during the fourth quarter continuing to ramp through the first half of fiscal '26. Transceiver manufacturing capacity expansion is also progressing as planned.”
During the call, there was a Q&A exchange that drilled deeper into a potential fourth customer:
“Ananda Baruah:
“On the new ramping customer, can you give us some sense of time frame when you think that it hits run rate and or any context around what run rate I know this is probably kind of somewhat project based, but when you think it hits run rate, what time frame? And then the second one is sounds like this year, calendar year '25, you're not going to run into sort of congestion between your EML chips and your own transceivers since they're largely SiPho (silicon photonics) based right now.”
Lowe:
“As far as the new customer reaching run rate, I'd say that's going to take some time. So don't be raising your projections for that customer. But as I said earlier, we're qualifying a second product there that should come on by the end of the calendar year. So, I'd say around the end of the calendar year, we should be in full motion with that customer. And then as we talked about in the script, the third customer start production in fiscal Q4 and really hit run rate, I'd say, by the fall time. So, September, October. That product is scheduled to ramp significantly faster. So that's the ramp color.”
Conclusion:
Lumentum is a small-cap company with differentiated technology that has caught the attention of heavyweight NVIDIA. We’ve been closely monitoring Lumentum for roughly a year, waiting patiently for their EML lasers for 200G to ship, enabling 800G and 1.6T bandwidths. Any progress here should continue into 2026-2027 for 400G data lanes and 3.2T bandwidths.
Their indium phosphide (InP) laser technology offers significant power efficiency advantages over traditional silicon photonics. This is increasingly important as power consumption becomes a central concern in scaling AI data centers. Lumentum’s collaboration with NVIDIA, integrating their high-efficiency lasers into NVIDIA’s Spectrum-X and Quantum-X photonics networking switches, is a nod towards the EML lasers and their ultra high power lasers for CPOs playing a critical role in future architectures.
With that said, Lumentum is in the high risk bucket due to being a small cap. The stock requires speculation as to when a shift in fundamentals will occur. To date, we have seen an inflection for one quarter, yet we need more evidence before an inflection becomes a meaningful trend. The sharp acceleration provided for in analyst consensus in H2 could wane if supply chain troubles trickle down and result in slower sourcing for AI systems.
Ultimately, it’s well worth our time to earmark companies like Lumentum and identify potential entry targets.
The I/O Fund recently launched our new Discovery tier which surfaces new ideas the I/O Fund does not own yet at a pace of 30-40 new stocks per year. Coverage will include AI hardware, AI software, crypto and more, from a leading tech portfolio. new Discovery tier which surfaces new ideas the I/O Fund does not own yet at a pace of 30-40 new stocks per year. Coverage will include AI hardware, AI software, crypto and more, from a leading tech portfolio.
Sample research we published in March and April:
Key supplier to TSMC’s new high-growth platform called Compact Universal Photonic Engine (COUPE)
Nuclear and natural gas supplier for AI data centers
A breakdown of the risks and opportunities for the biggest IPO in the AI sector
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.
Amazon is building an AI powerhouse in AWS, with the segment becoming an increasingly large driver of Amazon’s earnings power despite contributing just one-sixth of Amazon’s revenue. Management remains committed to continuously expanding AI offerings and expanding GPU capacity to meet demand, seeing long-term tailwinds to its AI growth story.
E-commerce remains Amazon’s bread-and-butter, but it has now found itself caught in the crossfire of fluctuating tariff policy, being quite heavily exposed to China. Amazon could feel impacts to both revenue and margins as a result, given that peer Walmart pulled its operating income growth forecast of 0.5-2.0% and simply said it has “widened” on tariff uncertainties. For Amazon, retail margins have also improved over the past few quarters to help (minimally) offset the effects.
AWS has helped drive strong expansion in Amazon’s margins with operating profit up 61% YoY to a record $21.2 billion. With that said, this previously bulletproof segment is also exposed to the trickle-down effects of tariffs as cloud customers exposed to China will likely cut their budgets in response. Mizuho believes this could be up to 50% of cloud customers.
How much of this is priced in? That’s a question the I/O Fund is working hard to answer as the risks in the market are immense yet shares of Amazon are near historically low levels.
Building An AI Behemoth in AWS
Amazon’s approach in AI mirrors a one-stop-shop, where it is providing access to a wide range of the industry’s leading models alongside its internal models, access to the newest generations of GPUs alongside extensive custom silicon deployment, and any service that customers could need in-between. Amazon stated in Q3 2024 that it had launched nearly 2x as many genAI and ML features as the other CSPs combined over the prior 18 months.
This is what Amazon sees as its differentiator versus Azure or GCP – its ability to offer more AI/ML features and services, a wide range of powerful custom silicon and GPU instances, and end-to-end platforms for developers to build on.
Amazon’s strategy is centered around offering efficient access to AI on a price-performance basis at any size or scale, either using its custom AI accelerators for training (Trainium) or inference (Inferentia), a wide range of Nvidia GPUs, or custom CPU instances (discussed below).
Customers can test, build, and customize genAI apps on the latest LLMs from a wide range of leading providers on Amazon’s serverless platform Bedrock, while SageMaker AI is AWS’ fully managed AI/ML platform spanning the training, fine-tuning and deployment lifecycle. Customers also have access to Amazon’s genAI assistant Amazon Q, and a wide range of storage, analytics and security tools.
AWS has developed several networking innovations to improve performance and scalability for AI workloads, such as Elastic Fabric Adapter (EFA) and Nitro. EFA can communicate directly with networking hardware using a technology called operating-system bypass (OS-bypass), reducing latency and allowing AI and HPC applications to easily scale to thousands of GPU or CPU cores. Nitro offloads typical virtualization functions such as networking, storage, and management tasks to specific Nitro cards, freeing up server resources and reducing costs.
Breakdown of Amazon’s Custom Chips, Instances
AI chips are a performance game, with Nvidia quickly upping the ante with its GB200 NVL72 systems, which offer up to 30x faster inference performance on trillion-parameter LLMs. Google and Amazon are responding quickly with their custom accelerators, with Google recently unveiling its 7th-gen TPU, offering up to 10x faster AI processing than v5p. Amazon says that its next-gen Trainium3, due later this year, is 2x as fast and 40% more energy efficient than Trainium2.
Amazon boasts a wide range of custom accelerator and CPU instance offerings alongside leading Nvidia GPUs:
Trainium2 AI Accelerator Offers up to 40% More Price Performance
Amazon launched its Trainium2 chips and instances in November 2023, built primarily for AI training tasks, offering up to 4x the performance of its first generation Trainium chip. Trn2 instances combine 16 Trainium2 chips to offer 20.8 petaflops of dense FP8 compute. Compared to its next-most powerful EC2 instances, Trn2 offers “30% more compute and 25% more high bandwidth memory.”
Amazon says that the Trn2 are “typically 30% to 40% more price performant than other current GPU-powered instances available.” For example, Databricks uses Trn2 instances to help “lower TCO by up to 30% for its customers,” while DataDog says the instances help users “cut AI infrastructure costs by up to 50% and boost model training and deployment performance.”
Next-Gen Trainium3 Set to Launch Later in 2025
Amazon’s next-generation Trainium3 chip is set to debut later in 2025, and are expected to offer up to 4x the performance of Trn2 UltraServers. Amazon has been sparse on details for the new chip, aside from the performance boost, speed upgrade and energy efficiency improvements. AWS CEO Matt Garman said the new clusters will let customers “iterate even faster when building models and deliver superior real-time performance when deploying them.”
Inferentia Tailored for Fast, Low Cost Inference
Amazon’s Inferentia chips, first launched in 2019, were tailored specifically for AI inference applications and aimed to tackle the barrier of high inference costs, which at the time could account for up to 90% of the infrastructure cost for building and deploying an AI/ML app.
The first-generation chip offered up to 2.3x higher throughput and 70% lower cost than comparable EC2 instances on inference tasks. Each chip featured 4 first-generation NeuronCores with 128 TOPS performance for lower precision FP16, BF16, and INT8 calculations, scaling up to 16-chip instances offering 192 GB memory for heavy inference tasks. Inferentia helped Amazon deliver high-throughput, low cost AI inference due to the chip’s inference-optimized design and low instance cost.
Inferentia2 Provides 10x Lower Latency
Inferentia2 was built for large-scale AI inference workloads, with Amazon saying customers could effectively deploy 175B parameter models for inference workloads on a single Inf2 instance, which features 16 Inferentia2 chips.
Inferentia2 can deliver up to 4x higher throughput and up to 10x lower latency compared to Inferentia, as it has 4x more memory capacity and 16.4x higher memory bandwidth than Inferentia. In addition, Inf2 are the first inference-optimized instances to support scale-out distributed inference for large-scale LLMs with high-speed connectivity between chips.
Inf2 is a core part of Amazon’s strategy to lead in AI inference as inference costs come down, offering low-latency, low-cost inference-optimized performance. Inf2 is said to cost ~$0.40 per 1 million tokens for a 70B model, versus ~$1.00 per 1 million for an H100 GPU ran on Azure. AWS customers can also build across Amazon’s full stack, Trn2 and Inf2, for quick, efficient and cheap AI training and inference tasks.
Graviton4 CPU Offers 40% Better Price Performance
AWS’ Graviton CPUs are its in-house Arm-based custom data center chips, with the most recent Graviton4 launched alongside Trainium2 in 2023. Amazon says the next-gen Graviton4 chips can offer up to “nearly 40% better price performance versus other leading x86 processors,” with 30% better performance, 50% more cores and 75% more memory bandwidth than the Graviton3.
Amazon has built more than 2 million Graviton chips to date since the first generation launched in 2018, with a wide range of instances powered by the chips. AWS offers EC2 instances for general purpose compute, as well as compute optimized, memory optimized, or storage optimized instances.
Unique ‘Burstable’ T3 Instances
Amazon’s EC2 T3 instances are a low-cost, burstable instance that offers balanced compute, memory, and network resources primarily for general purpose workloads. What’s unique is that T3 instances provide a baseline level of performance from the CPU and an ability to ‘burst’ CPU usage to access more compute and reach full-core performance.
The unique ‘burst’ factor of T3 instances is built upon credits, which are earned when running below baseline or when idle. As needed, at any time, credits can be used to throttle to maximum performance until credits run out, or for as long as required with customers responsible for the additional cost.
New UltraServers for Scale-out Compute
UltraServers are a new offering for Amazon, where it is linking four Trn2 instances to form a 64-chip cluster (4 Trn2 instances) for 83.2 petaflops of dense FP8 compute, utilizing its proprietary chip-to-chip interconnect NeuronLink. UltraServers aim to accelerate the training process further, letting customers deploy larger models faster – Amazon says customers could train 300B parameter models in just weeks as opposed to months.
UltraServers are then linked together to form UltraClusters, letting customers scale-out to tens of thousands up to 100,000 Trainium2 chips. This is the backbone of Amazon’s Project Rainier supercluster project for key partner Anthropic.
