AMD’s data center segment revenue increased 39% year-over-year to a record $5.4 billion, led by Instinct GPUs and EPYC CPUs. Of this, $390 million was from MI308 sales to China, which means the DC segment reported closer to $5.0 billion in revenue. That number is worrisome to the market because it would mean QoQ growth of 15% for Q4, down from 34% QoQ growth in Q3. This would also translate to YoY growth of 29% – which is not to grab the Street’s attention given Broadcom is in the 100% range on ASICs and Nvidia is reporting about 12X higher on a quarterly basis and 50X higher revenue on an annual basis.
The guide also implies a deceleration from $10.3 billion this quarter to $9.8 billion in the upcoming quarter. Although this beats the Street expectations of $9.37 billion, it’s not the beat/raise tempo being set from other AI stocks right now.
However, there were some bright spots on the call, primarily management stating to expect 60% growth in the data center over the next few years – including 2026. This hint is important as it communicates a strong acceleration into the second half of 2026 given we are starting out at data center growth of 39%, which is what we’ve been expecting.
Per my last earnings writeup: “AMD Q3: The Catalyst is Expected in H2 2026," which stated, “AMD is a stock where I’ve been intentional about managing expectations. The upside is compelling — as the second place in data center GPUs is wide open. Yet for those who have followed our coverage, the timing has always been key: meaningful execution in AI accelerators is not expected to materialize until the second half of 2026. In other words, the long-term opportunity is substantial, but patience remains part of the thesis.”
I could have simply republished last quarter’s write-up except for this one important update that offers 3-5 year visibility; rare for any management team but especially AMD.
Data Center to see 60%+ Growth for Next 3-5 Years
Buried in the call was a rather strong statement for this otherwise-conservative management team that AMD is “well positioned” to grow data center revenue by more than 60% annually over a 3-5 year time frame:
“With the launch of MI400 series and Helios representing a major inflection point for the business, as we deliver leadership performance and TCO at the chip compute tray and rack level. Based on the strength of our EPYC and Instinct road maps, we are well positioned to grow data center segment revenue by more than 60% annually over the next 3 to 5 years, and scale our AI business to tens of billions in annual revenue in 2027.”
Given the strength of the comment, an analyst asked about the comment on the call and if the 60% applies to 2026 with management replying this is certainly possible:
“We're not obviously guiding specifically by segment, but the long-term target of, let's call it, greater than 60% is certainly possible in 2026.”
There have been rumors that AMD may be late in delivering the MI400s, however, that was dismissed today on the call as management reiterated, they are on time for H2:
Later, the CFO also confirmed margins would improve by Q4 based on the launch: “So all those tailwinds we're seeing, we continue to see in the next few quarters. And MI450 ramp, of course, in Q4, our gross margin will be driven largely by mix. And I think we'll give you more color when we get there. But overall, we feel really good about our gross margin progression this year.”
There was more color provided on timing by CEO Lisa Su:
“But as we get into the second half of the year, the MI450 is really an inflection point for us. So that revenue will start in the third quarter, but it will ramp significant volume in the fourth quarter as we get into 2027. So that gives you a little bit of sort of what the data center ramp looks like throughout the year.”
Later she reiterated this again when asked for specific timing: “And our expectation is that we will be on track for our second half launch.”
However, I do expect some analysts to be turned off by the company choosing to not breakout Instinct GPU revenue from EPYC revenue in the data center segment. That would imply GPU revenue is pretty low still. That is not too relevant to our thesis, yet if you see negative notes coming out, it’s due to lack of visibility in the current size of the AI accelerator business.
The only visibility provided was a mention of “tens of billions” in revenue in 2027.
Aaron Rakers Wells Fargo Securities
Lisa, at your Analyst Day back in November, you seem to kind of endorse the high $20 billion AI revenue expectation that was out there on the Street for 2027. I know today you're reaffirming the path to strong double-digit growth. So I guess my question is, can you talk a little bit about what you've seen as far as customer engagements, how those might have expanded? I think you've alluded to in the past multiple multi-gigawatt opportunities. Just any — just double-click on what you've seen from the MI455 and Helios platform from a demand shaping perspective as we look into the back half of the year?
Lisa Su Chair, President & CEO
Yes. Sure, Aaron. Thanks for the question. So first of all, I think the MI450 Series development is going extremely well. So we're very happy with the progress that we have. We're right on track for a second half launch and beginning of production. And as it relates to sort of the shape of the ramp and the customer engagement, I would say the customer engagements continue to proceed very well. We have obviously a very strong relationship with OpenAI, and we're planning that ramp starting in the second half of the year going into 2027. That is on track. We're also working closely with a number of other customers who are very interested in ramping MI450 quickly, just given the strength of the product, and we see that across both inference and training. And that is the opportunity that we see in front of us. So we feel very good about sort of the data center growth overall for us in 2026. And then certainly going into 2027, we've talked about tens of billions of dollars of data center AI revenue, and we feel very good about that.
Financials:
By Royston Roche
Q4 Revenue Grew by 34%
AMD’s Q4 revenue grew by 34.1% YoY and 11.1% QoQ to $10.27 billion, beating estimates by 6.2%. Revenue growth was primarily driven by continued growth in the Data Center segment from both server and data center AI business, as well as a return to YoY growth in the Embedded segment. However, the company’s revenue this quarter included approximately $390 million from MI308 sales to China and excluding this revenue since it was not included in the guidance would yield only a 2.2% beat, the smallest in the last four quarters.
Management guided Q1 revenue of $9.8 billion at the midpoint, implying a YoY growth of 31.8% YoY and down (4.6%) QoQ and the guidance includes about $100 million of MI308 chip sales to China. It is primarily driven by growth in the Data Center and Client and Gaming segments and modest growth in the Embedded segment. Although this beats the Street expectations of $9.37 billion, it’s not the beat/raise tempo being set from other AI stocks right now.
Full year 2025 revenue grew by 34.3% YoY to $34.6 billion. Looking ahead, analysts expect revenue to grow 33.2% YoY to $46.1 billion in 2026 and accelerate to 37.9% YoY to $63.6 billion in 2027.
Q4 Data Center Revenue Grew by 39% YoY
The company’s Data Center segment revenue grew by 39% YoY and 24% QoQ to a record $5.4 billion, led by accelerating Instinct MI350 Series GPU deployments and server share gains. In server, adoption of fifth gen EPYC CPUs accelerated in the quarter, accounting for more than half of the total server revenue. AMD had record server CPU sales to both cloud and enterprise customers in the quarter and exited the year with record share.
In cloud, hyperscaler demand was very strong as North American customers expanded deployments. EPYC-powered public cloud offerings grew significantly in the quarter with AWS, Google and others launching more than 230 new AMD instances. In the enterprise, AMD is witnessing a meaningful shift in EPYC adoption, driven by leadership performance, expanded platform availability, broad software enablement, and increased go-to-market programs. Looking ahead, management expects server CPU demand to be very strong as hyperscalers are expanding their infrastructure to meet growing demand for cloud services and AI while enterprises are modernizing their data centers.
The company delivered record Instinct GPU revenue in the fourth quarter, led by the ramp of MI 350 Series shipments. In addition to the partnership entered in October with OpenAI to deploy 6 gigawatts of Instinct GPUs, AMD is in active discussions with other customers on at-scale multiyear deployments starting with Helios and MI450 later this year. AMD expanded the ROCm ecosystem in the fourth quarter, enabling customers to deploy Instinct faster and with higher performance across a broader range of workloads. The company remains on track to launch its MI500 chips in 2027 and expects to deliver a major increase in AI performance.
Client and Gaming Segment revenue grew by 37%
Q4 client and gaming segment revenue grew by 37% YoY and down (3%) QoQ to $3.94 billion. Client segment revenue grew by 34% YoY and 13% QoQ to $3.1 billion. Management highlighted the strong demand for Ryzen processors for laptops and PCs, which have been gaining market share against Intel.
In gaming, revenue grew by 50% YoY and down (35%) QoQ to $843 million. Management said in the Q4 earnings call, “Semi-custom sales increased year-over-year and declined sequentially as expected. For 2026, we expect semi-custom SoC annual revenue to decline by a significant double-digit percentage as we enter the seventh year of what has been a very strong console cycle.”
Embedded revenue returned to growth in Q4. It grew by 3% YoY and up 11% QoQ to $950 million, up from the YoY decline of (8%) and up 4% QoQ in Q3 2025.
Margins
The company’s profits are growing. However, near term margins are negatively impacted by higher operating expenses to support strong future AI opportunities. Management expects margins to improve by the end of Q4 due to favorable product mix, particularly the ramp of MI450 chips.
Q4 gross profits grew by 44% YoY and 17% QoQ to $5.58 billion. Adjusted gross profits grew by 41% YoY and 17% QoQ to $5.86 billion. Adjusted gross margin was 57%, and it benefitted from the $360 million previously written down MI308 inventory reserves. Excluding the inventory reserve release and MI308 revenue from China, gross margin would have been 55%, up 100 basis points YoY and QoQ, driven by favorable product mix. Management has guided 55% adjusted gross margin in Q1
Q4 operating income grew by 101% YoY and 38% QoQ to $1.75 billion. Operating margin improved by 600 basis points YoY and 300 basis points QoQ to 17%. Adjusted operating income grew by 41% YoY and 28% QoQ to $2.85 billion. Adjusted operating margin improved by 200 basis points YoY and 400 basis points QoQ to 28%. Management has guided an adjusted operating margin of 24% in Q1. The company’s near-term margins are negatively impacted by higher operating expenses to support strong future AI opportunities.
Net income was up 213% YoY to $1.5 billion or 15% of revenue, up 900 basis points YoY. Adjusted net income was up 42% YoY to $2.5 billion or 25% of revenue, up 200 basis points YoY.
Adjusted EPS grew by 40%
The company’s Q4 adjusted EPS grew by 40.4% YoY to $1.53 primarily driven by operating leverage, beating estimates by 16%.
Analysts expect Q1 adjusted EPS to grow 27.6% YoY to $1.22 and 185.8% YoY to $1.37 in Q2 2026.
Cash Flow and Balance Sheet
The company’s cash flows are growing primarily driven by higher revenue and profits.
Q4 operating cash flow grew by 77% YoY to $2.3 billion with an operating cash flow margin of 22%, up 500 basis points YoY.
Q4 free cash flow grew by 91% YoY to $2.1 billion with a free cash flow margin of 20%, up 600 basis points YoY.
The company had cash and short-term investments of $10.5 billion, up from $7.24 billion in Q3. While debt remained the same at $3.22 billion.
Inventories rose 8% QoQ to $7.92 billion to support the strong future AI demand.
Conclusion:
There was no big reveal in AMD’s earnings report, yet interesting enough, that has been the case for years and yet the stock is starting to see movement as AMD outpaced Nvidia’s returns last year by 3X. In my most recent Top 15 AI Stocks report, I highlighted how the element of surprise can work in AMD’s favor. Most importantly, AMD’s management team remains among the most conservative in the space, which makes it notable that we’re already hearing expectations for roughly 60% growth in the data center segment for 3-5 years. That’s a nice clue for what may be on the horizon.