Amazon Spending Same Amount as AWS’ Run Rate on Capex
Amazon has the benefit of becoming an AI behemoth due to AWS’ massive scale – the unit recently crossed a $115 billion annualized revenue run rate, more than doubling in the past four years, with revenue up 19% YoY to $28.8 billion in Q4. Unlike Microsoft, Amazon has not provided exact AI revenue figures for AWS, stating that it was at a “multi-billion dollar annualized revenue run rate” and growing triple digits YoY in Q4.
AWS has grown, and continues to grow, significantly quicker than Amazon as a whole. Ten years ago, AWS generated just $4.6 billion in revenue versus $89 billion for Amazon. Since then, AWS revenue has grown nearly 24x while Amazon has not even grown 7.5x.
AWS is also estimated to have nearly 4.2 million business customers, an enormous base to fuel continued growth in cloud services and AI. However, HG Insights estimated that 92% of these customers spend less than $1,000/month on AWS services, which might be an issue down the line considering that AI’s earliest adopters and largest spenders are likely to be in the Fortune 500.
Q4’s call also featured an important quote from CEO Andy Jassy about AI’s future at Amazon:
“I spent a fair bit of time thinking several years out. And while it may be hard for some to fathom a world where virtually every app has generative AI infusing it with inference being a core building block just like compute, storage and database, and most companies having their own agents that accomplish various tasks and interact with one another. This is the world we're thinking about all the time and we continue to believe that this world will mostly be built on top of the cloud with the largest portion of it on AWS.”
Amazon is buying fully into this vision, indicating that it is spending nearly AWS’ revenue amount in capex in 2025 to capture growing AI demand. Jassy explained that annualizing Q4 2024’s $26.3 billion capex would be a reasonably good way to view 2025’s capex. This implied Amazon is eyeing capex of ~$105 billion, up ~27% YoY after rising ~57% YoY in 2024. This would also be practically double Amazon’s $52.8 billion in capex from 2023, and comes after Amazon announced in early 2023 that it would be putting $150 billion towards new data centers over the next 15 years.
Amazon Quickly Outpacing Hyperscaler Peers When it Comes to Capex
We’ve closely tracked Big Tech’s capex for many quarters as a growth indicator for Nvidia and AI hardware suppliers. Amazon has been the largest spender of the hyperscalers (Amazon, Microsoft, Meta, Alphabet) over the last few years, accounting for nearly one-third of total capex in 2023 and 2024 and is set to be the top spender again in 2025.
On a quarterly basis, Amazon’s capex accelerated significantly beginning in Q2 2024, rising above $15 billion for the first time before ending the year $10 billion higher. Amazon far outspent Google and Meta, though Microsoft was a heavy spender as well in the second half of 2024.
Amazon is accelerating its capex as it continues to witness elevated AI demand that outstrips its available capacity. Jassy said both in Q3 2024, and in Amazon’s shareholder letter published in April 2025, that there was more demand AWS could fulfill even if they had more capacity, and that the “faster demand grows, the more datacenters, chips, and hardware we need to procure.” This suggests that AWS is continuing to see high levels of demand to give it the confidence to commit to $105B+ in capex for 2025.
AWS is monetizing this capex “many months” after it is spent and “over many years,” suggesting that 2024’s acceleration to 19% growth is likely from monetizing investments from 2023 with Nvidia’s Hopper GPU generation. This also suggests that the surge in capex in 2H 2024 has not been monetized yet and AWS will recognize the growth benefits through 2025, especially as component constraints ease in the second half of the year.
Data from Omdia shows Amazon was a significant deployer of both Nvidia’s GPUs and custom silicon in 2024, purchasing ~196K Hopper GPUs and deploying a combined 1.3 million Trainium and Inferentia chips (comparable to ~430K Hoppers). For 2025, Nvidia showed that the top CSPs (Amazon, Microsoft, Alphabet, Oracle) have already ordered 3.6 million Blackwell GPUs, versus 1.3 million at Hopper’s peak, with Amazon likely a large purchaser of the new generation.
Based on management’s comments about the monetization timeline, it’s possible that Amazon is only just beginning to recognize the growth tailwinds from these large investments and GPU deployments. This also sets the stage for consistent, strong growth through 2026 as Blackwell comes online along with Trainium3.
Despite this, there are still risks on the horizon –the recent tariff-fueled market turbulence has led to some heightened fears about the state of cloud spending, given that some customers may undergo budget optimization efforts in an effort to absorb higher costs in the coming months. For example, Mizuho’s James Lee predicts that up to 50% of cloud customers may eye budget reductions this year and be more “hesitant while assessing the economic impact before redeploying their capital.”
A Note on Amazon’s AI Pricing, 3P Model Support versus Azure, GCP
AWS is working to differentiate itself from Azure and Google Cloud (GCP) in two ways – offering a wide range of AI model support and on price-performance.
AWS offers access to more than 100 leading foundation models, from DeepSeek, Mistral, Meta, and others, as well as access to Anthropic’s models, Amazon’s internal models, and bring-your-own model support in Bedrock and SageMaker. Azure is tightly coupled with OpenAI due to Microsoft’s partnership, though it does also offer access to leading open-source models such as those from DeepSeek and Meta; however, not all of OpenAI’s leading models are accessible in every region. Google offers access to its full suite of models with broad open-source support, though usage of Google models is locked to GCP.
Amazon offers this range of model support to its entire custom silicon family of chips and Nvidia GPUs, while Azure lets customers build primarily on Nvidia’s ecosystem as its custom silicon effort is not nearly as extensive as AWS’. GCP mirrors Amazon with flexibility between TPUs and Nvidia GPUs.
For chip pricing, AWS and GCP can both offer custom silicon instances for fraction of the cost of Nvidia GPU instances, while also offering cheaper prices for inference workloads. A report from CloudExpat estimated that Amazon’s Trn1 instances cost ~$1.34 per hour versus $1.20/hour for GCP’s TPU v5e and $12.84/hr for an H100 on Azure; for inference, an Inferentia instance was estimated at $0.40 per 1M tokens, versus $0.30 on GCP and $1.00 for the H100 on Azure.
When it comes to running hosted inference on a model, such as for Meta’s Llama 405B, where costs range from as low as $6 per 1M tokens on Fireworks.ai to ~$21.33 on both Azure and AWS, all the way to $30 on Snowflake, Amazon’s performance sets it apart. While smaller platforms like Fireworks may hit users with ‘spike arrests’ where usage is throttled, Amazon can offer compellingly fast performance and token generation due to its breadth of instances. So for a model like Llama 3.1 405B, AWS says that independent inference performance tests “showed that Amazon Bedrock, running on Trn2 instances, delivers more than 3x higher token-generation throughput compared to other available offerings by major cloud providers.”
Amazon is extending its competition with Microsoft to the AI startup sphere, backing OpenAI competitor Anthropic with a total investment so far of $8 billion in the Claude developer.
Anthropic is quickly adding new models with improved capabilities and new features as it targets both consumer and enterprise AI demand as competition heats up. The startup unveiled an AI Research feature, which conducts multi-step autonomous research, with in-line citations and an ability to “synthesize findings and deliver holistic, source-backed summaries.” It also launched deeper integrations with Google Workspace, where Claude can now connect directly to Gmail, Google Calendar, and Google Docs, allowing it to scan emails, summarize docs, identify meetings, and surface files requested.
Anthropic also recently introduced new subscription plans for Claude, building on top of its $20/month Claude Pro plan and supplementing its free tier. The new Max plan for $100/month offers 5x the usage as Pro, while the $200/month option offers up to 20x the usage. Additionally, Anthropic is said to be preparing to roll out a new “voice mode” AI assistant that users can speak to, as early as this month, following ChatGPT’s footsteps. The company is also working on a new major ‘hybrid’ model that can switch between deep reasoning and quick responses, offering developers a sliding scale customization tool to shift speeds and help reduce costs.
Amazon is working to support Anthropic’s future growth with Project Rainier, a massive supercomputer being developed primarily to serve Anthropic’s growing compute needs. AWS is also optimizing its platform to offer faster access to Claude, with Bedrock offering a new "latency-optimized mode" for Claude 3.5 Haiku which runs 60% faster on Trainium2 instances.
Project Rainier, in a way, is Amazon’s internal competitor to OpenAI and Oracle’s Project Stargate, as the supercomputer cluster, uniquely housed in multiple facilities in different locations linked by Amazon’s Elastic Fabric Adapter, will feature hundreds of thousands of its Trainium2 chips. Rainier is also unique in that it will be built entirely with Amazon’s custom chip stack, while heavily leveraging AWS’ Nitro system and its Neuron SDK for maximum performance.
Rainier will be composed thousands of UltraServers linked together to form an UltraCluster accessing those chips. UltraClusters will help bring training times from weeks down to just a few days, according to AWS, as customers can access thousands of chips in tandem for large-scale training tasks.
When completed, it will be one of the largest clusters ever built globally, though specifics about the exact amount of chips, cost and scale are still secret. Amazon is no stranger to large-scale cluster development though, having collaborated with Nvidia on Project Ceiba, a 20,736-Blackwell GPU cluster offering 414 exaflops performance.
Investing in Anthropic can be seen as a net positive for Amazon despite the large capital outlays in its upfront investment and in Rainier – not only does Amazon have a guaranteed, leading AI customer to utilize the massive compute cluster it’s building, but it’s also receiving “intense feedback” from the startup on its custom accelerators and how to better optimize each new generation for performance or cost.
AWS Margins Very Strong, but Risks Ahead
AWS has historically enjoyed a strong operating margin profile that has strengthened considerably since early 2023, nearing the 40% range.
In Q4, AWS reported an operating margin of 36.9%, down from its peak of 38.1% in Q3. This was a substantial improvement from a 29.6% margin in Q4 2023. Management noted that the strength of margins was due to strong growth and cost control – operating income for AWS has risen >30% YoY in each of the last six quarters, and 48% YoY in Q4. However, quarterly fluctuation is expected depending on the level of investment Amazon is making in the segment.
On a TTM basis, AWS had only briefly surpassed 30% in the beginning of 2022 prior to a macro-induced growth slowdown that dented margins. Since then, AWS has managed to substantially improve its profitability, with TTM operating margin at 37% in Q4, up more than 12 points from 24.7% in Q2 2023.
Should AWS continue to see operating income grow in the double-digits YoY, it could possibly reach the 40% threshold by the end of the year, providing a strong tailwind to EPS growth as it contributed half of Amazon’s total operating income in Q4.
AI, Useful Life Classifications May Present Margin Headwind
Management has talked about the AI versus non-AI margins, with AI margins being significantly lower due to the massive investments Amazon is undertaking at the moment. CFO Brian Olsavsky explained that AI “does come originally with lower margins and a heavy investment load,” and in the short-term that will be a headwind to margins, but over the long-term, he expects “margins will be comparable in non-AI business as well.”
This is likely due to a much lower revenue return per dollar of spending presently, given that AI is only at a multi-billion dollar run rate. TD Cowen analysts put this in perspective, estimating that AWS historically has generated $4 in incremental revenue for every $1 of capital spending, but with surging AI investments, the ratio is now likely ~$0.20 cents for every $1. TD Cowen expects the incremental revenue to reapproach its usual $4 in the next several years.
Another risk to operating margins that lies ahead is useful life calculations for servers, primarily that useful lives may continue to decrease as the pace of GPU upgrades accelerates, as the performance gaps between each generation widens.
Amazon noted that AWS’ operating margin in Q4 benefited from a ~200 bp YoY positive impact from increasing the estimated useful life of servers in 2024. Stripping out this impact, operating margin would’ve been 34.9%.
However, Amazon said that in Q4, they “completed a useful life study for our servers and networking equipment and observed an increased pace of technology development,” and as a result decreased the useful life for a “subset” of servers and networking from 6 years to 5 years. Management estimated that this change would decrease FY25 operating income by ~$700 million.
Amazon also retired a subset of servers and networking equipment early, recording a $920 million accelerated depreciation expense, which they estimate will also negatively impact FY25 operating income by ~$600 million. Both of the impacts are expected to primarily be felt in AWS. This combines for a $1.3 billion negative headwind to operating income, which would be about a (2.7%) impact assuming 20% YoY growth in operating income.
However, Nvidia is pushing ahead with a break-neck product upgrade cycle for its GPUs, maintaining an annual cadence that rival AMD is working to match. Each generation upgrade, such as from its Hopper generation to its Blackwell generation, promises significant leaps in compute and thus performance.
These rapid performance upgrades could mean that current depreciation schedules still do not account for how quickly older generations of GPUs become obsolete simply from an inability to remain competitive performance-wise, whether that be in sheer compute or performance-per-watt. Should depreciation schedules more accurately be in the 3 to 4 year range, recognized depreciation expenses would be much larger, providing another headwind to operating margins.
E-Commerce in Tariff Territory
On the e-commerce side, Amazon is facing risks from increased tariff uncertainty, considering its rather high exposure to China. Online stores revenue growth has leveled off in the high-single digits, while third-party (3P) seller services revenue has decelerated rather sharply over the last few quarters. In light of the rising China risk, Amazon has also pushed forward with a large US warehouse expansion plan.
According to Bloomberg, Amazon canceled several inventory orders from China after tariffs on the country were raised to 104%. The orders were said to contain numerous consumer-oriented products, such as air conditioners and scooters. Additionally, Amazon has high China exposure from first-party vendors, which account for ~40% of its items; Morgan Stanley estimates that 25% of the cost of goods sold from 1P vendors comes from China. There are also reports that Chinese vendors are considering hiking prices or leaving the marketplace due to the high tariffs.
Tariffs also could weigh on 3P vendors, as higher import prices could squeeze margins and again lead to price hikes to offset higher costs. Amazon is said to have offered price concessions to some vendors with high-demand products in order to mitigate potential impacts.
In light of the heightened tariff risk, Amazon is doubling down on its domestic US business, as it is reportedly looking for capital partners for a $15 billion project to expand its warehouse footprint. The plan would call for construction of 80 new logistics facilities, with the majority being delivery hubs and a few being multi-story, larger-scale fulfilment centers.
North America revenue did reaccelerate to 10% YoY in Q4 to $115.6 billion, though growth has slowed over the past few quarters – Q2 and Q3 both saw the segment grow just 9% YoY, versus 11% YoY in the same period in 2023.
International revenue decelerated rather sharply in Q4, with growth of 9% YoY to $43.4 billion, down from 12% growth in Q3. This would mark the segment’s slowest growth since the start of 2023.
Segment Breakdown:
Decelerations in 3P seller services (which includes commissions, fulfillment and shipping fees) and advertising occurring simultaneously raise red flags for Amazon heading into Q1, as it suggests that its seller flywheel may be losing some momentum as tariff risks escalate. The softness of 3P seller services noted below hints that marketplace volume may be slowing, while the slowdown in advertising growth (in the holiday quarter) could mean sellers are spending less on advertising due to weaker sales performance, tighter margins or lower product demand.
Amazon’s online store revenue was $75.6 billion in Q4, up 8% YoY, maintaining the same growth pace from Q3 yet accelerating from the 6% growth seen in Q1.
On the other hand, 3P Seller Services growth has decelerated dramatically, from 19% in Q4 2023 to just 9% in Q4 2024.
Advertising growth has mirrored that deceleration, from 26% in Q4 2023 to just 18% in Q4 2024.
Recent Strengthening in E-commerce Margins Provides Some Cushion for Tariffs
Tariff risk looks to be amplifying some underlying weaknesses in 3P seller services and advertising; two signals of marketplace volume and demand. However, Amazon does have levers it can pull to mitigate these effects, as it is recognizing cost savings in inventory management and has a much stronger operating margin profile now that can absorb some costs.
Amazon’s management explained in Q4 that some genAI applications they built for inventory management have led to “10% better forecasting on our part and 20% better regional prediction,” and improvements in robotic fulfilment efficiency have combined for “significant productivity and cost savings.” Management sees more opportunities to further reduce costs via more refined inventory placement, expansion of same-day delivery networks, and accelerated robotics and automation throughout the fulfillment process.
On the margin side, Amazon has seen solid improvement in both North America and International that should offer it some leeway when it comes to absorbing tariff impacts:
North America operating margin reached 8.0% in Q4, up nearly 2 points YoY, while TTM operating margin reached 6.4%, up 2.2 points YoY.
In dollar terms, North America’s TTM operating income rose 68% to $25.0 billion, or a $10.1 billion YoY gain.
International operating margin was 3.0% in Q4, a 4 point YoY improvement from (1.0%), while TTM operating margin was 2.7%, a notable improvement from (2.7%) last year.
Revenue Guide Misses Estimates
Amazon’s soft Q1 guidance pointed to growth in the high-single digit range, a sharp sequential deceleration and what would be the first single-digit print in the last eight quarters.
For Q1, Amazon projected revenue between $151 billion and $155.5 billion, for 5% to 9% YoY growth; which was well below the consensus estimate for $158 billion.
Amazon said the softness was partially due to an “unusually large, unfavorable impact of approximately $2.1 billion” from foreign exchange rates (150 bp YoY growth), as well as the lack of leap-year impact, which added $1.5 billion of revenue (or 120 bp YoY growth impact). Even backing out both headwinds, growth would still be projected in the single-digits using the midpoint of the range. It also likely does not account for increased macro turbulence on the retail side as the full breadth and severity of tariffs on key trading partners was not known when guidance was provided.
At the midpoint of 7% YoY growth, this would be Amazon’s slowest quarterly growth rate since Q3 2001. This also does not account for some of the escalated tariff risks and seller impacts that have arisen throughout April. Analysts expect Amazon to fare much better, with the current consensus estimate calling for 8.2% growth to $155 billion in revenue, at the upper end of the guided range. Analyst estimates range from $153.2 billion to $157.2 billion, with none of the 45 analysts expecting Amazon to report below 7% growth this quarter.
EPS Growth Outpacing Revenue on Margin Strengths
Despite the weak revenue guide and sequential deceleration, Amazon’s margin strengths, primarily from AWS but also from improvements on the e-commerce side, are aiding robust earnings growth.
Amazon is currently estimated to generate $1.38 in EPS in Q1, up nearly 39% YoY, or more than 5x the rate of revenue growth at midpoint. EPS has rebounded from 2023’s lows, rising 216% in Q1 2024 and >50% YoY in each quarter of 2024.
Q1’s EPS estimate does face some risks from tariffs impacting margins or weaker revenue growth, while broader macro risks in consumer spending are becoming more prevalent. Consumer spending fell by the most in four years in January, and while it rebounded in February, much of the rebound was driven by price increases as inflation-adjusted spend barely rose. Consumer spending is being closely watched as consumers remain a bit more cautious, with Richmond Fed President Richard Barkin saying that while spending is not yet showing “troubling signs of a decline,” it is the metric he is most closely watching as the “trigger” on the economy.
EPS growth is forecast to slow rather dramatically as Amazon continues to face tough comps and slower growth – each quarter in 2025 is expected to see sub-10% revenue growth. Q2 and Q3 are expected to see EPS growth of just 12% to 13%, while Q4 runs the risk of falling flat, with estimates pointing to just 2% YoY growth.
For 2025, EPS growth is projected to outpace revenue growth by 5 points at just over 14% YoY, as Amazon benefits from the improved margin base it built throughout 2024, which drove EPS growth of 91% YoY last year.
Gross margin is on the verge of breaking 20% as high-margin AWS and advertising increase their revenue share, while operating margin sharply expanded to the double-digits for the first time, up from the low single digits at the start of 2023. Should Amazon be able to drive and maintain an operating margin >40% for AWS, it is well on the path to see mid-20% gross margins and mid-teens operating margins over the next few years.
While Amazon is lapping strong growth this year, EPS growth is expected to accelerate through FY27 on the back of this margin strengthening. FY26 EPS growth is currently projected at 19% YoY before accelerating to nearly 25% YoY in FY27, representing a 10 point acceleration in two years.
From FY25 to FY27, EPS growth is currently around a 19% CAGR. With the way that AWS is quickly reshaping Amazon’s margin profile and putting it on a trajectory for mid-teens operating margins, Amazon could likely see EPS growth accelerate to low to mid-20% CAGR simply from the growing margin and profit tailwinds from AWS.
Operating Cash Flows Remain Robust, FCF Impacted by Capex
Amazon has driven significant growth in operating cash flow, though FCF growth has been minimal due to accelerating capex.
Amazon generated $115.9 billion in operating cash flow in 2024, up 36% YoY. OCF margin expanded 3.4 points to 18.2%. This has been a remarkable turnaround from 2022, where TTM OCF dropped to just $35.6 billion.
However, FCF growth has stalled, with Amazon reporting minimal FCF generation in the first three quarters of 2024 due to sharply increased capex. For 2024, FCF rose just 4% YoY to $38.2 billion, but on a TTM basis, FCF has declined sharply from its peak at $53 billion in Q2. This was because Amazon reported just $4.7 billion in OCF in Q3 and $5.1 billion in Q1, or margins of just 3% and 3.5% respectively.
While FCF’s growth trajectory had tracked OCF rather closely through 2021 all the way to 2024, FCF has detached and began declining due to capex acceleration. With management signaling ~$26 billion in capex each quarter in 2025, it’s likely that FCF generation and growth will lag OCF through next year.
Valuation at Historic Lows
Amazon is trading at historically low valuations on an earnings and cash flow basis, while it continues to benefit from improving operating leverage as AWS and ads help drive margins and earnings growth higher.