Royston Roche, 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 AMD at the time of writing and may own stocks pictured in the charts.
Meta is a stock that our firm is watching this year with great anticipation. The headline results are good and important to review, yet what’s most important is what is up ahead as Meta begins to merge LLMs directly into their recommendation systems, a shift away from back-end optimizations based on algorithms. This is how management described the shift: “Our feeds will become more interactive overall. Today, our apps feel like algorithms that recommend content. Soon, you'll open our apps, and you'll have an AI that understands you and also happens to be able to show you great content or even generate great personalized content for you.”
There are many advertising businesses using AI, yet there is something unique about how Meta is approaching the problem – and this is showing up in the results. Per our Q1 2026 Top 15 AI Stocks report: “Margins matter, cash matters, but what matters more is the 3X growth Meta has seen in its Advantage+ segment in less than a year, as the company had reported $20 billion about three quarters ago, with the recent update from last quarter at $60 billion. If this runaway growth continues, then Meta will easily be outpacing Search and Google Cloud combined on AI revenue.”
That “something unique” is Meta’s data. What does Search really know about you other than intent? Meanwhile, Meta offers rich contextual, behavioural data that expands over time. The dataset offers preferences, emotions, social context, network effects, etc, whereas Google would struggle to truly “know you.” This can make the LLMs more personal and effective especially as we enter the agent era. Notably, this is not an either/or discussion about Google Search, rather it’s an opportunity to discuss Meta’s strengths and why the company is currently number two in AI revenue despite spending a less capex than its counterparts.
Below, I discuss the earnings call discussions that were centered around the plans Meta has for 2026 and also the numbers that drove strong after hour performance.
Meta to Replace Legacy, Rule-Based Algorithms with AI Agents
Meta is an “eyeballs” company, and thus, an important lever to growth is increasing user engagement. In the most recent quarter, the company drove incremental engagement from ranking and product improvements. Primarily, the company optimized their systems to consider longer interaction histories to better identify a person’s interests. This led to the highest lift in feed views that the company has seen in two years: “The optimizations we made in Q4 drove a 7% lift in views of organic feed and video posts on Facebook, resulting in the largest quarterly revenue impact from Facebook product launches in the past two years.”
Moving forward, Meta’s goal this year is to scale their training data to offer more personalized recommendations. By moving away from algorithms driving the feeds to LLMs, Meta can make he systems more responsive to real-time interest.
This may seem like a subtle shift, but it’s actually not subtle at all – Meta is proposing a complete overhaul in how their systems surface content. Moving forward, LLMs will offer reasoning for a level of personalization not possible in the current approach, which is more pattern recognition based. Think of how Spotify works – it surfaces music you’ve already listened to. Facebook feeds are similar. However, moving forward, Meta can offer a personalized agent approach to where AI optimizes a feed to suggest content that does not require a direct signal.
Here is what was stated on the call:
“We're seeing in our early testing that personalized responses drive higher levels of engagement, and we expect to significantly advance the personalization of Meta AI this year. This dovetails with our investments in content understanding, which will enable our systems to develop a deeper understanding of each person's interests and preferences while also identifying the most relevant content across our platform to pull into responses.”
Although Meta uses AI in its recommendations, the current systems are based on pattern and behavior-driven algorithms. For 2026, Meta will offer content that goes beyond the bounds of what you’ve already searched for/engaged with AI agents that can more intelligently infer your interests.
The result will be more time spent on the platform and with higher engagement. Even incremental gains here will lead to more advertising dollars.
Ad Platform Driven by Superintelligence
The second area that Meta is making “big bets” is the increasing monetization efficiency. Last quarter alone, the company doubled the number of GPUs used to train their GEM model for ads ranking. Similar to what was stated above, part of the improvements is using longer sequences of user behavior to inform the feed plus which ads are placed and when: “This new sequence learning architecture is significantly more efficient than our prior architectures which should enable us to further scale up the data, complexity and compute we use in our future ranking models to deliver performance gains.”
Meta’s main approach to increasing the effectiveness of ad placements remains user targeting, but just smarter user targeting. This results in 4X better results than using AI to increase overall ad load: “In fact, in the second half of 2025, our initiatives on Facebook to redistribute ads across users and sessions delivered a nearly 4x larger revenue impact than Facebook ad load increases.”
As you’ll see below in the Financials section, these improvements are making a material difference with Q4 revenue growing 17% QoQ and with a forward guide that implies the highest YoY growth rate for Meta since Covid-fueled 2021.
Driving Down Costs:
There are two primary ways that Meta plans to drive down costs. The first is to leverage AI internally to reduce their workforce.
Here is what was stated about using AI internally to replace engineers: “Since the beginning of 2025, we've seen a 30% increase in output per engineer with the majority of that growth coming from the adoption of agenetic coding, which saw a big jump in Q4. We're seeing even stronger gains with power users of AI coding tools, whose output has increased 80% year-over-year. We expect this growth to accelerate through the next half.”
The second is to use a mix of custom chips, lower-cost AMD GPUs and Nvidia GPUs to achieve their goals: “Procuring sufficient infrastructure capacity is central to these initiatives, and we're working to meet our silicon needs by deploying a variety of chips that optimally support each of our different workloads. To that end, in Q4, we extended our Andromeda ads retrievable engine, so it can now run on NVIDIA, AMD and MTIA. This, along with model innovations, enabled us to nearly triple Andromeda's compute efficiency. In Q1, we will extend our MTIA program to support our core ranking and recommendation training workloads in addition to the inference workloads it currently runs.”
The paragraph above spells out headwinds to Nvidia to where even if the market continues to grow, Nvidia’s overall percentage of the market will erode. We covered this in the Q3 2025 Top 15 AI Stocks report under the subheading “AI is Diversifying”
Financials
By Royston Roche
Q1 Revenue guide suggests fastest growth since Sept 2021
Q4 revenue grew by 23.8% YoY and 16.9% QoQ to $59.9 billion, beating estimates by 2.4%. Although strong sequential growth in Q4 is seasonal and Meta posted a 19.2% QoQ increase in Q4 2024, the current sequential growth is being achieved on a substantially higher revenue base of $51.2 billion versus $40.6 billion in the prior-year period. The strong revenue growth was primarily driven by robust demand stemming from AI advancements in ad recommendations, monetization, and user engagement.
Management issued strong revenue guide of $53.5 billion to $56.5 billion, implying a 30% YoY growth and a sequential decline of (8.2%) at the midpoint. While the QoQ contraction reflects normal seasonality, the implied 30% YoY growth represents the fastest pace in the last 4.5 years.
The company’s 2025 revenue grew by 22.2% YoY to $200.97 billion. Looking ahead, revenue growth is expected to accelerate 2.8 percentage points to 25% YoY growth to $251.3 billion in 2026 and will moderate to 16.9% YoY to $293.8 billion in 2027.
Record Q4 Advertising Revenue of $58.1 billion
Meta is already seeing tailwinds from AI recommendation models driving higher ROI for advertisers following increased time spent across its family of apps.
Q4 advertising revenue grew by 24.3% YoY to a record $58.1 billion. Notably, absolute advertising revenue growth reached $11.3 billion in the quarter, surpassing the $10.2 billion increase recorded in Q3.
Key Metrics
Q4 ARPP reaches a record $16.56
Perhaps the most important metric for Meta’s ad monetization is Family ARPP (average revenue per person). It reached a record $16.56 in Q4 2025, highlighting that Meta’s AI-driven ad performance improvements and monetization efforts are bearing fruit. While the 16.2% YoY growth in Q4 reflects a deceleration from the 17.7% seen in Q3, such a trend is common on a higher base and less of a concern given Meta is guiding for a modest acceleration this fiscal year. Notably, Q4 ARPP outpaced the 15.6% growth recorded in the prior-year period.
In Q4, the total number of ad impressions grew by 18% YoY, accelerating from 14% growth in Q3 and up from 6% in the year ago quarter. Impression growth was broad-based across regions, driven primarily by higher engagement and user growth, with incremental support from ad load optimizations.
Pictured Above: Ad impressions saw outsized growth this past quarter due to the new sequence learning architecture discussed above. Source: Meta investor relations.
The average price per ad continues to rise and it grew by 6% YoY in Q4, benefiting from increased advertiser demand, largely driven by improved ad performance. However, the YoY growth has decelerated from 10% in Q3 and 9% in Q2 and this metric is to be watched despite ad impressions offsetting the deceleration.
Family of Apps daily active people (DAP) grew by 6.9% YoY to 3.58 billion in Q4, though the growth rate decelerated slightly from 7.6% growth in Q3.
Margins
Meta delivered a sequential improvement in operating margin in Q4 as its continued investments in AI are beginning to show early signs of payoff, even though margins were lower than a year ago quarter.
Q4 gross margin was 81.8%, up 10 basis points YoY and down 20 basis points sequentially.
Q4 operating income increased 5.9% YoY to $24.7 billion, with an operating margin of 41.3%. The margin improved by 130 basis points sequentially, but declined 700 basis points from a year ago, largely due to higher AI-related operating expenses.
Looking ahead to 2026, management expects total expenses to grow by 40.6% YoY to $165.5 billion—driven largely by increased infrastructure spending and continued investment in AI talent. Even with those higher costs, the company still expects operating income to grow in 2026. Management also emphasized that losses at Reality Labs are not expected to increase next year.
Q4 net income grew by 9.3% YoY to $22.77 billion with a net profit margin of 38% compared to 43.1% in the same period last year.
GAAP EPS beat of 8%
Q4 GAAP EPS grew by 10.7% YoY to $8.88, beating estimates by 8%, driven primarily by higher revenue from stronger AI monetization. Analysts expect EPS to grow 2.4% YoY to $6.59 in Q1 2026 and down (0.3%) YoY to $7.12 in Q2 2026.
Looking ahead, GAAP EPS is expected to grow 26.5% YoY to $29.71 in 2026 and 16.3% YoY to $34.55 in 2027.
Cash Flow and Balance Sheet
Meta’s cash flows improved in Q4 driven by higher profits.
Q4 operating cash flow grew by 29.4% YoY to $36.2 billion with an operating cash flow margin of 60.5% compared to 57.8% in the same period last year.
Q4 free cash flow grew by 7% YoY to $14.1 billion with a free cash flow margin of 23.5% compared to 27.2% in the same period last year. Q4 capex grew by 49.2% YoY to $22.14 billion. 2025 capex grew by 84.1% YoY to $72.22 billion. Management expects 2026 capex to be $115 billion to $135 billion, implying a YoY growth of 73.1% at the midpoint primarily due to higher AI investments.
The company had cash & marketable securities of $81.6 billion and debt of $58.7 billion compared to $44.45 billion and $28.8 billion at the end of Q3. The company issued debt of $30 billion in Oct 2025. Meta also entered a joint venture with Blue Owl Capital to fund its development at the Hyperion data center in Louisiana. Thereby, helping it to keep about $27 billion in debt off-balance sheet, where it would sit in a special-purpose vehicle tied to Blue Owl. While this approach may improve reported leverage and financial ratios, it carries inherent risks as the company is indirectly responsible for the off-balance sheet debt.