Amazon is trading at a <29x forward earnings, at 2024 and below 2022 lows, which would be the lowest forward P/E ratio the company has traded at in more than a decade. It’s also well below the 36-38x multiple it commanded through much of 2024, and a ~30% discount to its 5-year average forward P/E of 39.4x.
Given the strong growth in OCF, it is no surprise that Amazon is trading at its lowest P/OCF multiple at just 17x. This is also a 30% discount to its 5-year average P/OCF multiple of 25.7x, and a 20% discount to retail peer Walmart, which is trading at 20.6x OCF for just 2% YoY growth compared to Amazon’s 36%.
While Amazon is trading at historic lows for forward P/E and P/OCF, its top-line multiples have been consistently expanding as high-growth, high-margin AWS and advertising continued to grow their share of revenue and boost Amazon’s earnings power. Amazon is currently trading at 2.8x forward P/S, nearly double its multiple from early 2023 but in line with its 5-year average at 2.7x.
Conclusion
E-commerce made Amazon famous, yet its AWS segment is quietly and quickly becoming a driver of Amazon’s growth story.
Amazon is positioning AWS to become an AI behemoth with a focus on offering everything customers will need to fully harness AI efficiently and cheaply. AWS continues to launch new Trainium and Inferentia instances to offer a broader range of powerful, affordable training and inference-optimized compute resources to its customers. It also provides access to 100+ leading models and bring-your-own model support, while launching 2x more AI features than other CSPs combined from early 2023 through Q3 2024. Project Rainier is showcasing AWS’ powerful in-house chips and networking, aiming to scale up to hundreds of thousands of chips housed in different facilities as one of the largest GPU clusters built to date.
Capex spending hints that growth for AWS could remain strong through 2025 and into 2026 as Amazon begins to monetize 2024’s accelerating investments. While the e-commerce side bears the brunt of potential tariff impacts, Amazon’s valuation has gotten attractive recently as P/E and P/OCF multiples have fallen toward historic lows.
Want exclusive, fast-paced deep-dive research on 30-40 stocks per year in the market’s hottest sector, from AI to data center, software, crypto and more, from a leading tech portfolio? This is what you’ll get with the I/O Fund’s new Discovery tier.
Sample research we published in March and April:
Key supplier to TSMC’s new high-growth platform called Compact Universal Photonic Engine (COUPE)
Nuclear and natural gas supplier for AI data centers
A breakdown of the risks and opportunities for the biggest IPO in the AI sector
Current Pro and Advanced Members: To subscribe to Discovery with 40% off, please click here to email usclick here to email us or email premium@io-fund.com and mention code DISCOVERY40.
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.
Marvell reported strong Q2 FY2025 results. The company beat the revenue consensus estimates by 1.5% and non-GAAP EPS estimates by 2.3%. Revenue grew 10% sequentially, and the guide for the next quarter beat estimates by 2.8%. The guide suggests the company returns to growth from FQ3.
AI led the way, with data center revenue growing 92% YoY. The consumer segment revenue recovered, more than doubling sequentially, and management believes that enterprise networking, carrier and auto and industrial end markets have bottomed in the second quarter. All the end markets are expected to grow in Q3.
Revenue
FQ2 revenue declined by (5.1%) YoY and grew by 10% sequentially to $1.27 billion, primarily due to solid growth in the data center end market led by AI. This compares to a (-12.2%) YoY decline in FQ1. Next quarter, management has guided for a revenue to accelerate to 2.2% YoY growth to $1.45 billion at the midpoint. This beat the consensus estimates by 2.8% and the positive price action further implies we will see analysts raise consensus.
Matt Murphy Chairman and CEO said in the earnings call, “For the third quarter, we are forecasting consolidated revenue to grow 14% sequentially at the midpoint of guidance. We expect this growth to be primarily driven by data center AI, and further augmented by the start of a recovery in our enterprise networking and carrier end markets.”
Margins
Margins are improving. The higher custom ASIC revenue will weigh on the gross margins. However, operating leverage and non-recurring engineering (NRE) cost benefits will help to improve the bottom line, particularly the management pointed to the next fiscal year.
The gross margin was in line with the guide at 46.2%, up from 38.9% in the same period last year and 45.5% in the previous quarter. Management has guided for 47.2% in the next quarter.
Adjusted gross margin is 61.9%, up from 60.3% last year and down from 62.4% in the previous quarter. The management guide for the next quarter is 61%, down sequentially primarily due to the lower margins for custom ASIC business.
The operating margin was (-7.9%), up from (-15.3%) last year and (-13.1%) in the previous quarter. The management guide for the next quarter is (-0.6%). Adjusted operating margin was 26.1%, down from 26.9% in the last year and up from 23.3% in the previous quarter. It beat the guide marginally by 0.5%. The guide for the next quarter is 28.9%. The improvement in margin is due to operating leverage and also non-recurring engineering (NRE) cost benefits.
Net loss was ($193.3 million) or (-15.2%) of revenue compared to a net loss of ($207.5 million) or (-15.5%) of revenue in the same period last year. Adjusted net income was $266.2 million or 20.9% compared to $290.2 million or 21.6% last year.
EPS
Adjusted EPS was $0.30, beating estimates by 2.3% compared to $0.33 in the same period last year. The management GAAP loss per share guide for the quarter is ($0.09) +/- $0.05 and adjusted EPS guide is $0.40 +/- $0.05, representing a YoY decline of (-2.4%).
Analysts expect adjusted EPS to accelerate to 5.2% growth in Q4 and 108.7% in Q1 FY2026.
Cash Flow and Balance Sheet
The cash flows have improved when we compared to last year. This is positive particularly since the company has high debt.
Operating cash flow was $306.4 million or 24.1% of revenue compared to $112.5 million or 8.4% of revenue in the same period last year and 28% in the previous quarter.
Free cash flow was $253 million or 19.9% of revenue compared to $1.2 million or 0.1% in the same period last year and 20% in the previous quarter. The cash flows were lower in the last year due to higher days sales outstanding of 82 days compared to 76 days in the recent quarter and also higher severance-related cash restructuring charges last year.
Cash was $808.7 million and debt of $4.13 billion compared to $847.7 million and $4.15 billion in the previous quarter.
Inventory was $818 million compared to $826 million in the previous quarter.
The company paid $52 million in dividends and repurchased shares worth $175 million. The company expects to increase share repurchases further in FQ3.
Key Segments
Data Center
Data Center end market grew by 92% YoY and 8% sequentially to $880.9 million led by strong AI revenue.
Matt Murphy said in the earnings call, “These above-guidance results were driven by strong demand for our electro-optics products, custom silicon beginning its anticipated ramp, as well as growth in our storage and switch revenue. Strong bookings continue for our market leading 800 gig PAM products and 400ZR data center interconnect, or DCI products, and we are looking forward to starting shipments of our next-generation 200 gig per lane, 1.6 terabit DSPs in the third quarter. As a result, we expect our electro-optics revenue will continue to grow every quarter this fiscal year on a sequential basis.”
The management also highlighted that custom silicon business is moving in the right direction and custom silicon customers are here to stay. “Our AI custom silicon programs are progressing very well with our first 2 chips now ramping into volume production. Development for new custom programs we have already won, including projects with the new Tier 1 AI customer we announced earlier this year, are also tracking well to key milestones.
Looking ahead to the third quarter of fiscal 2025 for our data center end market, we are forecasting revenue growth to accelerate into the high teens sequentially on a percentage basis. We expect the largest contributor to this growth will be our AI custom silicon programs as they begin to ramp meaningfully in the third quarter, further augmented by ongoing growth from our optics portfolio.
Although custom has a lower gross margin than our merchant products, it benefits from inherently lower operating expense levels, given NRE offsets from customers and the sharing of IP with our merchant business. As a result, as custom silicon becomes a larger part of our overall revenue, we see a path for operating expenses as a percentage of revenue decreasing below our current target operating model.”
Carrier Infrastructure
Carrier Infrastructure revenue declined by (-72%) YoY and up 6% sequentially to $75.9 million. Management expects aggregate revenue from carrier infrastructure revenue and enterprise networking to grow sequentially in the mid-single digits in the next quarter and further improve in the fourth quarter.
Enterprise Networking
Enterprise Networking revenue was down (-54%) YoY and (-1%) sequentially to $151 million. Management expects aggregate revenue from carrier infrastructure revenue and enterprise networking to grow sequentially in the mid-single digits in the next quarter and further improve in the fourth quarter. After several quarters of inventory correction, the company is seeing signs of growth in both Carrier Infrastructure and Enterprise Networking.
Consumer
Consumer end market was down (-47%) YoY and up 112% sequentially to $88.9 million following the gaming inventory correction. Management expects revenue to grow slightly on a sequential basis in the next quarter.
Automotive/Industrial
The automotive and industrial end markets revenue declined by (-31%) YoY and (-2%) sequentially to $76.2 million due to the broad inventory correction in the automotive end market. Management expects auto and industrial end market to grow sequentially in the mid-single digits in the next quarter.
Earnings Call:
No Update on AI Revenue:
The CEO had stated the following at the beginning of the call: “Given the strong start in the first half of the fiscal year from AI and our expectations for accelerated growth in the second half, we remain confident in our ability to significantly exceed the full year AI revenue target discussed earlier this year at our AI event” – yet, there was no update to the $1.5 billion floor provided for this year and the $2.5 billion floor for AI revenue provided for next year. Analysts poked quite a bit to get an update, but to no avail, with the CEO stating: “we're not calling those numbers out typically by quarter.”
Here is an example of what transpired in the Q&A, to where the CEO insisted the number was higher yet did not provide specifics:
Question Quinn Bolton (Analysts)
Matt, I'll ask a question, but if you don't answer it, maybe I'll follow up. You guys are talking about nice upside, the $1.5 billion and $2.5 billion target for AI. Is that something you think you're closer to $2 billion than $1.5 billion when all is said and done this year? I mean, can you give us any sort of quantification of the upside in AI revs? And if not, I'll follow up with a product question.
Answer Matthew Murphy (Executives)
Yes. I don't think we're — we're sort of fresh off the $1.5 billion update from a few months back when we had our AI day. But I think if you look at the — even like Q3, right, where overall revenue for the whole company is growing in mid-teens and then obviously guiding up data center much higher than that with AI driving it and then saying also Q4 is going to be extremely strong in data center and AI, you can probably draw a line of sight to it. But we're clearly, clearly exceeding the $1.5 billion. That's for sure. And then again, the [indiscernible] for next year is really good, because from an exit standpoint, we'll be at a very healthy level by the fourth quarter.
1.6T Transceivers Ramping in Q3:
The earnings report provided a surprise in terms of the 1.6T transceivers ramping in Q3. We covered this recently here (and a few other times in archived Marvell analysis on our site). Per our previous analysis: “Marvell offers 200-gig, 400-gig and 800-gig PAM-based electro-optics. The 800-gig is the primary interconnect for AI deployments. The company is qualifying a 1.6T solution with 200-gig per lane for the next leg up in AI acceleration. For the 1.6T solution, Nvidia will be a lead partner.”
Here is an update from the call which points toward next year being a big year for Marvell on electro-optics:
Question Christopher Rolland (Analysts)
Congrats on the results. I have an Inphi question primarily. And Matt, you talked about 1.6T. That's pretty exciting that, if I understood that correctly, you're going to be ramping in 3Q. Perhaps you can talk about the profile of this ramp, what it looks like in 3Q, 4Q and into '25. I think The Street's all over the place on this node. But if you could talk about that and maybe even the economics, what it means for you guys, that would be fantastic.
Answer Matthew Murphy (Executives)
Yes. Thanks, Chris. Yes, it's still early. We were first to market in this area, having introduced these products at OFC a year ago. It's great to see that they're going into production. We'll know a lot more when we start to see how our customers are planning to deploy it. But we are seeing initial shipments now. The way I would think about it though is 800 gig right now and into — and through next year is still going to be the workhorse high-volume platform. And even some of the newer launches that are coming, as an example, are going to still support 800 gig as well as 1.6T.
So it's really hard to call exactly. It's going to be an important transition. There's no question. The only question is the timing of which. And Chris, I think we'll be in a better position to comment on that probably as we get closer to the end of the year, we start looking at the setup for calendar '25 and what our customers are thinking. But either way, we're going to be well positioned for both of those opportunities for next year.
– End Quote
Conclusion:
Marvell is bottoming and we are looking for an entry. We can’t say right now if we will enter this quarter or enter closer to calendar year 2025. Although the rebound is a step in the right direction, the lack of an update on AI revenue leads us to the assumption there isn’t much of an update to report. Today there are stronger AI stories, whereas next year Marvell could rival some of the AI stories we own today.
Royston Roche, Equity Analyst at the I/O Fund, contributed to this article.
Marvell is a stock we watch closely and a stock we ultimately want to own due to its abundant AI potential, yet the company’s AI potential is buried by other segments in a steep, cyclical trough. Looking ahead, next year has the makings of a solid comeback for this often-overlooked AI stock, with revenue growth set to return as early as Q3 before accelerating strongly in fiscal 2026 (CY2025).
Marvell reports fiscal Q2 earnings on August 29, with AI revenue at the forefront after management hinted at exceeding its previously set $1.5 billion AI revenue floor. This will be the key metric to watch alongside data center revenue, which is Marvell’s only end market segment with both YoY and QoQ growth at the moment. We see potential for ASICs to become a meaningful part of data center build-outs, with Big Tech touting performance and cost advantages from custom silicon hosted in the cloud, and we’re closely tracking both Broadcom as the leader and Marvell as the runner-up in the space.
Revenue
Marvell guided for $1.25 billion in revenue in the quarter, for a YoY decline of (6.5%) and QoQ growth of 7.8%. This is a notable acceleration from (12.2%) YoY in fiscal Q1, where Marvell reported $1.16 billion in revenue.
This acceleration is expected to persist through the second half of fiscal 2025, with consensus estimates for Q3 pointing to $1.41 billion, for a YoY decline of just (0.8%), a 570 bp acceleration from Q2, and QoQ growth of 12.8%. Q4 is currently estimated at $1.59 billion, for YoY growth of 11.5% and QoQ growth of 12.8%.
However, the real growth story lies in fiscal 2026, with Q1 through Q3 all currently estimated to record >30% YoY growth.
For fiscal 2025, Marvell is expected to report $5.4 billion in revenue, or a (2%) YoY decline. Revenue growth is expected to sharply accelerate to 32.2% YoY to $7.14 billion in fiscal 2026.
Key Metrics
As is the case with other AI semiconductors, Marvell’s data center segment drove Q1’s growth, nearly doubling YoY and the only end market segment with both YoY and QoQ growth. For a refresher on Marvell’s key products, refer to our prior analysis “Marvell: Tons of AI Potential Obscured by Underperforming Segments.”
Data center is Marvell’s strongest segment, with $816.4 million in revenue in Q1, up 87% YoY and up 7% QoQ. This came on the heels of another historic data center quarter of $765.3 million, up 54% YoY and up 38% QoQ. Management guided for mid-single-digit sequential growth in the segment as ASICs continues to ramp, implying revenue of ~$865 million in Q2 at ~6% QoQ growth. Electro-optics and interconnects were primary growth drivers in Q1, with initial custom silicon shipments starting, so we’re expecting to see more ASICs in the mix this quarter.
Outside of the data center, Marvell’s other segments are showing deep YoY declines.
Carrier infrastructure revenue declined (75%) YoY and (58%) QoQ to $72 million in Q1, with Q2 guided to be near flat sequentially, with management saying the recovery will be harder to predict.
Enterprise networking declined (58%) YoY and (42%) QoQ to $153 million in Q1, with Q2 similarly expected to be near flat sequentially, before recovering in the second half of this fiscal year.
Consumer revenue declined (70%) YoY and (71%) QoQ to $42 million in Q1, dragged down from softness in the gaming market. However, management believes the gaming inventory correction has cleared, with the segment expected to double on a sequential basis in Q2.
Automotive revenue declined just (13%) YoY and (6%) QoQ to $78 million, and once more is expected to be flat sequentially, with growth resuming in the second half of the fiscal year.
Margins
Margins are a bit lower historically as volume production of ASICs ramp this year, with management “expecting a very substantial ramp in the second half of this year, followed by a full year of high volume production in fiscal 2026.” This will weigh on gross margins in the near term, but will help to drive stronger operating margins, primarily due to non-recurring engineering costs.
Marvell guided for GAAP gross margin to be 46.2%, a 70 bp QoQ improvement from 45.5% in Q1 and a 730 bp YoY improvement. Adjusted gross margin was guided to be 62%, down 40 bp QoQ from 62.4% in Q1 but up 170 bp YoY.
Management provided some context on the path of gross margins, saying that the sequential decrease in adjusted gross margin stems from a shift in product mix as consumer revenue rebounds and ASICs ramp. Management added that the “substantial” ramp in H2 in ASICs “is likely to be dilutive to our current gross margins, but to be accretive to operating margin and earnings.”
GAAP operating margin for Q2 was guided at (8.8%), an improvement from (13.1%) last quarter and (15.3%) last year. While it is a step in the right direction, Marvell still has a lot of ground to cover to return to GAAP operating profitability. Adjusted operating margin is expected to be 25.6% in Q2, up 230 bp sequentially.
EPS
Though Marvell did not provide any guidance on net margins for Q2, we can infer from management’s EPS guides that net margin is likely to marginally improve sequentially.
GAAP EPS is guided to be ($0.20), +/- $0.05 for Q2, compared to ($0.25) in Q1, a slight improvement at midpoint, though the lower end of the range comes in line with Q1’s result.
Adjusted EPS is guided to be $0.29, +/- $0.05, compared to $0.24 in Q1, again a slight improvement at midpoint with the low end of the range being flat to Q1. This reflects a YoY decline of (11.1%), before improving to (7.5%) in Q3 and returning to bottom-line growth of 5.1% in Q4.
For FY2025, Marvell is projected to report $1.40 in adjusted EPS, for a YoY decline of (7.2%).
The real growth story for Marvell’s bottom line arises in FY2026, once ASICs reaches full volume production, as noted previously by management to be accretive to both operating margin and EPS. Q1 FY26 is currently estimated to see Marvell report nearly 108% YoY growth in adjusted EPS to $0.50.
On a full year basis, FY2026’s adjusted EPS growth is currently estimated at 71.9% YoY to $2.41, reflecting the significant operating leverage opportunity that lies ahead from ramping ASICs to full volume production.
Cash Flows and Debt
Cash flows have remained strong for Marvell despite the end market weakness in a majority of its segments aside from AI; however, debt is one concern as debt-to-EBITDA ratios are high.
Operating cash flow in Q1 was $324.5 million for a margin of 28%, while free cash flow was $232.5 million, for a margin of 20%. This is lower than usual due to annual employee cash bonuses.
Inventory was $826 million in Q1, decreasing $38 million QoQ and $200 million YoY. Days sales outstanding decreased 8 days to 69 days.
Cash and equivalents totaled $847.7 million at the end of Q1, down from $950.5 million at the end of the prior quarter, primarily due to the employee bonuses.
Debt totaled $4.15 billion. Marvell’s gross debt-to-EBITDA ratio at the end of Q1 was 2.27x and net debt-to-EBITDA was 1.8x.
What to Expect for AI Revenue
Marvell’s AI revenue will be closely watched in Q2, given that other AI semiconductors in AMD and Broadcom both recently raised AI revenue outlooks for this year. Marvell had guided for $1.5 billion in AI revenue exiting the fiscal year, setting a floor for next year at $2.5 billion.
Marvell had given clues that ASICs AI revenue is trending around $500 million per quarter in Q1, and offered more color on the full year outlook:
CEO Matt Murphy: “If you look at our Q2, most of the growth in the data center segment is coming from custom. So that's a positive. And then, the whole thing inflects meaningfully in the second half and I'd say from a full year perspective, the way to think about it, maybe some additional color would be, we talked about a floor of $1.5 billion for AI revenue for Marvell for this fiscal year with about two-third in electro-optics and a third in custom. And we see now both of those exceeding that number.”most of the growth in the data center segment is coming from custom. So that's a positive. And then, the whole thing inflects meaningfully in the second half and I'd say from a full year perspective, the way to think about it, maybe some additional color would be, we talked about a floor of $1.5 billion for AI revenue for Marvell for this fiscal year with about two-third in electro-optics and a third in custom. And we see now both of those exceeding that number.”
On the topic of AI revenue growth in fiscal 2026 (calendar 2025), management said:
“Our programs continue to kind of upsize in terms of the magnitude we're looking at. I'd just say we didn't give the breakout exactly for next year, but we did talk about incremental $1 billion is the floor for next year from this year, so going from $1.5 billion to $2.5 billion.we did talk about incremental $1 billion is the floor for next year from this year, so going from $1.5 billion to $2.5 billion.
And a lot of that is going to be due to the custom programs hitting their first full year of volume. So we're not calling out the exact split for next year, but maybe that will help you triangulate in the middle in terms of where we are today, where we're trying to drive the business and then also where we see the overall AI business for next year. “And a lot of that is going to be due to the custom programs hitting their first full year of volume. So we're not calling out the exact split for next year, but maybe that will help you triangulate in the middle in terms of where we are today, where we're trying to drive the business and then also where we see the overall AI business for next year. “
At the moment, analysts are already looking above that guided range. For example, JP Morgan believes Marvell’s AI revenues will easily surpass management’s $1.5 billion floor and rise to $1.6 to $1.8 billion in calendar 2024, before soaring more than 70% YoY in 2025 to $2.8 billion to $3.0 billion, as ASICs programs ramp at Amazon and later Microsoft.
Valuation
Marvell is currently trading above historical norms on top-line and bottom-line valuations, but it’s likely the market is pricing Marvell at this premium on expectations that custom silicon will ramp and accelerate ahead of schedule, and drive margin improvement and stronger bottom line growth.