Conclusion
Meta was able to put up strong results primarily by increasing ad impressions, despite seeing a decelerating average price per ad from 14% growth last year to 6% growth this year. The combination resulted in average price per person reaching $16.56 up from $14.25 in the year ago quarter.
Facebook feeds saw an improvement in its sequencing learning architecture. In other words, Meta tracks user history longer to infer context on what the user is most interested in seeing next, which led to higher user engagement and a spike in ad impressions. Early signs of this improvement are evident with a forward guide that implies the highest YoY growth rate for Meta since 2021 at 30% YoY growth guided for Q1.
Most importantly, management is stating there will be a substantial shift to their underlying recommendation systems by re-architecting its traditional algorithms that are more pattern-based to now become more intuitive and forward-thinking in terms of anticipating what a user will want to see next.
Royston Roche, 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 META at the time of writing and may own stocks pictured in the charts.
Lumentum is benefiting from outsized demand for its EML lasers, reaching a quarterly company record in EML laser shipments. While EMLs are largely spoken for with InP wafer fab capacity fully allocated with long-term agreements, the company is expanding its capacity with additional supply expected to come online in the second half of this calendar year.
The transition to 1.6T is moving faster than management originally anticipated, which contributed to the beat/raise with management stating: “We achieved another quarterly company record in EML laser shipments led by 100 gig line speeds and bolstered by a ramp in 200 gig devices. Simultaneously, we expanded our footprint in next-generation architectures shipping CW lasers for 800 gig manufacturers and increased volumes of ultra-high-power laser shipments for CPO applications.”
There are additional growth levers for Lumentum as we look further out, driven by optical circuit switches and co-packaged optics. Optical circuit switches are beginning to move the needle now with a $400 million backlog, although currently at around $10 million in revenue. In 2027, co-packaged optics (CPOs) will represent another important market for Lumentum alongside UHP chips and ELS modules will help expand the company’s serviceable addressable market.
To learn more about upcoming networking shifts, read our Top 15 report here.our Top 15 report here.
Increasing Capacity for H2 2026
Management emphasized that indium phosphide wafer fab capacity is fully allocated, with the company indicating they have already delivered half of their expansion target over one quarter alone due to strong customer demand necessitating they pull forward delivery. Thus, the natural question for an investor is whether Lumentum can add more capacity. The company stated they foresee more capacity coming in the second half of the calendar year:
“We are scaling rapidly through precision tool optimization and yield gains. This execution will help to ensure that additional capacity comes online as planned over the next two quarters and beyond. While not able to size it, we now have line of sight to a significant block of additional capacity starting in the second half of 2026 both recurrent activities in Sagamihara and better utilization of our Caswell, United Kingdom and Takao, Japan fabs.”
200G/1.6T Ramping Faster than Expected
Although minimal right now, 200G is ramping faster than expected, representing 5% of unit volume yet represents 10% of laser chip revenue. According to management, demand for 200G EMLs is about a quarter faster than they originally anticipated with the goal of ending the year with 25% of unit volume from this new product mix – with these seeing higher average sales prices than the 100-gig.
Management stated the following: “Our 200-gig line speed, as we said, is actually doing a little bit better than we expected. I think on the last call, we had said that the 5% revenue of — 5% of mix would be this quarter. It was a quarter earlier than we had expected, and that's primarily because 1.6T is coming on, I think, faster than we initially anticipated, and that is heavily being driven by 200-gig EMLs.”
This was discussed further in the call with management stating 1.6T was stronger than it was 90 days ago:
Ruben Roy Stifel, Nicolaus & Company, Incorporated
Great. Just a very quick follow-up. Has anything changed with the way you guys are thinking about 800 gig versus [ 1.6T ] module mix this year one way or the other? Is it accelerated towards 1.60T for any reason in terms of volumes from a single customer or multiple customers? Or is it relatively unchanged from how you're thinking about it 90 days go?
Michael E. Hurlston President, CEO & Director
Yes. 1.6T is definitely stronger than we felt than we felt 90 days ago. So 1.6T is definitely accelerating. Our 800-gig volume actually is doing better than we would have expected. So an 800-gig that what you're seeing right now from us is an acceleration in revenue on our 800-gig shipments. But in the market, to your question, Ruben, 1.6T is definitely going better. We have exposure to a couple of customers, a couple of large customers on 1.6T, and we've been surprised by how quickly they're trying to push us to deliver and their forecast to us relative to the different SKUs that we're being asked to deploy.
OCS Starting to Move the Needle, CPOs Incoming for 2027
Optical circuit switches are starting to move the needle. By the end of fiscal 2027, management now expects roughly $400 million in revenue, up from the $100 million originally guided—representing an incremental $300 million upside to prior expectations.
Management stated the following on the call:
“If you remember, you and I discussed last time, we believed our calendar Q4 would be around $100 million. It looks like it will be quite a bit higher than that, although we're not breaking up that $400 million between the two quarters. So we — it's a broad-based — there's multiple customers making up that backlog. We've talked about shipping to three customers, and that continues — but those customers are increasing their demands rather significantly. And thus, the demand on us has gone up quite appreciably. So we feel pretty good about that. I think as we enter calendar year 2027, it should go up from there in terms of what we see in our backlog and in terms of our revenue.”
Co-packed optics are expected to contribute early 2027 with stronger contribution by year-end 2027: “An industry pivot is underway to bypass the scaling limits of copper. By late calendar 2027, we would expect our first scale-out CPO shipments, replacing longer copper connections.”
Financials
Revenue Growth to Accelerate to 89% in Q3
Lumentum delivered Q2 revenue at the upper end of its guided range, yet its guidance stands out as it not only points to YoY growth accelerating almost 24 points to 89.3% YoY, but it also was a larger magnitude beat in dollar terms versus last quarter.
Q2 revenue was $665.5 million, a modest 2% beat to estimates and in the upper end of Lumentum’s guidance for $630-$670 million. Revenue growth accelerated 7.1 points to 65.5% YoY, while sequential growth was robust at 24.7% QoQ, its fastest growth in eight years and accelerating 13.7 points.
For Q3, Lumentum guided for revenue between $780 million and $830 million, accelerating 23.8 points to 89.3% YoY at midpoint. Sequential growth will remain strong with guidance pointing to growth of 21% QoQ at midpoint. What’s impressive here is that Lumentum’s guidance beat consensus by a larger margin than it did last quarter – at the $805 million midpoint, this would be nearly $99 million ahead of the $706.4 million estimate, whereas Q2’s guide for $650 million at midpoint beat by ~$88 million.
Considering the scope of this raise for Q3, it’s likely that estimates for Q4, which currently are pegged at just $770.4 million, are revised much higher in the coming days/weeks. As a result, it’s likely that consensus estimates for FY26, currently at $2.64 billion, move ~8-10% higher.
For a brief recap, Components include laser chips, laser subassemblies, line subsystems and wavelength management subsystems, while Systems includes full stand-alone products such as optical transceivers, optical circuit switches and industrial lasers.
Lumentum expects Components to be the cornerstone for revenue growth and profitability while Systems will scale rapidly with transceivers, OCS and other high-performance solutions.
Components Revenue Accelerates Slightly
Components revenue was $443.7 million in Q2, up 68.3% YoY and 17% QoQ. YoY growth accelerated 3.4 points, though QoQ growth decelerated 1.4 points from last quarter. Components accounted for 66.7% of revenue in Q2, down from 71% in Q2.
Driving this growth were EML shipments, with Lumentum saying both 100G and 200G EMLs reached new company records. Ultra-high-power lasers for CPO continued to grow, with Lumentum outlining a broader ramp in the second half of calendar 2026, aligning with Nvidia’s Spectrum-X switch roadmap with its Vera Rubin platform.
Lumentum also noted that Q2 saw an eighth consecutive quarter of sequential growth for narrow linewidth lasers for data center interconnect (DCI) applications, while pump lasers for scale-across and sub-sea applications reached a record.
Systems Revenue up 43% QoQ, New Record for Transceivers
Systems revenue saw much stronger sequential growth than Components as it is coming off a much smaller base, with cloud transceivers likely to be the main driver as optical circuit switching (OCS) is still very early in its ramp.
Systems revenue rose 43.5% QoQ and 60.1% YoY to $221.8 million, a sharp acceleration from a (3.6%) QoQ decline and 46.5% YoY increase in Q1. This was driven by record cloud transceiver shipments.
Lumentum did note that OCS shipments exceeded a $10 million quarterly run rate in the quarter, while manufacturing readiness is proceeding ahead of schedule as the company prepares to begin fulfilling its >$400 million OCS backlog later in 2026.
Margins Expand Substantially, Operating Margin up 22.5 Points YoY
Complementing the strong revenue growth is substantial margin expansion, with Lumentum showing GAAP operating margin expand 22.5 points YoY to nearly crack into double digit territory. Lumentum’s cost profile also shows that operating margin expansion will continue as costs rise at a much slower pace than revenue growth.
Lumentum reported solid expansion in gross margins in Q2, with GAAP gross margin up 2.1 points QoQ and 11.3 points YoY to 36.1%, and adjusted gross margin up 3.1 points QoQ and 10.2 points YoY to 42.5%.
However, operating leverage was quite prevalent and visible in the quarter as opex rose just 0.6% QoQ and 16.3% YoY. This, combined with gross margin expansion, drove significant expansion in operating margins on a YoY and QoQ basis.
GAAP operating margin in Q2 was 9.7%, up 8.4 points QoQ and 22.5 points YoY, while adjusted operating margin was 25.2%, up 6.5 points QoQ and 17.3 points YoY (and ahead of guidance for 20-22%). Lumentum forecast this operating margin to continue at a similar rate, projecting adjusted operating margin of 30-31% in Q3, up 5.3 points QoQ and 19.7 points YoY.
To note, Lumentum is well ahead of its target financial model, which called for >20% adjusted operating margin and 39-42% adjusted gross margin at a $750 million/quarter. Lumentum is already above both metrics on $665 million/quarter base.
Net margins followed, with GAAP net margin expanding 11 points QoQ and 26.9 points YoY to 11.8%. Adjusted net margin expanded 5.4 points QoQ and 14.1 points YoY to 21.6%. An important takeaway here is that this AI-driven growth is driving strong earnings leverage for Lumentum, as adjusted net margin was single-digits just three quarters ago.
Adjusted EPS Beats by 18%, Q3 Guided 40% Above Estimates
As just mentioned, this AI growth and margin expansion is driving visible earnings leverage for Lumentum. Not only did adjusted EPS beat estimates by 18% in Q2, but Lumentum’s Q3 guide was more than 40% above estimates, signaling >290% growth will be maintained for another quarter.
Q2 GAAP EPS was $0.89, up from just $0.05 in Q1 and beating the $0.50 consensus estimate by 78%. Adjusted EPS was $1.67, up 51.8% QoQ and 297.6% YoY, and beating the $1.40 estimate by 18.4%.