Marvell is currently trading just above 11x sales, on both a TTM and forward basis, given the low growth environment for the current fiscal year. This is approximately a 30% premium to Marvell’s 3-year average forward revenue multiple of 8.6x; however, this 11x revenue multiple has been Marvell’s average for 2024.
On the bottom-line, Marvell trades near the higher end of its range over the past three years, given that margins down the line have struggled and not yet recovered. Marvell trades at nearly 50x forward adjusted EPS, much higher than its 3-year average of 34x. Marvell is expected to grow into this multiple in FY26 on the back of strong adjusted EPS growth as ASICs ramp.
Conclusion
On one hand, Marvell’s rebound and growth story might be too early, and there will be a couple of less-spectacular earnings reports until we get to the rebound in 2025. On the other hand, Marvell may start to move quicker than current consensus is forecasting as it’s participating in a few explosive trends, namely ASICs and data center optics.
How the market perceives Marvell’s non-AI segments until a material recovery arises is anyone’s guess. In a risk-on environment, these segments will be dismissed and the number of times a management team mentions the words “AI” on an earnings call is all that matters. In a risk-off environment, Marvell’s unfortunate exposure to telecom and gaming will mute upside. As it stands, we remain cautiously optimistic on Marvell.
We will cover the report for our premium members briefly BMO on Friday.
Damien Robbins, Equity Analyst at the I/O Fund, contributed to this article.
GPU sales are surging at the moment, primarily from Big Tech’s $200 billion in capex for AI infrastructure services. Critical data center components, including networking, are required for GPU systems. There is a well-quoted discussion from Dell executives earlier this year, that by the end of the systems lifecycle, $2 to $3 will be spent on networking and storage for every $1 spent on GPUs. Granted, the effects of this 2-X market demand will be spread across many more players compared to GPUs. Yet, there is ample evidence that networking is sparking a remarkable growth trajectory of its own. For example, Nvidia’s InfiniBand has seen triple digit growth, Cisco has provided strong AI commentary, and Arista Networks’ view is that networking is mission-critical to improve GPU utilization.
Arista Networking is positioning itself as a pure-play in AI-driven networking, and management sees tailwinds to growth not only via Ethernet establishing itself as the go-to choice in networking, especially for AI training, but also stemming from the broader market opportunity arising from the massive shipment volumes of Nvidia’s GPUs.
Why Data Center Spend Is Accelerating
The current AI landscape is spearheaded by Big Tech. Microsoft and Amazon are touting multi-billion-dollar cloud revenue run rates from AI, Google sees a clear path to monetizing AI features, and Meta is aggressively investing in AI with some initial evidence it’s boosting average revenue per user (ARPU). What’s unfolding is an AI ‘arms race’, in that Big Tech, and a handful of startups including OpenAI, Mistral, Anthropic, and others, are competing to develop, deploy and commercialize the next cutting-edge AI model. This race also spreads over to who can develop the best AI assistants/Copilots, increase adoption of GenAI tools, and accelerate revenue growth in the cloud.
Nvidia CEO Jensen Huang explained exactly why this race is rapidly unfolding, and why Big Tech’s AI expenditures are increasing not only this year, but likely for the next few years: “time is really, really valuable to them. Let me give you an example of time being really valuable, why this idea of standing up a data center instantaneously is so valuable and getting this thing called time to train is so valuable. The reason for that is because the next company who reaches the next major plateau gets to announce a groundbreaking AI. And the second one after that gets to announce something that's 0.3% better. And so the question is, do you want to be repeatedly the company delivering groundbreaking AI or the company delivering 0.3% better? And that's the reason why this race, as in all technology races, the race is so important. And you're seeing this race across multiple companies because this is so vital to have technology leadership, for companies to trust the leadership and want to build on your platform and know that the platform that they're building on is going to get better and better.”
The conclusion is that Big Tech firms are snapping up Nvidia’s GPUs as fast as they reach the market, and this demand spills over into AI servers and networking components, as both are crucial for AI systems.
Networking Becoming Indispensable
Big Tech is deploying thousands to (soon) millions of GPUs and in-house AI accelerators, and networking is a mission-critical piece. Switches are crucial for communication inside the GPU clusters, allowing quick, efficient communication and transfer of data between each node, which is essential for parallel processing, and thus overall job completion time when it comes to training large-scale AI models and completing larger workloads.
A cohesive networking layer with high-quality switching technology can lower power consumption needs significantly and improve performance job completion times. As a result, the industry is going all-in on Ethernet, Nvidia included, in part due to its compatibility, cost and performance advantages, and security.
According to Arista, Ethernet has advantages over Nvidia’s InfiniBand: “AI workloads are placing greater demand on Ethernet, as they are both data and compute-intensive across thousands of processes today. Basically, AI at scale needs Ethernet at scale. AI workloads cannot tolerate the delays in the network, because the job can only be completed after all flows are successfully delivered to the GPU clusters. All it takes is one culprit of worst-case link to throttle an entire AI workload.”
Broadcom’s management seconded this, explaining that as companies scale GPU clusters, they are “going to have to use the best networking technology. And we believe that the best networking technology is Ethernet.” Broadcom’s Ram Velaga added that whether GPUs are “connected inside a data center or across data centers, you cannot get around the fact that you have to connect multiple GPUs. Once you accept the fact that it is a distributed computing problem and you need a network, then I would make a very strong case for you that the best network in the world, over multiple generations, over and again, has been Ethernet.”
Velaga used Meta as an example as to why networking (and Ethernet) is so important: when Meta is running “large workloads, anywhere between 18% to 57% of the time, the traffic is just sitting in the network. That means during this period of time, the GPUs are actually sitting idle. Now think about it. If on an average somebody is charging somebody between $20,000 to $30,000 per GPU and you've got 100,000 GPUs, you're talking about a $2 billion to $3 billion infrastructure. And if $2 billion to $3 billion infrastructure is sitting idle for 18% to 57% of the time, that's a lot of money, right?”
By creating more efficient lines of communication between GPUs, Ethernet can help accelerate job completion times, and in turn, allow more jobs to be completed on the clusters. Velaga touched upon the performance advantages of Ethernet versus Nvidia’s InfiniBand, noting that Meta tested both on a 24,000 GPU cluster and found that Ethernet provides up to 10% better performance at half the cost, which, when translated over to overall infrastructure costs, could equal hundreds of millions to billions saved.
Nvidia is also prioritizing Ethernet with its new Spectrum-X solution, despite seeing strong triple-digit networking revenue growth (accelerating from 94% YoY to 242% YoY in 4 quarters to over $3 billion in quarterly revenue) driven by InfiniBand. CEO Jensen Huang said Nvidia is “all-in on Ethernet” with an “exciting road map coming.” He added that “Spectrum-X is ramping in volume with multiple customers, including a massive 100,000 GPU cluster. Spectrum-X opens a brand-new market to NVIDIA networking and enables Ethernet only data centers to accommodate large-scale AI. We expect Spectrum-X to jump to a multibillion-dollar product line within a year.”
For more information on how Ethernet compares to InfiniBand, reference our analysis: “Broadcom: Networking/ASICs Giant and The Second Largest by AI Revenue” where we go through a side-by-side comparison including why Big Tech is pushing for Ethernet over Nvidia’s in-house InfiniBand.Broadcom: Networking/ASICs Giant and The Second Largest by AI Revenue” where we go through a side-by-side comparison including why Big Tech is pushing for Ethernet over Nvidia’s in-house InfiniBand.
Arista Confident in Ethernet Opportunity
Arista echoed much of Broadcom’s comments on Ethernet’s performance advantages versus InfiniBand, reiterating that Ethernet is “proving to offer at least 10% improvement of job completion performance across all packet sizes versus InfiniBand.”
Arista added that they are “witnessing an inflection of AI networking and expect this to continue throughout the year and decade” as Ethernet emerges as “a critical infrastructure across both front-end and back-end AI data centers.”
Per Arista’s Q1 call, we are “progressing well in four major AI Ethernet clusters that we won versus InfiniBand recently,” and in the four clusters, they are “migrating from trials to pilots, connecting thousands of GPUs this year, and we expect production in the range of 10K to 100K GPUs in 2025.”
Arista’s management reiterated the company will reach an AI target of $750 million in 2025. Barclays believes Arista “can top its guidance of $750M in AI-related back-end revenue for 2025,” commanding 18% market share in data center switching (versus 27% share for Nvidia and 22% for Cisco).
Arista also has a few core risks, particularly in that revenue is heavily concentrated in two major customers, Microsoft and Meta, accounting for 38% of company-wide revenue ($2.2 billion of Arista’s $5.8 billion in revenue in 2023). While we are seeing both companies spending quite aggressively in AI this year and next, revisions to capex plans intra-year (much like how we saw Meta increase its full year capex guide last quarter) can move Arista’s stock price.
Microsoft and Meta accounted for 39% of Arista’s revenue in 2023, down slightly from 42% in 2022 but up significantly from less than 25% in 2021. In dollar terms, the two are both billion-dollar customers, with revenue from Microsoft increasing more than 50% in 2023 and nearly 59% in 2022.
Both Microsoft and Meta increased 2024’s capex plans, with Microsoft’s Q1 capex rising 80% YoY to $14 billion, and full fiscal year capex up 50% YoY to $50 billion. Microsoft is reportedly seeking to triple its GPU supply this year to 1.8 million GPUs to support AI demand on Azure, and demand for networking components should rise hand in hand.
Meta boosted its full year capex range to $35-40 billion, pointing to 33% YoY growth and $4 billion more than previously anticipated, to build out AI infrastructure and support its internal AI roadmap. Meta’s Q1 capex was only $6.7 billion, implying that the bulk of this spend will hit in the second half of the year, possibly accelerating at a ~20% QoQ rate and exiting 2024 above the $11 billion range – hinting that Meta’s contributions to Arista’s revenue growth may not be felt in full force until the back half of 2024.
While the capex growth is positive, competition in the networking space is high, from Broadcom to Nvidia to Cisco to others. Cisco noted that it has been seeing strong momentum in Ethernet AI fabric deployment at three of the top four hyperscalers, possibly alongside Arista’s solutions, while Nvidia has recently released its Spectrum-X Ethernet solution which it expects to become a multibillion-dollar product line with a year. According to Nvidia, Spectrum-X delivers 1.6X better networking performance than traditional Ethernet.
Our previous Broadcom analysis points toward the Ethernet networking giant being second in AI revenue, primarily from AI networking revenue. Forward-looking, Broadcom is expected to end the year with $2.75 billion per quarter in AI revenue for $11B per year. Compare this to Nvidia’s networking revenue at $3.2 billion.