For Q3, Lumentum guided for $2.15 to $2.35 in adjusted EPS, pointing to YoY growth of 294.7% and QoQ growth of 34.7%. At midpoint, this represented a 40.6% beat to the consensus estimate of $1.60.
Combined, Q3 and Q4 adjusted EPS came in $0.91 ahead of estimates, and if Q4 is to see a similar ~$0.60 to $0.70 upward revision to match Q3’s beat, full-year adjusted EPS estimates could move to around the $7.50 range, up from the current $5.92 prior to the report.
Cash Flows and Balance Sheet
Cash and equivalents increased slightly to $1.16 billion while debt was $3.29 billion.
Inventories were $570.4 million, up more than 7% QoQ, while accounts receivable surged nearly 23% QoQ (again) to $376.8 million, supporting the upcoming product ramps over the coming quarters.
Conclusion:
AI networking is entering a new phase, one where silicon photonics plays a much larger role alongside system-level shifts such as optical circuit switches and co-packaged optics.
The goal is to continually optimize for AI workloads rather than rely on traditional networking as ever-expanding AI clusters seek speed, lower latency and reduced power consumption. Pluggable optics work well enough today but are quickly becoming a limiting factor as power and port density doesn’t scale well with traditional optics. This dynamic is pushing hyperscalers to rethink network architectures, and positions Lumentum well, as its solutions help future-proof AI networks for the years ahead.
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 LITE at the time of writing and may own stocks pictured in the charts.
Palantir reported another very strong quarter in Q4, with revenue accelerating to 70%, an impressive 57 point acceleration over the last ten quarters, while guiding for revenue to accelerate further to 73.6% in Q1.
US commercial momentum remained unabated, with revenue accelerating 16 points sequentially to 137% YoY, surpassing the $500 million mark in the quarter. When looking at the strength of both QoQ and YoY growth, it’s likely Palantir represents the highest AI segment growth across the AI universe.
On top of that, Palantir initially guided for fiscal 2026 revenue to accelerate from 56.1% to nearly 61% YoY, driven by US commercial revenue accelerating six points to >115% YoY. Driving such an acceleration at these growth rates is undeniably difficult, yet there are hints that Palantir could go above and beyond these figures by this time next year.
Unlike 99% of other companies reporting this season, Palantir’s call offered little substance to sift through, with more clues on its growth opportunities hidden in the key metrics – detailed for you below.
Q4 Revenue Growth Accelerates to 70%, FY26 Guided to Accelerate to 61%
Palantir reported revenue of $1.41 billion in Q4, accelerating to 70% YoY while QoQ growth ticked 1.5 points higher to 19.1%, and marking a 50 point acceleration over the last two years. This also is Palantir’s highest revenue growth in their history as a public company.
More impressively, Palantir guided for this revenue acceleration to continue into Q1 and for 2026, suggesting that the AI-driven growth engine that propelled shares higher through 2024 and 2025 is still intact, and potentially strengthening. Q1 revenue was guided to be $1.532 billion to $1.536 billion, accelerating 3.6 points to 73.6% YoY at midpoint (and what would be a fresh record growth rate), though QoQ growth would be just 9%.
For 2025, Palantir reported $4.48 billion in revenue, up 56.1% YoY, and accelerating 27.3 points from 28.8% growth in 2024. Total commercial revenue rose 60% YoY to $2.07 billion, while government remained a key contributor with revenue up 53% YoY to $2.40 billion. 2025’s initial guidance also helps provide some perspective on just how strong Palantir’s year was, and how 2026 could materialize if this robust AI-driven momentum continues (which key metrics support) – Palantir initially guided for 30.9% growth, but ended the year more than 25 points higher at 56.1%.
For 2026, Palantir offered an initial guide for $7.182 billion to $7.198 billion, up 60.7% YoY at midpoint, or $900 million ahead of consensus for $6.29 billion for 42.8% growth. This would also mark a 4.6 point acceleration, a significant feat considering the swift acceleration the company saw through the back half of 2025.
US Commercial Growth Guided to Accelerate in FY26 to 115%
Palantir’s AIP-driven US commercial segment remains the company’s core revenue driver, with growth accelerating once again in Q4 to the fastest rate in four years. What’s more impressive is that Palantir not only has guided for US commercial revenue to more than double in 2026, but that it was guided to accelerate from 2025’s already-rapid 109% growth.
US commercial revenue rose 137% YoY and 29% QoQ to $507 million in Q4, surpassing a $2 billion annualized run rate in the quarter, up from a $1 billion run rate at the start of 2025. QoQ growth accelerated only one point from 28% in Q3, though accelerating sequentially at this pace is difficult.
On a YoY view, US commercial continued to accelerate, with the 137% growth in Q4 marking a 16 point acceleration from 121% YoY in Q3. Since the start of the year, US commercial revenue growth has accelerated a tremendous 66 points.
While US growth continues to accelerate, Palantir lags greatly in International. International commercial revenue is not nearly as strong as US commercial with revenue growth of just 8% YoY and 12% QoQ to $170 million in Q4. For context, International commercial revenue has not grown faster than 10% YoY in the last six quarters. However, despite the softness in international, overall commercial revenue growth managed to accelerate from 73% to 82% YoY in Q4.
For 2025, US commercial revenue rose 109% YoY to $1.465 billion, driven by the strong acceleration in the back half with Palantir reporting back-to-back quarters with >120% growth.
For 2026, Palantir delivered an initial guide for $3.144 billion in US commercial revenue, representing growth of >115% YoY, a six point acceleration. It’s hard to understate the strength of this initial guide – not only is Palantir doubling revenue once again after doubling in 2025, but that there is a high likelihood that high AIP demand will translate into stronger growth as the year pans out.
What Palantir’s Beat/Raise Pattern from 2025 Suggests about 2026
Palantir’s recent beat/raise trends from 2025 shed light into how 2026 could shape up for US commercial revenue, if the company can maintain a similar trajectory throughout this year on strong AI momentum.
Looking back to 2025, Palantir’s first guide for US commercial revenue was for 54% YoY growth to $1.08 billion. This was raised to 68% YoY to $1.18 billion by Q1, and raised again to 85% YoY to $1.30 billion by Q2. By Q3, management raised US commercial growth to 104% YoY to $1.43 billion, with the final number being the 109% reported this quarter, for a total raise of 55 points throughout the year.
The challenge for Palantir is that its initial guide for 2026 starts at a much higher base at >115%, though its pattern through 2025 and strong key metrics (outlined below) suggest that upside to this forecast is likely.
Plotting out a modest outperformance through the year and a similar outperformance as 2025 show two different trajectories for US commercial revenue growth. Under the modest outperformance scenario, or along the lines of a five point raise to growth each quarter and a small Q4 beat, this would project revenue out to ~133% YoY, or 18 point upside versus this initial guide and a 24 point YoY acceleration. To put this in dollar terms, this scenario would project revenue of $3.42 billion, or ~$280 million above guidance (smaller than 2025’s $385 million beat).
If Palantir can outperform to a similar degree as 2025, such as 45-50 points above the first guidance, the revenue projection for US commercial would look much different. This scenario would need around a 10 to 12 point raise each quarter, and could project revenue as ~160% YoY, a 51 point acceleration. In dollar terms, this would project $3.82 billion, or ~$680 million above guidance.
The main takeaway here is that even a modest outperformance and guidance raises of a few points each quarter could easily drive US commercial revenue growth to a double-digit acceleration from 2025’s 109% growth.
Government Revenue Remains Strong
Outside of US commercial, Palantir’s government remained strong with 60% YoY and 15% QoQ growth in the fourth quarter to $730 million, accelerating from 55% YoY and 14% QoQ in Q3.
It cannot be ignored that government still remains critical to Palantir’s success despite the unwavering US commercial momentum, as government accounted for nearly 52% of revenue in Q4.
US government revenue rose 66% YoY and 17% QoQ to $570 million, accelerating from 52% YoY and 14% QoQ in Q3, whereas international revenue grew 43% YoY and 9% QoQ to $160 million, driven by work in the UK.
Palantir said this US growth was driven by its mission impact across the DoD and accelerating momentum in civil agencies, highlighting that it was awarded an up to $448 million contract with the US Navy to “modernize the shipbuilding supply chain and accelerate delivery of naval vessels.” Palantir added that Maven usage reached all-time highs and will “continue to be rolled out to all combatant commands and many more networks over the rest of this government fiscal year.”
However, CEO Alex Karp downplayed Palantir’s international growth outlets and ability to meaningfully accelerate growth especially in Europe, saying that “Palantir is in a unique position where we really don't have the bandwidth to do anything that's difficult outside of America. So — and as this learning curve goes on, it's more and more difficult to help people understand how to implement these things and the demand in the U.S. is so great.”
Net Retention Rate Expands 5 Points, Rule of 40 Expands to 127%
While Palantir’s AI-driven revenue growth metrics are the strongest among AI-exposed software, its key metrics remained robust in Q3 and support this strong AI-driven growth curve persisting through 2026.
Net Retention Rate Expands 5 Points
Palantir’s NRR accelerated another 5 points sequentially to 139% in Q4; on a YoY basis, NRR has risen 19 points. Should Palantir continue to drive similar NRR expansion through 2026 as it expands the 680 >$1 million deals it signed in 2025, there is a chance they move above the 150% threshold at some point in 2026.
RPO Surges – Up 62% QoQ and 143% YoY
RPO saw a meaningful step up in Q4, rising 62% QoQ to $4.21 billion, with YoY growth accelerating from 65.6% in Q3 to 143.4% in Q4. This also represented the company’s strongest RPO growth since the start of 2023 on both a YoY and QoQ basis.
This strength was also reflected in billings, which rose 91.1% YoY and 21.5% QoQ to $1.49 billion, a sharp acceleration from 49% YoY and 11.5% QoQ in Q3. Both of these key metrics witnessing this sharp step-up in tandem provides further confidence in Palantir’s 2026 accelerations panning out with the potential for upside to its initial guidance as each quarter progresses.
Rule of 40 Expands 13 Points QoQ and up 46 Points YoY
Palantir also stands out for its exceptional Rule of 40 score which continued to expand in Q4, which the company defines as revenue growth plus adjusted operating margin. It’s not just the fact that Palantir’s Rule of 40 score is now well beyond 100%, but the fact the company is expanding this margin by a significant degree:
“Our Rule of 40 score reached new heights at 127%, up 46 points year-over-year and 13 points quarter-over-quarter, proving that hyper growth and exceptional profitability aren't mutually exclusive, but rather the inevitable outcome of Palantir delivering transformational impact at scale.”
For 2025, Palantir had a Rule of 40 score of 106%, with management guided for an initial Rule of 40 score of 118% for 2026.
Second Highest TCV Growth on Record
Palantir also booked record TCV of $4.26 billion, up 138% YoY, with commercial TCV of $2.6 billion, up 161% YoY and 83% QoQ, accelerating from 132% YoY and 32% QoQ in Q3.
RDV was $11.2 billion, increasing 105% YoY and 29% QoQ, accelerating from 91% YoY and 21% QoQ; however, US commercial RDV slowed from 199% YoY and 30% QoQ in Q3 to 145% YoY and 21% QoQ in Q4.
Again, the simultaneous accelerations in Palantir’s key metrics in Q4 signals that the company’s growth engine through 2026 is visibly strengthening, providing more confidence in upside to Palantir’s guidance for both revenue and US commercial revenue.
Margins Continue to Strengthen
While its revenue growth and acceleration are second-to-none in AI software, so are Palantir’s margins, with the company showcasing an impressive ability to drive margin expansion of >10 points while simultaneously accelerating revenue. For example, Palantir’s adjusted operating margin in Q4 was a record 57.4%, well ahead of its guidance for 52.4% and expanding 12 points YoY. This is a remarkable feat as it highlights Palantir’s ability to maintain its cost profile despite meaningfully accelerating revenue quarter after quarter. Adjusted EBITDA margin also showed strong expansion, coming in at 57%, up 6 points QoQ and 11 points YoY.
Looking down the line, gross margins expanded nicely in Q4, with GAAP gross margin at 85%, up 6 points YoY and 3 points QoQ. Adjusted gross margin also expanded but at a smaller degree, up 3 points YoY and 2 points QoQ to 85%.
The operating margin expansion was where Palantir shone. GAAP operating margin was 41% in Q4, up 40 points YoY (coming against a low comp due to the one-time stock appreciation rights (SARs) expense) and 8 points QoQ. As noted above, adjusted operating margin was 57.4%, up 12 points YoY and 6 points QoQ. Palantir guided for adjusted operating margin to remain strong in Q1 to 56.8% at midpoint, up 13 points YoY and down marginally QoQ.
Turning to net margin, the expansion was less pronounced but still visible. GAAP net margin was 43%, up 33 points YoY (again vs the low SARs comp) and 3 points QoQ. Adjusted net margin was 46%, up 5 points YoY and 1 point QoQ.
It’s easier to see how Palantir’s margins have strengthened when looking at the full-year. GAAP operating margin for 2025 was 32%, up 21 points. Adjusted operating margin expanded 11 points to 50%, and Palantir has initially guided for adjusted operating margin to expand 7.5 points in 2026 to 57.5%.
Down the line, GAAP net margin was 36% for the year, up 20 points, and adjusted net margin was 43%, up 8 points.
EPS
Palantir reported $0.25 in adjusted EPS in Q4, up 78.6% YoY, with GAAP EPS coming in at $0.24, up 700% YoY and beating estimates by 8.7% and 33.3% respectively. Palantir did not provide guidance for Q1, though consensus estimates currently call for adjusted EPS of $0.21, up 60.2% YoY, and GAAP EPS of $0.16, up 100% YoY.
For the full year, Palantir delivered adjusted EPS of $0.75, up 82.9% YoY, and GAAP EPS of $0.63, up 231.6% YoY. Again, Palantir did not provide guidance, though current consensus points to adjusted EPS up 39.7% to $1.01 and GAAP EPS up 30% to $0.82.
Cash Flows Robust
Palantir’s cash flows were robust in Q4, and management guided for adjusted FCF margin to expand in 2026 from an already strong 51% in 2025.
Operating cash flow was $777.3 million in Q4 for a 55% margin, down slightly from a 56% margin in the year ago quarter but rebounding solidly from a 43% margin in Q3. For the year, Palantir delivered operating cash flow of $2.13 billion, or a 48% margin, up from 40% in 2024.
Adjusted free cash flow was $791.4 million in Q4 for a 56% margin, down from a 63% margin a year ago but up from 46% in Q3. For 2025, Palantir generated $2.27 billion in adjusted FCF for a 51% margin, up from 44% in 2024.
For 2026, Palantir guided for a step up in adjusted FCF, projecting it to increase more than 77% YoY to $3.925-$4.125 billion. This would represent an adjusted FCF margin of 56%, a five point expansion from 2025.
Palantir’s balance sheet remained extremely healthy with cash of $7.18 billion and zero debt.
Valuation
Palantir’s valuation remains in uncharted territory, though forward multiples have come down quite a bit when factoring in FY26’s guidance and the new fiscal year adjustment.
On a forward PS basis, Palantir is trading at a 50x multiple with the initial 2026 guide for $7.19 billion in revenue. This is the cheapest shares have been valued since late April 2025, and a meaningful ‘discount’ to the >100x multiple Palantir commanded at the end of 2025.
On the bottom line, Palantir’s multiple has also come down, with shares now trading at 145x forward PE, nearly half of the 260-280x multiple it held from the end of 2025 and only a ~30% premium to its 112x average five-year multiple.
Conclusion
Palantir’s Q4 report showed that the company’s AIP-driven momentum remains robust with no signs of slowing, further supported by most key metrics accelerating in unison. Palantir’s NRR expanded 5 points to 139%, its Rule of 40 score expanded 46 points YoY to 127%, and record TCV and RPO were the cherry on top of a strong quarter.
Palantir also guided for revenue to accelerate to nearly 61% YoY in 2026, driven by US commercial revenue accelerating to >115% YoY. Driving an accelerate at this multi-billion dollar scale is difficult, yet the company’s key metrics suggest growth rates may continue to move higher.
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 PLTR at the time of writing and may own stocks pictured in the charts.
SanDisk’s second quarter report was a blowout on all accords, with the company reporting an impressive 31% QoQ growth for revenue to $3.03 billion and a tremendous 408% QoQ growth to $6.20 in adjusted EPS, capitalizing on strong demand and strong pricing from undersupply dynamics.
However, the guide was even more impressive, with SanDisk forecasting $4.4 to $4.8 billion in revenue, up 52% QoQ at midpoint, and adjusted EPS more than doubling QoQ to $12 to $14, roughly 200% above consensus at midpoint.
To put in perspective just how large of a beat this was, SanDisk was not expected to see this level of revenue or EPS at the end of 2027 – consensus for the Dec 2027 quarter was $4.19 billion and $9.29 in EPS heading in to this report.
There were a handful of important comments from management in the call regarding the NAND market, that it will be even more undersupplied in fiscal Q3, while data center growth forecasts raised yet again.
NAND Market Remains Critically Undersupplied
SanDisk attributed its revenue and margin over-delivery versus its guidance to strong pricing during the quarter, a byproduct of the tightening supply-demand environment. Management offered some commentary on supply-demand dynamics for next quarter and for 2026, which points to strong pricing growth remaining a tailwind as supply is expected to lag demand by a widening margin.
Management stated: “In the December quarter, we experienced a clear and significant improvement in market conditions across end markets, which led to higher pricing."
SanDisk noted that it was unable to fulfill customer demand in Q2, yet management added that it anticipates “the market to be more undersupplied [in Q3] than it was in the second quarter” with bit growth down mid-single digits QoQ compared to a mid-single digit increase QoQ in Q2.
Management also added that they expect “customer demand well above supply beyond calendar year 2026, which requires careful allocation planning and alignment with our customers.”
This is rather important as analysts were currently expecting NAND pricing to peak in the calendar Q1 quarter on a QoQ basis, yet ASP growth may end up higher for longer considering the supply-demand imbalance is widening.
For example, analysts were projecting NAND ASPs to accelerate to the low-20s to low-30s QoQ in calendar Q1 and then slow to the mid-teens in calendar Q2, yet a widening imbalance could potentially push prices up to 40% QoQ and 20% QoQ, respectively.
For SanDisk, this could have strong implications for both revenue and earnings, as the company is showing revenue growth at a multiple of this QoQ growth in prices. For example, SanDisk’s revenue was up 21% QoQ last quarter, or 3-4X estimated ASP growth, while this quarter revenue rose 31% QoQ, or 1.5-2X estimated ASP growth.
For fiscal Q3 (CQ1), the company’s 52% QoQ guide is ~1.5-2.5X estimated ASP growth, meaning that if this trend persists and ASP growth estimates move materially higher, to maybe ~20% for CQ2, this could roughly project 30% QoQ growth for SanDisk in its fiscal Q4, or potentially well above $6 billion.
Moving back to the topic of supply, there was one major factor that management highlighted that could somewhat hinder its ability to continue growing at a multiple of QoQ ASP growth – long-term agreements. SanDisk explained that it is looking to evolve from quarterly supply negotiations (as has been typical with NAND historically) to multi-year agreements with firmer supply and pricing commitments. This in part stems from data center customers offering visibility into 2026-2028 demand needs and some even out to 2030, requiring a “substantial” amount of exabytes, with SanDisk seeing LTAs as an avenue to offer “confidence in supplying that level of demand on a sustained basis.”
Management is aiming to have supply plans aligned with “predictable long-term demand at current and forecasted market prices,” hinting that they will look to align these LTAs to capture potential price increases through the rest of the year to minimize lost revenue that could be captured under quarterly deals. It is a possible two-edged sword though, in that it does lock in supply prices over quarters to years, insulating SanDisk from when prices (ultimately) revert lower, but also potentially capping price-driven growth as prices would then be locked in and not shifting quarter-to-quarter.
SanDisk disclosed that it signed and closed one LTA in the quarter, opting not to disclose terms other than stating that it included a pre-payment component, with several more LTAs in the queue.
Data Center Growth Forecasts Accelerating Rapidly
SanDisk once again upped its data center exabyte growth forecast, this time to a larger degree, with management going so far as to hint that this updated forecast may still be too low.
Now, SanDisk explained that they are now “looking at high 60% exabyte growth in that market for '26,” a more than 20 point raise, and yet this increased outlook “doesn't include any CapEx raises on this earnings cycle.”
To note, Meta already provided a rather large capex increase, guiding initially for $115-135 billion in capex for the year, up 73% YoY and well ahead of estimates for ~$108 billion. Microsoft did not provide a guide, yet quarterly capex was $37.5 billion, ahead of estimates for $34.3 billion, and Amazon and Alphabet have yet to report, though if its peers are any indication, it’s that capex will exceed estimates once more.
This accelerating forecast for data center exabyte growth ties into NAND’s increasing role in AI infrastructure, as we had recently outlined with KV cache requirements and Nvidia’s new inference memory platform. On this exact topic of Nvidia’s KV cache discussion and TB of content per GPU, management explained that “none of that demand is in the numbers we're talking about, demand numbers at this point,” but “our initial looks at it when we look at, let's say, '27 demand, we think that's roughly maybe 75 to 100 additional exabytes. And then a year after that, you can double that. So it is a significant amount of demand.”
For context, 75-100 EB of demand in 2026 would account for roughly 6-8% of the entire flash market, while doubling that to 150-200EB in 2027 would correspond to 10-13% of the market – a significant new demand driver.
As a result of the increasing role of NAND and enterprise SSDs in AI inference applications, and expectations for a “meaningful increase in NAND content per deployment,” management expects data center revenue to “grow meaningfully in both the near and long term.”
Data center revenue witnessed a sharp acceleration in the quarter with revenue up 64% QoQ, up from 26% QoQ in fiscal Q1, with management expecting this to accelerate in the back half of the year with a “substantial step-up next quarter.”