InfiniBand increases dependency on Nvidia, requires a new networking stack, and lags Ethernet on raw bandwidth. Improvements in Ethernet systems are expected to offer better load balancing and congestion control to help close the gap with InfiniBand’s low latency. Broadcom’s Jericho3-AI switch platform is the company’s AI fabric that competes with InfiniBand on AI training completion times, and it allows for more than 32,000 GPUS to be linked for a massive AI training system.
Regarding Arista, the company has partnered to offer a holistic solution, where a remote Arista-based AI agent will help customers optimize and manage their AI clusters with a single control point; however, investors should expect competition in the market to remain fierce as hyperscalers continue to build out data center infrastructure.
Turning to fiscal Q1’s earnings — Arista delivered a solid report, with revenue ahead of expectations as margins remained strong. However, headline revenue growth has decelerated rather quickly as Arista faced difficult comps in Q1. Despite the deceleration, the bottom line remains strong and in fact is strengthening.
In terms of AI revenue, management did not provide a figure for 2024, but its $750 million target for 2025 would represent close to 10% of total revenue, with consensus for FY2025 at $7.82 billion.
Revenue and EPS:
Arista reported revenue of $1.57 billion in Q1, representing YoY growth of 16.3% and beating expectations by nearly $24 million. This is down from 54% growth in the year-ago quarter. Growth is expected to decelerate more than 4 percentage points on a sequential basis in Q2.
GAAP EPS of $1.99 represented YoY growth of 44.2%, beating estimates by $0.40.
For Q2, Arista guided revenue between $1.62 billion and $1.65 billion, representing YoY growth of 12.1% at midpoint, a fifth consecutive quarter of decelerating revenue growth. However, Q2 is expected to mark the bottom, with analysts expecting growth to reaccelerate to the 14%+ range by Q4.
Margins:
Gross margin was 63.7% in Q1, down 120 bp QoQ from a six-year high in Q4 at 64.9%. Gross margin has been relatively stable in the 63% to 64% range aside from a dip to the 60% range in the first half of 2023.
Operating margin reached a record high at 42.0% in Q1, and represented a 50 bp QoQ and 610 bp YoY expansion. Operating margin has expanded steadily since 2021, increasing nearly 10 percentage points from the low 30% level.
Net margin was 40.6%, up 80 bp QoQ and 830 bp YoY. Arista’s bottom line strength should not be overlooked, especially as leverage improves down the line even with decelerating revenue growth. In the market’s top AI stocks at the moment, Arista has one of the strongest bottom lines outside of Nvidia.
Cash and Debt:
Arista reported cash and equivalents of $5.45 billion, and has zero debt.
Cash flow margins are strong — operating cash flow increased over 37% YoY to $514 million, for a 32.7% margin. Free cash flow also increased 37% YoY to $504 million, for a 32% margin.
How Arista Networks Ranks on AI Revenue:
Here’s a quick glance on the rankings for AI networking revenue, with Arista near the bottom of the list. Nvidia and Broadcom lead the sector, with Nvidia recently surpassing a $13 billion annual run rate in networking.
Nvidia is in first with $3.2 billion in networking revenue in fiscal Q1, after recently surpassing a $13 billion annual run rate in Q4. Nvidia is expecting its new Spectrum-X product to reach a multi-billion dollar run rate “within a year.”
Broadcom is in second place with AI revenue of $3.1 billion in fiscal Q2, projecting an annual run rate of more than $11 billion, or more than $2.75 billion per quarter. Based on management’s commentary, networking likely contributed upwards of $1 billion in the quarter, and could exit the year in the mid-$4 billion range.
Marvell reported approximately $500 million in AI revenue in the most recent quarter, with management eyeing a “a floor of $1.5 billion for AI revenue” for this fiscal year, with two-thirds coming from electro-optics and one-third from ASICs.
Juniper Networks reported $321.2 million in the AI enterprise segment, or 23.5% of revenue. Recently, it was announced that Juniper is being acquired by HPE.
Arista has not broken out AI revenue on a quarterly basis yet, but is targeting $750 million in AI revenue in 2025, which is equivalent to ~10% of revenue for that year.
Valuation:
Arista’s top line valuation has surpassed historical peaks. Bottom line strength and improved operating leverage are driving increased earnings power exiting 2024 with a bottom line valuation in line historically.
Arista currently trades at 19.4x sales and 17.3x forward sales, both above historical highs – in late 2021, Arista peaked at just under 17x sales, the same level where it pulled back from in early May following its post-earnings rally. Buying in the 8x sales range offers a higher probability for upside, although notably, AI stocks have not offered low entries over the past year.
On the bottom line, Arista trades slightly below 52x earnings and nearly 47x forward earnings, which on the surface is expensive, and above its 5-year average of 34x. Arista peaked at 60x earnings at its 17x sales peak multiple in late 2021, providing a tiny bit of breathing room for the bottom-line valuation on improved earnings power later this year and through next; however, downside risk is more prevalent as both the top and bottom-line valuations become stretched.
There are currently two counts that I see playing out in this final push higher. Both counts have us in the final swing higher of the larger 3rd wave, they only differ on how high this swing can go before we see a larger pullback:
Technical Analysis
By Knox Ridley
There are currently two counts that I see playing out in this final push higher. Both counts have us in the final swing higher of the larger 3rd wave, they only differ on how high this swing can go before we see a larger pullback:
Green – If we can breakout over $390 and hold this level, then the odds favor this path higher. This would target the $440 – $480 region directly. The final target should be between $490 – $520 before we see the larger 4th wave pullback.
Red – If we fail to breakout over $390, and instead see a breakdown below $335. This would signal that we are in the 4th wave drop, which I generally have targets between $260 – $200. If this plays out, as long as we hold $195, this would be a decent buying opportunity for the final 5th wave swing higher.
Conclusion
InfiniBand has been reporting up to 500% growth and more recently 300% growth, yet for the first time since the AI surge, Nvidia reported that networking declined sequentially this past quarter.
Despite Nvidia’s GPU moat being fully intact, its networking lead is in question. Gartner recently reported that by 2028, 80% of hyperscalers will “opportunistically” prefer Ethernet over proprietary technologies. According to Broadcom, this shift is happening quickly with management predicting that as soon as next year “all mega-scale GPU deployments will be on Ethernet.”
Arista Networks is a stock to watch in this space, primarily for its defensibility on the bottom line. The company sees $750 million in back-end AI revenue in 2025 stemming from growth among hyperscalers. Heavy revenue concentration in Meta and Microsoft is a core risk to watch, yet capex spend is accelerating for the time being, providing growth tailwinds for the networking industry as a whole. Next week, we will revisit another Ethernet networking play with more revenue and a stronger growth story for 2025, despite being weaker on the bottom line.
Damien Robbins, Equity Analyst at the I/O Fund, contributed to this article.
We recently covered Big Tech’s massive capex plans for 2024, with Microsoft, Meta, Amazon and Alphabet likely spending close to or upwards of $200 billion this year, and signaling a high likelihood of increasing capex in 2025. These spending plans are primarily for AI infrastructure, and while a majority of this spend is likely flowing to GPU leader Nvidia and then downstream to AI servers from Super Micro and Dell, the data center networking component shouldn’t be overlooked.
Arista Networks is a major supplier to Meta, and provided upbeat guidance this quarter. Lumentum, an optical and photonics components manufacturer did, as well, signaling an important opportunity for revenue growth from the data center. Management said they are “making excellent progress on the huge opportunities the long-term demand for data center photonics creates for us, driven by the exponentially increasing compute requirements of artificial intelligence, machine learning, and advanced data centers.”
Lumentum is in the high-risk bucket due to being a small cap, and because it requires speculation as to when a shift in fundamentals will occur. The stock will be reserved for the Advanced tier’s momentum portfolio for now, until we see more fundamental strength. This means technical analysis plays a primary role. We will close the position quickly (as soon as one day) if the setup fails. Or, we will hold for many months and increase its allocation if we see the stock shift to meet more of our criteria.
Background on Optical Interconnects
Optical interconnects help data centers accelerate data throughput between data centers, inside the data center between servers or racks, while reducing latency and power consumption. AI is driving cloud demand higher from the hyperscalers, leading to more data being created and processed, thus helping drive a need for these interconnects to meet demand for high-speed, low power data transmission in data centers.
It’s no secret that Big Tech is investing tens of billions in expanding and building physical data center infrastructure over the next five to ten years – just in May, AWS, Google and Microsoft announced tens of billions in investments to expand data center infrastructure globally, which we covered here. These continuous investments in physical infrastructure will not only require GPUs, but also other networking and optical components.
Lumentum’s management has been bullish on the data center optical component market for some time, saying in August 2023 that the “data center optical component market is projected to grow sharply over the next 4 to 5 years to accommodate the increased traffic associated with AI as customers employ ever higher bandwidth interconnects between racks within racks in between servers and storage. We also believe that Datacom VCSEL growth will be meaningful in the next several years as copper is replaced by short-reach multimode optical links.”
As a result, management is eyeing a rather quick ascent to $500 million in quarterly revenue by the end of next year, up 60% from Q4’s projected $302.5 million, while also expecting its cloud business to transform into a multi-billion dollar opportunity in the longer term.
Financials Suffering from Sharp Inventory Correction
Though Lumentum’s management expressed optimism on the longer-term potential of AI-driven data center revenue growth in its recent fiscal Q3 earnings call, the steep inventory correction in telecom has led to a substantial revenue decline and significant margin erosion, presenting a major fundamental headwind for data center growth to overcome.
Management noted in August 2023 that they were “facing significant headwinds as both our direct and end customers actively work to reduce their elevated inventory levels,” and estimated the correction to last through 2023 with shipments “well below end-market demand.” Lumentum reiterated in November 2023 that they were continuing to ship below end market demand.
This created a sharp revenue decline in barely a year and a half — from peak to trough (in six quarters from Q2 FY23 to Q4 FY24’s estimates), TTM revenue has fallen more than (26%).
Telecom Headwinds are Late to Clear
TTM (trailing twelve months) revenues peaked in fiscal Q2 2023 at $1.83 billion, and are now projected to sit at $1.35 billion based on Q4’s guidance. This telecom softness has continued to weigh on results with the inventory correction lasting longer than expected. Lumentum said in its Q3 earnings call in early May 2024 that “revenue will continue to be burdened by telecom customer inventory challenges. The pace of telco carrier spending has slowed more than previously anticipated. Because we continue to ship below end market demand, customer inventory of our products is decreasing, indicating that we are getting closer to the end of this lower demand phase in our industry.”
Once the inventory correction clears, revenue headwinds should subside, though there are lingering doubts about when exactly these issues will clear up, given that we’re now two quarters later than previously anticipated.
Fiscal Q3 Earnings Overview:
Turning to fiscal Q3, Lumentum reported revenue of $366.5 million, down (4.4%) YoY and flat QoQ. Management guided for $290 million to $315 million in fiscal Q4, representing a YoY decline of (18.4%) at the midpoint, far below the $331 million consensus estimate and suggesting its revenue recovery may not be out of the woods until after fiscal Q1 2025 (Sept 2024 quarter).
Margins have rapidly eroded, weighing on Lumentum’s bottom line. TTM gross margin was 20.4%, down from 36.4% a year ago and down from a peak of 46.0% in June 2022. Fiscal Q3’s gross margin of 16.2% suggests the decline in margins is not over yet.
Operating margin has fallen to (22.3%), down from 1.1% a year ago and 17.7% in the June 2022 quarter. Fiscal Q3’s operating margin was (31.3%), and Q4’s guided range points to operating margins declining again.
This is a pretty significant erosion in Lumentum’s operating margin that can’t be fixed solely via cost cuts; in order to drive operating margin to positive, revenue growth will need to rebound alongside improvements in gross margin, most likely back to 30% and higher.
Because of the revenue slump and margin weakness, Lumentum has burned through a substantial bit of cash, reporting $871 million in cash and equivalents in fiscal Q3, down from $1.67B a year ago and $1.22 billion last quarter. The QoQ decrease was primarily due to $323 million in debt repayment in Q3, as operating cash flow was only thinly negative, at ($7 million).
However, total long-term debt is $2.52 billion, raising the likelihood of a capital raise down the line to smooth out any issues with a shrinking cash balance while working to expand production capacity.
Management Quite Optimistic About Data Center Revenue Growth
Lumentum is quite optimistic about the data center opportunities that are arising from Big Tech spending tens of billions on data center infrastructure.
Nvidia’s earnings shed clarity on why Big Tech is investing in the data center at break-neck speeds: CEO Jensen Huang said that “Everybody is anxious to get their infrastructure online. And the reason for that is because they're saving money and making money, and they would like to do that as soon as possible.”
He further explained that Big Tech and other AI firms “need to make money today. They want to save money today. And time is really, really valuable to them. Let me give you an example of time being really valuable, why this idea of standing up a data center instantaneously is so valuable and getting this thing called time to train is so valuable. The reason for that is because the next company who reaches the next major plateau gets to announce a groundbreaking AI. And the second one after that gets to announce something that's 0.3% better. And so the question is, do you want to be repeatedly the company delivering groundbreaking AI or the company delivering 0.3% better? And that's the reason why this race, as in all technology races, the race is so important.”
Lumentum believes that this AI-fueled boom in physical data centers will help drive the next leg higher for its cloud segment. Management said in February 2024 that they expected “revenue from data center transceivers to temporarily dip in the June and September quarters, and then grow significantly through the end of the year and into calendar year '25,” as they work to build out leading-edge transceiver manufacturing capacity in Thailand.
This ties in to a more optimistic view on 2025, with management eyeing exiting the year at a $2B+ annualized revenue run rate, up from $1.35 billion based on Q4’s guidance. Here’s what management said:
“To summarize, the combination of explosive growth in cloud data center and AI-driven demand, our customer traction and capacity additions for new data center products and strong early demand for our new telecom products makes me confident and bullish about calendar 2025. We expect significant growth next calendar year as our investments in new data center products and manufacturing capacity this year translates into significant new revenues. This, combined with the telecom industry inventory correction abating, makes the outlook for calendar 2025 and beyond very promising. We have multiple cloud customer engagements which will drive meaningful revenue growth and drive total company quarterly revenue to exceed $500 million exiting calendar 2025. Additionally, we expect that significant growth will continue into 2026 and 2027. We are working on several significant opportunities today that we expect will propel our cloud business into a multi-billion dollar annual run rate business in the coming years.”We expect significant growth next calendar year as our investments in new data center products and manufacturing capacity this year translates into significant new revenues. This, combined with the telecom industry inventory correction abating, makes the outlook for calendar 2025 and beyond very promising. We have multiple cloud customer engagements which will drive meaningful revenue growth and drive total company quarterly revenue to exceed $500 million exiting calendar 2025. Additionally, we expect that significant growth will continue into 2026 and 2027. We are working on several significant opportunities today that we expect will propel our cloud business into a multi-billion dollar annual run rate business in the coming years.”
CEO Alan Lowe doubled down on this, saying “we would certainly be disappointed if we don't more than double our datacom business by then from today's or from the Q3 run rate.”we would certainly be disappointed if we don't more than double our datacom business by then from today's or from the Q3 run rate.”
Given the billions committed to data center infrastructure from Big Tech over the course of the next 3 to 5 years, there certainly is room for Lumentum to capture such growth, stemming from an increasing need to transmit data in and in between data centers. However, one risk arises here, in that Lumentum is walking a tightrope in working to simultaneously expand production capacity and meet demand without missing the mark.
Management said that “a lot of this incremental capacity is really new customers and diversified customers, both in the cloud space as well as the AI infrastructure space. And so, the challenge is it's a chicken and egg thing in that if you don't have the floor space and capacity, you're not going to get the orders. And … if you have the orders and you don't have the floor space, you're not going to be able to perform. So, we're working hand in hand with our customers to make sure that we're pulling the trigger at the right time to not have too much capacity, but at the same time to build confidence that we're making the investments on behalf of them and the growth that they see in calendar 2025 and beyond.”And so, the challenge is it's a chicken and egg thing in that if you don't have the floor space and capacity, you're not going to get the orders. And … if you have the orders and you don't have the floor space, you're not going to be able to perform. So, we're working hand in hand with our customers to make sure that we're pulling the trigger at the right time to not have too much capacity, but at the same time to build confidence that we're making the investments on behalf of them and the growth that they see in calendar 2025 and beyond.”
This is critical to note, not only because it is new customers and new capacity coming online, but also because it can be quite difficult to perfectly align supply and demand in a fast-moving market.
A Look at Cloud & Networking Revenue
Lumentum’s Cloud and Networking revenue was $313.8 million in Q3, up 9.5% QoQ and 7.1% YoY, with management saying the growth was driven by data center demand and its Cloud Light acquisition two quarters ago.
Cloud and networking revenue has grown sequentially since fiscal Q1, but management’s guidance calls for a sequential decline, including ~$40 million reduction from softer telecom demand, to ~$247.5 million in fiscal Q4. In order to reach management’s goal of exiting calendar 2025 with $500 million plus in quarterly revenue, Cloud and Networking revenue would likely need to grow upwards of 70% to $425 million or above.
Despite Optimism, Analysts Not Convinced of Growth
Despite management’s expressed optimism and explicit statements about reaching $500 million plus in quarterly revenue by the end of 2025, analysts aren’t convinced of this growth.
Current consensus estimates call for revenue of $489 million in the Dec 2025 quarter, just over 2% below management’s target, with the lowest estimate below $475 million, more than 5% below management’s target. Revenue growth also is not expected to return to double-digits until the June 2025 quarter, with estimates pointing to a nearly 40 percentage point acceleration from 6.8% to 45.6% growth.
In addition, revenue revisions are downward, signaling eroding confidence in the turnaround story. FY25’s revenue estimate of $1.51 billion has been revised (9.1%) lower over the past 3 months, while adjusted EPS of $1.67 has been revised (29.7%) lower, suggesting margin headwinds are expected to remain. The upcoming September and December quarters (fiscal Q1 and Q2 2025) have both seen revenue revised more than (13%) lower.
The doubts here about reaching and surpassing these revenue targets likely arises from a combination of factors – persisting softness in telecom weakness for multiple quarters more than anticipated, the tightrope walk for bringing capacity online to book new orders, and in general the six quarter-plus time frame between now and then. These types of revenue accelerations brought about by AI are what the Street has been rewarding recently, especially in hardware and data center players, hence why we’re tracking Lumentum and adding it to our momentum watchlist.
Valuation
Because of the top-line growth headwinds at the moment, Lumentum is trading at a discounted top-line valuation relative to its historical average since its IPO. Shares have traded at an average 2.9x PS multiple, as low as 1x around IPO and more recently 1.5x in October 2023, and as high as 5x in early 2021. Shares are currently trading at approximately 2.2x PS, and 2.3x forward PS due to the revenue declines we’re seeing.
Lumentum’s bottom line valuation is much more stretched, trading at 46.6x adjusted EPS for fiscal 2024, with adjusted EPS projected to decline (77%) this fiscal year due to the margin erosion we previously discussed. However, analysts forecast EPS to rebound quite rapidly in the back half of fiscal 2025 and continue this growth in 2026, with shares trading at an estimated 12.1x adjusted EPS for 2026.
Margins will be the first tell-tale sign of the pace and timing of this EPS rebound and growth, but the current pace of acceleration is attractive should Lumentum be able to hit or exceed analyst estimates.
As you can see, mid-2025 is currently the estimated time for Lumentum to see its rebound. This may seem far off, yet we are building our pipeline now for 2025 potentials and this is one we will continue to watch for when fundamentals and technicals align.
Competitors
Lumentum has a strong competitor in Marvell, with analysts seeing Marvell in “pole position” in the optical space with Inphi. Marvell reported over $1 billion in data center optics revenue in fiscal 2024, and projects electro-optics revenue to surpass $1 billion this year.
Marvell reported $816 million in data center revenue in Q1, up 87% YoY, driven by cloud AI, cloud infrastructure, and electro-optics. Management projected this “robust growth to continue from AI with the expected ramp in our cloud, custom AI programs to augment our substantial base of electro-optics revenue, which we expect will remain correlated to accelerator shipments.”
Electro-optics is expected to be a $1B+ business for Marvell this year, explaining that “from a full year perspective, the way to think about it, maybe some additional color would be, we talked about a floor of $1.5 billion for AI revenue for Marvell for this fiscal year with about 2/3 in electro-optics and a third in custom.” For more context, Marvell’s exit rate in Q4 was “well north of $200 million. And the bulk of that, as we said, was in optics.” Management also provided some clues that electro-optics growth will mirror the pace of AI accelerator growth, from 50% to 100% YoY. This growth is likely outpacing Lumentum’s growth, where management sees 60% growth form $300 million to $500 million over six quarters.
Conclusion
Lumentum has caught our attention as a potential beneficiary of AI capex spending from Big Tech, and while the majority of Big Tech’s budgets will be allocated to GPUs from Nvidia, physical data center infrastructure buildouts will require both AI servers and racks, and networking and optical components.
Lumentum’s management is eyeing data center demand to help drive a push to $500 million plus in quarterly revenues, or nearly new record levels, by the end of calendar 2025, though analysts aren’t fully convinced of the story, with estimates below management’s explicitly stated targets. The telecom inventory correction continues to weigh on revenues while margins have eroded substantially, providing two near-term headwinds to top line and bottom line growth, though forecasts point to strong accelerations in early calendar 2025.
We continue to watch Marvell for similar reasons, which is that telecom headwinds may subside while electro-optics and AI data center growth powers forward in the medium-term. Marvell is a stock we’ve owned in the past with coverage dating back to 2019, covered the Inphi acquisition, and various other reports found here. There is also a potential CXL memory catalyst, described here.back to 2019, covered the Inphi acquisition, and various other reports found here. There is also a potential CXL memory catalyst, described here.