Discussions on Margins, Opex
Because NAND has traditionally been very cyclical, analysts questioned about where true cycle gross margins will land considering the expectations for NAND to become more secular in nature. Unpacking the true cycle margins peak to trough is important considering pricing power with strong data center growth and other factors such as the BiCS8 ramp and lower per-bit costs are tailwinds to margin expansion, yet when supply-demand tightness begins to ease and prices reverse, margins will likely face stiffer headwinds.
Q, Asiya Merchant, Citigroup:
“How are you thinking about your true cycle margins, gross margins, seems like that was quite a long time ago when you were hitting those levels. But how are you thinking about gross margins here structurally?”
A, Luis Visoso, SanDisk CFO:
“I think the way I would answer your question about through-cycle margins is similar to where David left it, which is in a high CapEx, high R&D industry or company. Frankly, 35% [gross margin] is not where we would like to be, right? So we're not going to give you a new number today. But clearly, that's not where we want to be. What I'll tell you is this is the first quarter, right, that we are above 35% with 51%. We're guiding, call it, midpoint of 66%. So we're making progress and we're getting to a place where we believe we can justify the CapEx. We can justify the investments in R&D that the business requires.”
Management also explained that the current opex run rate – $476 million for GAAP opex in Q2 and to ~$514 million guided for Q3 at midpoint, and $413 million for adj opex and $460 million guided for Q3 – will not move significantly higher as revenue scales: “a long way of saying the level of spending we had last quarter, what we're guiding this quarter, those are kind of more sustainable levels for now.”
This allows for a rough view of where true cycle earnings power could lie. Should the true cycle gross margins sit at around 35%, despite management aiming for higher, this offers a rough view into where true cycle earnings could sit. At a ~$16 billion revenue base (~4% above current FY26 consensus at $15.46 billion on Jan 30), a 35% gross margin with ~$1.7 billion in adjusted opex and roughly $600 million in additional expenses would project true cycle adjusted earnings around ~$21.
Management also discussed the following: “We're literally able to trade out the lowest margin business for now the highest margin business, and that provides a significant tailwind to the business as well.”
Financials
Revenue Showing Sharp Acceleration into Q3
SanDisk reported $3.03 billion in revenue in Q2, beating estimates by ~12.5%, with SanDisk attributing the growth to higher prices across its three segments with prices strengthening through the quarter. Revenue growth accelerated more than 38 points to 61.2% YoY, while sequential growth accelerated nearly 10 points from 21.4% QoQ in Q1 to 31.1% QoQ in Q2.
Q3’s guide was a blowout versus consensus, with SanDisk forecasting $4.4 to $4.8 billion in revenue, more than 58% ahead of consensus for just $2.91 billion. This also points to a significant 110 point acceleration to 171.3% YoY at midpoint and 21 points to 52% QoQ. Initial estimates for fiscal Q4 point to 191% YoY growth and 20% QoQ growth from Q3’s midpoint to $5.53 billion.
For the full year, current consensus points to 110% YoY growth to $15.46 billion, though this is subject to change as revisions have only just now begun to roll in.
Data Center Growth Accelerates 38 Points to 64% QoQ
SanDisk’s data center revenue growth was robust in Q2 with the company reporting growth of 76% YoY and 64% QoQ to $440 million, accelerating 86 and 38 points respectively. Data center still accounts for a smaller portion of overall revenue at almost 15% in the quarter, though this is up from 12% last quarter.
Management said they are seeing strong adoption of data center products from cloud hyperscalers, enterprise and edge data centers, and system integrators. SanDisk completed qualification of its PCIe Gen5 high-performance TLC SSDs at a second hyperscaler in the quarter, while two major hyperscalers are advancing with qualifications for its BiCS8 QLC ‘Stargate’ products, set to begin shipping in the next several quarters, providing another tailwind for growth.
Perhaps more importantly was management’s commentary about sequential growth through the second half of the fiscal year – we had outlined previously that management had stated in Q1 that they foresee sequential growth through the year with faster growth in 2H.
As we had mentioned above, management said that they believe data center growth will accelerate from here with a substantial step-up in Q3. Assuming this means a similar mid-30s acceleration to ~100% QoQ, this would project data center revenue to be ~$880 million in Q3, or more than 3X higher than where it entered the year. This also means reaching a $1 billion revenue quarter for the data center is increasingly plausible by Q4.
Looking at SanDisk’s other segments, edge revenue was up 21% QoQ and 63% YoY to $1.68 billion, with demand meaningfully exceeding supply with management citing PC replacement and AI adoption as driving higher storage content per device. Edge growth did decelerate from 26% QoQ in Q1 though YoY growth accelerated 33 points.
Consumer revenue was up 39% QoQ and 52% YoY to $907 million, accelerating from 27% YoY and 11% QoQ in Q1. SanDisk said product mix shifted toward higher-value configurations and premium products, supporting content growth.
Margins See Strong Expansion
SanDisk saw strong gross margin expansion in Q2 stemming from higher prices, while unit cost reductions served as an operating margin tailwind.
Q2 GAAP gross margin was 50.9%, up 21.1 points QoQ and 18.6 points YoY, while adjusted gross margin was very similar at 51.1%, up 21.2 points QoQ and 18.6 points YoY.
GAAP operating margin was 35.2%, up 27.6 points QoQ and 24.8 points YoY, while adjusted operating margin was 37.5%, up 26.9 points QoQ and 25.1 points YoY. Operating margins also benefitted from a (7%) QoQ decline in opex, which management said was due to a change in how they sell products: “basically, we're now moving into charging for our qualification units. So in the past, we used to record costs as they were incurred, right? They were period cost. And this is a nonrecurring element, which is a onetime gain as we move from period cost into inventories as we're now selling these qualification units.”
GAAP net margin was 26.5%, up 21.6 points QoQ and 21 points YoY, and adjusted net margin was 32%, up 24.2 points QoQ and 20.5 points YoY.
For Q3, SanDisk projected margins to expand further, guiding GAAP gross margin to be 64.9% to 66.9%, up 15 points QoQ and 43.4 points YoY at midpoint, while GAAP operating margin was implied to be 54.7% at midpoint, up 19.5 points QoQ. Adjusted gross margin was guided to be 65% to 67%, with adjusted operating margin guided to be 56% at midpoint.
Earnings
SanDisk reported a large beat on EPS in Q2, though arguably the Q3 guide could be one of the largest beats in tech, with management forecast Q3 adjusted EPS 200% above consensus estimates. Rough estimates for Q4 show EPS potentially moving even higher from Q3’s strong growth.
GAAP EPS was $5.15 in Q2, up 587% QoQ and 615% YoY, and nearly $2 ahead of consensus estimates for $3.20. Adjusted EPS was $6.20, beating the $3.78 estimate by 64% and representing 408% QoQ and 404% YoY growth.
For Q3, management guided for $12 to $14 in adjusted EPS, up 110% QoQ, and coming in 200% above consensus estimates for $4.33 at midpoint. Based on current estimates for ~20% QoQ growth to $5.53 billion and slight margin expansion of 3-5 points, this would project Q4 adjusted EPS in the $19.50-$20.50 range.
This assumption would take FY26 adjusted EPS up to ~$40.40, at the midpoint of Q3 and Q4’s rough estimate, or growth of ~1,250% YoY. Current consensus points to $34.92 in adjusted EPS, up ~1,068% YoY.
Cash Flows and Balance Sheet
Cash flow margins improved significantly, and SanDisk paid down a substantial amount of debt in the quarter.
Operating cash flow was $1.02 billion, up 973% YoY and for a 33.7% margin, QoQ and 28.6 points YoY.
Free cash flow was $980 million and adjusted FCF was $843 million for a 27.9% margin, up 8.5 points QoQ and 23 points YoY – note the margins are for adjusted FCF as it includes expenses related to the Flash Ventures JV with Kioxia for manufacturing. Management commented that any material increases in capex would “require high confidence that demand at attractive pricing levels is durable over a several year horizon,” suggesting that SanDisk will not preemptively increase capacity if it does not believe pricing will support strong profitability.
Cash was $1.539 billion, and debt was $603 million, down from $1.85 billion two quarters ago as SanDisk paid off another $750 million in debt this quarter.
Inventories were $1.97 billion, up marginally from $1.91 billion in the prior quarter.
Valuation
For a quick update on the valuation, shares are trading at 5.7x forward PS on the current $15.46 billion consensus, while on the bottom line, shares are trading at 17x forward PE on the current consensus of $34.92, or 14.8x on our estimate for $40.40.
Conclusion
SanDisk reported a strong Q2 with Q3 showing meaningful acceleration across revenue and earnings, with the company riding strong demand and strong ASP tailwinds. QoQ growth is expected to accelerate for the data center segment in Q3 with the back half being much stronger, per management, while overall revenue growth was guided to 52% QoQ for Q3, the highest among the AI semiconductor names that we track.
Although we don’t expect memory stocks to go up in a linear fashion, it’s hard to imagine an outcome where 2026 is not dominated by this subsector of AI stocks. What the I/O Fund is trying to wrap our heads around, is that we are seeing Blackwell and Blackwell Ultra impact, whereas incoming Rubin supports a sustained trend for NAND. Management commentary around the KV cache—something we outlined in our Q1 2026 Top 15 report—was particularly notable, and perhaps the most important takeaway from the call was the expectation that this acceleration continues beyond next quarter.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund do not own shares in SNDK at the time of writing and may own stocks pictured in the charts.
GE Vernova exited 2025 with one of the strongest AI demand and backlog profiles in the energy industry. This quarter, management emphasized accelerating slot reservations, rising pricing, improving backlog margins and multi-year visibility extending into the end of the decade.
The company signed 6GW of incremental gas contracts in the final three weeks of December, bringing total Q425 as contracts to about 24GW. As a result, the Gas Power backlog plus slot reservation agreements (SRAs) expanded from 62GW to 83GW sequentially.
Management now expects to reach 100GW under contract in 2026, an upward revision from the 60GW discussed in mid-2025. Notably, the current 83GW under contract is heavily allocated toward 2029 delivery. By the time that 100GW is reached, both 2029 and 2030 capacity will be sold out.
The Importance of Slot Reservation Agreements (SRAs)
On the earnings call, management stated that reservation agreements signed today are priced approximately 10 to 20 points higher than legacy backlog, confirming the capacity scarcity could translate to higher growth.
Here is what was stated about the potential for SRAs to grow incremental growth due to higher pricing:
“When we look at where we're trending with our slot reservation agreements today versus our existing backlog, there's another 10 to 20 points of pricing strength in the SRAs today. We are pleased — you're right, we were talking in the middle of last year at 60 gigawatts and landed at 83 gigawatts because the intensity of the discussions, really late summer, fall, right through the holidays have continued to be very intense. When you think about this year getting to 100 gigawatts by the end of the year, what I would tell you is it's likely going to be a larger proportion of orders. Today, with the 83 gigawatts, it's 40 gigawatts of orders, 43 gigawatts of SRAs, that probably shifts towards more of a 60-40 split with 60% on order over the course of 2026.”
As these higher-priced reservations convert into firm orders over time, management expects backlog margins and future earnings power to continue improving:
“We expect significant growth again in Power and Electrification's backlog in '26 at better margins as we convert higher-priced gas slot reservation agreements into orders and benefit from strong demand and pricing for grid equipment.”
Equipment Orders Surge to 91% YoY Growth
Organic orders surged with growth of 65% YoY to $22.2 billion and equipment orders grew 91% YoY to $16.2 billion. Services grew 22% to $6 billion.
GE Vernova exited this year with total backlog of $150 billion, up 26% YoY. Equipment backlog reached $64 billion, a 49% increase, while services backlog grew to $86 billion. Management stated they added an incremental $8 billion in incremental margin dollars to equipment backlog in 2025 alone, exceeding the prior two years combined. Because equipment delivery cycles are long, the majority of these higher-margin orders will not begin contributing to revenue until 2027 and beyond, creating a lag between reported growth and underlying earnings power.
Electrification backlog reached $35B, up $11B year-over-year and represented GEV’s largest-ever growth in this segment. This is due to demand for substations, switchgear, HVDC and grid equipment.
Prolec Acquisition:
Prolec designs and manufactures medium-and-large power transformers that was owned 50% by GE and 50% by Mexican industrial group Xignux. The acquisition makes GE Prolec a wholly owned subsidiary. For the purchase price of $5.275 billion at closing, Prolec GE expects $3 billion in revenue at a 25% adjusted EBITDA margin with “low double-digit revenue growth in the coming years.” The electrification segment is guided to $13.5 billion to $14 billion in 2026 with Prolec representing 20% to 22%.
Transformers are a leading bottleneck with lead times of 2-4 years, to where they’ve become a gating item for AI data center power connections and grid expansion. Even if generation is available, transformers are needed to deliver power to an AI data center, which leads to a direct path for Prolec’s importance in the AI buildout. Now that Prolec is fully integrated, GEV’s electrification segment will benefit from owning one of the more capacity-constrained parts of the AI buildout. In exchange, this will lead to higher margins, pricing power and backlog visibility for GEV. Transformers are higher-margin with GEV’s electrification EBITDA margin at 14.9% in 2025 and now guided to 17% to 19%. The deal is also accretive to free cash flow within the first year.
Financials
Q4 Revenue Beat of 7.1%
GE Vernova Q4 revenue grew by 3.8% YoY to $10.96 billion, beating estimates by 7.1%, driven by rising AI energy demand. Organic revenue grew by 2% YoY to $10.8 billion. The company is a major beneficiary of the increasing energy requirements from the global AI infrastructure build-out, positioning the company as a key beneficiary of this secular trend. The continued slowdown in the Wind segment was offset by the growth in power and electrification segments that are benefitting from rising electricity consumption driven by data centers and artificial intelligence demand.
The company’s revenue growth is expected to accelerate to 9.8% YoY growth to $8.8 billion in Q1 and is expected to grow 7.8% YoY to $9.82 billion in Q2 2026.
Full-year 2025 revenue grew 9% year over year to $38.1 billion. Management expects 2026 revenue of $44–$45 billion, up from prior guidance of $41–$42 billion provided at the December 2025 Investor Update, now reflecting the acquisition of the remaining 50% stake in Prolec GE, which is expected to close in February.
Management also increased the by 2028 revenue outlook to $56 billion from $52 billion with low-teens organic growth during the period 2025 to 2028.
Segments
Q4 Power Orders grew by 77%
Q4 power orders increased 77% YoY to $11.7 billion, driven primarily by a sharp acceleration in gas power equipment orders, which more than tripled on higher volumes and favorable pricing. Gas turbine orders rose 71% YoY to 41 units, while power services orders grew 15%, reflecting continued customer investment in existing fleets.
Q4 power segment revenue grew organically by 5% YoY to $5.7 billion. Management expects high single-digit organic growth in Q1.
EBITDA margin improved by 360 basis points sequentially and 200 basis points YoY to 16.9%, primarily driven by pricing and productivity gains more than offsetting incremental costs associated with capacity expansion, R&D investments, and inflationary pressures. Management expects the EBITDA margin to be 240 basis points lower sequentially, due to seasonality, to 14.5% in Q1. However, it would be up 300 basis points YoY.
Wind Segment recovery expected in 2H 2026
Q4 Wind Segment organic revenue was down (25%) YoY to $2.34 billion primarily due to lower onshore wind equipment deliveries. Management expects organic revenue to be down high teens in Q1 due to lower onshore equipment deliveries.
Wind orders increased 53% YoY to $3.1 billion, driven by stronger onshore equipment demand, primarily outside North America. Management remained cautious about calling an inflection in U.S. orders, citing ongoing project delays and tariff-related uncertainty. In offshore, the company continues to prioritize execution of its challenged backlog.
Q4 EBITDA losses were ($225 million) or EBITDA margin of (9.5%) compared to 0.60% in the same period last year and (2.3%) in Q3. EBITDA losses widened, driven by higher losses on Offshore Wind contracts, including the impact of the recently issued U.S. order to halt construction of all offshore projects and lower Onshore Wind equipment volumes, partially offset by improved performance in Onshore Wind services. Management expects Q1 EBITDA losses of $300 million to $400 million due to lower onshore wind volume and tariffs.
Management expects a strong recovery in the second half of 2026. The company’s CFO, Kenneth Parks, said in the earnings call, “Looking at 2026, we expect significant improvement in Wind revenue in the second half of the year given only 30% of our expected onshore turbine shipments are in the first half as almost 70% of our 2025 equipment orders came later in the year. Also, the volume we're shipping in the first half has fewer contractual protections for tariffs since we signed these orders before their implementation. As a result, we expect EBITDA losses in the first half to be partially offset by profitability in the second half.”
Electrification Q4 Orders 2.5x of revenue
Electrification orders were 2.5x revenue and were up 50% YoY to $7.4 billion primarily due to growing grid equipment demand, particularly for synchronous condensers, substations partially to support data center growth and switchgear. The company also witnessed strong equipment orders growth in the Middle East, which increased over $1 billion and in North America, which more than doubled YoY.
Q4 organic electrification revenue grew by 32% YoY to $2.9 billion primarily driven by strong growth in switchgear and High-Voltage Direct Current (HVDC) equipment. Management expects a similar revenue as Q4 in the next quarter which will also include Prolec GE.
Q4 EBITDA margins improved 410 basis points YoY to 17.1% primarily due to strong volumes, productivity gains, and favorable pricing. Management expects Q1 EBITDA margin of 16.5%.
Adjusted EBITDA grew by 7.3% in Q4
The company’s Q4 adjusted EBITDA grew by 7.3% YoY to $1.16 billion with an adjusted EBITDA margin of 10.6%, an improvement of 250 basis points sequentially and 40 basis points YoY. Organic adjusted EBITDA margin improved 10 basis points YoY to 10.7%.
2025 adjusted EBITDA margin improved 260 basis points YoY to 8.4% and was in-line with the management mid-point guidance of 8.5%. Management expects 2026 adjusted EBITDA margin to improve to 12% in 2026 driven by growing backlog, favorable pricing, and improved operational execution. Management also expects adjusted EBITDA to be more second half weighted with highest revenue and adjusted EBITDA in Q4 2026.
Q4 net income was $3.7 billion or 33.5% of revenue compared to $484 million or 4.6% of revenue in the same period last year. The Q4 net income included a one-time tax benefit of $2.9 billion.
EPS
Q4 GAAP EPS was $13.39, up from $1.73 in the prior-year period, reflecting a one-time tax benefit of $10.58. Excluding this benefit, GAAP EPS would have been $2.81, below the consensus estimate of $3.13, primarily due to losses in the Wind segment.
Analysts expect strong EPS growth in the coming quarter with Q1 EPS expected to grow 127.7% YoY to $2.07 and Q2 EPS to grow 65.1% YoY to $3.07.
Cash Flow and Balance Sheet
GE Vernova is funding this growth from a position of improving financial strength. In December 2025, S&P and Fitch upgraded their investment grade credit rating to BBB from BBB-, and BBB+ from BBB, respectively. Both maintained positive outlooks on their upgraded ratings.
The company exited 2025 with $8.85 billion in cash and generated $3.7 billion in free cash flow, more than double the prior year. Gross debt will remain below 1x EBITDA even after funding the Prolec GE acquisition. In early February, the company expects to issue roughly $2.6 billion of debt in order to complete the previously announced acquisition of the remaining 50% ownership stake of Prolec GE.
Capital returns accelerated alongside growth investments, including a doubled dividend for 2026 and an expanded $10 billion share repurchase authorization. The company had cash of $8.85 billion and no debt at the end of Q4.
The company’s cash flows are improving driven by growth in profits and also improvement in working capital.
Q4 operating cash flows grew by 169% YoY to $2.48 billion with an operating cash flow margin of 22.6% compared to 8.7% in the same period last year. The company benefitted from down payments on higher orders and slot reservations at Power as well as higher orders at Electrification.
Q4 free cash flow grew by 214.7% YoY to $1.8 billion with a free cash flow margin of 16.5% compared to 5.4% in the same period last year.
Conclusion
GEV is a rare, quality stock in the AI space that is buffered from competition. The company will see the full weight of the United States behind its efforts as its well positioned to offset the many GWs the AI buildout needs. The demand is unquestionably high, but how fast GEV can manufacture it and what price can GEV get for that capacity. Although discussions are stretching into 2029-2030, the SRAs signed in previous years for equipment orders can offer a pricing uplift and equipment margin expansion. GEV is also expediting gas turbines with 200 machines installed in 2025 and another 200 planned for 2026. With the Prolec acquisition, GEV is also becoming a strong contender on the electrification side.
It is my best guess that when higher-beta AI stocks sell off (as they inevitably always do) that GEV offers a steadier and more of a safer, quality hedge for that trade. Keep in mind, since the AI boom began on Jan 1st 2023, GEV has outperformed Broadcom, AMD, Micron and TSM by 2X or more – proving GEV is anything but a sleepy energy stock.
Royston Roche, Equity Analyst at I/O Fund contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
SanDisk was the best stock in the S&P 500 in 2025 with a 559% return, and the company is continuing this strong performance in 2026 with shares up 41% in barely two weeks. Much of this performance is rumored to be linked to Nvidia’s CES presentation where the GPU leader discussed context windows as the next bottleneck for AI inference, and hinted at rack-scale and network solid state drives (SSDs) becoming key components to address this.
On a broader level, data center/enterprise SSDs are often overlooked but equally critical as HBM when it comes to AI training and inference. This is because data center SSDs offer higher read-write speeds critical for accessing and transferring data rapidly, along with higher performance and energy efficiency, vital factors for larger-scale AI training and inference workloads.
SanDisk operates independently after being spun out of Western Digital in February 2025. The company expects to ride enterprise SSD demand tailwinds with management projecting sequential growth in its data center segment through 2026, with two hyperscaler qualifications underway and an additional hyperscaler expected in 2026.
However, data center remains a smaller portion of revenue, contributing $269 million last quarter or less than 12% of revenue, with client (PC/smartphones) and consumer products (SD cards/USB) remaining core to its business.
The Context Bottleneck, and Extending KV Cache to SSDs
Moving to some longer-term tailwinds for SSDs, Nvidia discussed how context windows could soon be the new bottleneck for AI inference performance at CES this week, as it unveiled its new Inference Context Memory Storage platform (ICMS) to address growing key-value (KV) cache capacity limits. KV cache capacity is a known pain point when working to balance long-context reasoning and memory capacity.
Put simply, the KV cache is a memory optimization technique that stores calculations during the inference phase, allowing the model to remember those prior calculations instead of repeating them, thus enabling faster response times; without it, latency would be much higher and computation much slower as every new token would require recalculation of all prior tokens. KV cache essentially serves as a model’s long-term memory that is reused and extended throughout many steps or requests.
However, the KV cache has a substantial memory footprint, especially for long contexts, and during deployment it can consume ~30% of GPU memory, making it a major bottleneck for large-context applications, such as coding, natural language processing, or handling simultaneous requests from many users on large models.
Let’s first take a look at KV cache from a tensor parallelism perspective, and how distributing memory across tens to thousands of GPUs significantly increases available KV cache memory, translating directly to larger maximum token context windows.
For example, a single AMD MI300X GPU running a Llama-70B model would have ~17GB of memory capacity available for the KV cache, after accounting for 140GB to store model parameters (2 bytes per parameter on FP16, so 70 billion * 2), and 35GB for the activation buffer (estimated ~25% of model storage), per TensorWave. With an estimated 2.6MB required per token (or 2.6GB per 1K tokens), a single GPU can handle a max request of ~6,500 tokens.
When you distribute model parameter and activation buffer across an 8-GPU server, or ~17.5GB and 4.4GB per GPU, this frees up 170.6GB per GPU for the KV cache, a ~10X increase; for the server, KV cache memory availability is now 1,360GB. This means that an 8-GPU server could now handle a max request of 523,000 tokens, an ~80X increase, with gains that only compound as server size and memory increase. This can then be optimized for longer contexts, or 8 requests of 64k context lengths, for high-throughput, or 64 requests of 8k context lengths, or other combinations.
Here’s where SSDs fit in – by extending or offloading KV cache to local SSDs, model prefill and time to first token can be significantly reduced, thus significantly decreasing latency and increasing throughput.
AI inference acceleration startup WEKA states that when it tested Llama-3.1 70B with no optimizations, a 100K token prompt took 24 seconds to prefill into the model before any output could be generated, but “extending GPU memory to ultra-fast storage [NVMe SSDs] can dramatically improve token processing efficiency.” When configuring an Nvidia DGX H100 server with an 8-node exabyte-scale NVMe SSD pod, WEKA says its “tests demonstrated a staggering 41x reduction in prefill time on LLaMA3.1-70B, dropping from 23.97 seconds to just 0.58 seconds,” significantly improving model efficiency with zero optimizations – simply from adding SSDs to extend GPU memory.
Google ran tests using an 8-GPU H100 server on Llama-3.3 70B, extending system memory to larger, lower-cost CPU RAM and enterprise SSD tiers. For a ~4 million token cache, Google found that utilizing CPU RAM and SSDs in conjunction with HBM decreased end-to-end latency and time to first token by 64% to 79%, while increasing throughput by 179-264% on 50k to 100k token prompts.
Nvidia Working to Tackle the Context Bottleneck with its ICMS Platform
With its new ICMS platform, Nvidia is working to mitigate context windows becoming a major bottleneck as agentic AI and physical AI scale over the coming years. Scaling of models to trillions of parameters and the shift to multi-step reasoning, multi-agent workflows or advanced multimodal applications will generate substantial volumes of context data and require significant KV cache reuse to maintain accuracy and context in prolonged interactions.
As such, Nvidia believes that “AI factories need a complementary, purpose‑built context layer that treats KV cache as its own AI‑native data class rather than forcing it into either scarce HBM or general‑purpose enterprise storage.” For example, the current inference context hierarchy begins with HBM (G1), providing near-instant access to latency-critical context in active generation, down to SSDs (G3) in the third tier to handle ‘warm’ data, or data that is used regularly but less frequently and still requiring efficient, cost-effective storage. Enterprise or shared storage sits at the bottom of the hierarchy (G4), handling ‘cold’ data, or data stored for long-term retention but much less frequently accessed.
Nvidia is essentially proposing an architectural redesign of this hierarchy, positioning the new ICMS platform between G3 and G4, or as it calls it, G3.5. ICMS is a new Ethernet-attached NVMe SSD storage tier, likely integrated into the fabric and optimized specifically for KV cache usage at the pod level. It is powered by Nvidia’s new BlueField 4 data processing unit (DPU) packing 512GB of on-board SSD capacity, a 4x increase from BlueField3’s 128GB, and is combined with Nvidia’s GPUDirect Storage, which bypasses the CPU and provides direct memory access from GPU to SSDs to reduce latency.
ICMS will provide petabytes (millions of GB) of shared KV memory capacity per GPU pod, capable of storing context for many models or agents simultaneously, while being located close enough to GPUs to frequently share inference context with lower power consumption and better efficiency versus shared storage. Nvidia claims ICMS can enable up to 5x improvements in power efficiency and 5x increases in tokens per second versus shared storage.
It is this new platform and Nvidia CEO Jensen Huang’s comments relating to the storage hierarchy redesign that have fueled optimism for SanDisk and SSDs, as Huang believes it is a completely new market, integrating SSDs to the fabric, that could ultimately become the largest storage market:
“For storage, that is a completely unserved market today. The way that storage works is SQL. SQL is structured data. Structured database is lightweight. AI database KV caches insanely heavy weight. You're not going to hang that off of your north-south network. I mean that's just a horrible waste of network traffic. You want to put it right into the computing fabric, which is the reason why we introduced this new tier.
This is a market that never existed. And this market will likely be the largest storage market in the world, basically holding the working memory of the world's AIs. And that storage is going to be gigantic, and it needs to be super high performance.”
Because Nvidia is positioning NVMe SSDs to become the backbone for this new shared memory tier, there is the potential for SSD suppliers to see solid medium/long-term tailwinds from increased SSD capacity requirements in inference-optimized deployments over the next few years. For example, Bernstein estimates that Huang’s CES comments on SSDs and KV cache requirements suggest an additional 16TB per GPU, compared to 3-4TB per GPU today, or 4-5X growth. This will be more weighted towards year-end and into 2027 as ICMS rolls out with Rubin.
SanDisk’s BiCS8 Tech, Kioxia JV and Data Center ‘Stargate’ SSD Line
SanDisk is eyeing strong growth in the enterprise SSD market with its ‘Stargate’ NVMe SSD products, based on its BiCS8 architecture jointly-developed with Kioxia, offering industry-leading capacity, energy efficiency and performance.
NVMe (Non-Volatile Memory Express) is a protocol designed specifically for NAND-flash based SSDs that optimizes performance by reducing latency and increasing data transfer speeds by utilizing the PCIe bus, enabling high throughput and fast data transfer speeds necessary for AI training and inference.
SanDisk and Kioxia’s joint venture is one of the longest-standing JVs in the industry, signed in 2000 and lasting through 2034. It is primarily a shared manufacturing and capex strategy, with the two both splitting JV capex and wafer output and then selling NAND products independently. For example, the mega-fab in Yokkaichi, Japan produces nearly one-third of all global NAND bits (500K wafers per month), with the JV taking 80% capacity, split 50/50 between SanDisk and Kioxia, and Kioxia taking the remaining 20% (for an overall split of 60-40 for Kioxia and SanDisk).
The two also recently started operations at their new second fab in Kitakami, Japan, which is geared towards BiCS8 3D NAND and future advanced 3D NAND production, with output expected to ramp meaningfully in the first half of 2026. This will help scale SanDisk’s enterprise SSD line, based on BiCS8, and potentially aid in future development of high-bandwidth flash. BiCS8 accounted for 15% of bits shipped in fiscal Q1 and is expected to reach majority of bit production exiting FY26, providing a clue into the ramp profile for the year.
BiCS8 is the duo’s eighth-generation BiCS (bit cost scalable) 3D NAND architecture, which stacks NAND cells vertically, creating more layers and reducing costs per bit. BiCS8 scales to 218 layers from 162 layers in BiCS6, with SanDisk saying that BiCS8 increases memory density by more than 50%, program and read bandwidth by 35% and 26%, and data transfer speeds by more than 80% versus BiCS6. The two also have previewed the next generation of BiCS, scaling to 332 layers and further improving interface speeds by ~33%.
Additionally, SanDisk believes that its BiCS8 QLC (quad-level cell) die underpinning its Stargate data center SSDs delivers substantial performance, latency and efficiency advantages over competitors: 11% to 67% faster input/output speeds in Gb/s, along with 27% to 34% lower latency. Management expects its BiCS8 QLC line to go from ~20% to 40% of its data center business by the end of FY26.
SanDisk’s ‘Stargate’ line, built on BiCS8, debuted this year with 64TB and 128TB capacities now shipping. 256TB products are scheduled for launch in mid to late 2026 and 512TB targeted in 2027, with the combination of fast performance, low latency and high capacity making the SSDs suitable for managing massive AI datasets and workloads. SanDisk says Stargate “is growing in demand with 2 hyperscaler qualifications underway and a third hyperscaler along with a major storage OEM planned for calendar year '26,” with current qualification focused on 128TB.
SanDisk also says that its SN861 NVMe SSDs were the first to be certified to support Nvidia’s GB200 NVL72, though this does not mean that its shipments will be correlated 1:1 with Nvidia’s racks. Management explained that the certification puts them on the approved vendor list, and “when the ODMs are picking their design, they pick vendors from the approved vendor list. So that's how we are getting into qualifications with the system partners that are building on top of Nvidia, not necessarily the Nvidia build.”
‘AI SSDs’ and a Path to ~33X Increase in Performance
Nvidia’s ICMS also could create a new market for SLC-based ‘AI SSDs’ in the data center, which had predominantly focused on TLC and QLC-based SSDs. SLC, or single-level cell, stores one bit of data per cell, offering the fastest data retrieval and best performance, though it is typically the most expensive; TLC and QLC (triple-level and quad-level) store three and four bits per cell, increasing storage capacity significantly and reducing cost but with slower performance versus SLC. Overall, SanDisk’s role in these first performance-based ‘AI SSDs’ remains somewhat unclear, though the company is playing a much more integral role in high-bandwidth flash (HBF) development.
The AI SSD push is currently being spearheaded by SK Hynix and Kioxia, and ultimately aims to boost SSD performance by up to 33X by 2027. SK Hynix is creating a three-product family, ‘AI-N’, with its AI-N P focusing on maximizing performance, AI-N B on maximizing bandwidth, and AI-N D on maximizing storage density. Reports suggest that Hynix’s first-gen AI-N P product will target approximately 25 million input/output operations per second (IOPS), or how many read/write operations a device can perform per second, with a higher number meaning faster performance. This is about an 8X leap from current SSDs at ~3 million IOPS today, while Hynix’s second-gen and Kioxia’s product in conjunction with Nvidia are said to be targeting 100 million IOPS by 2027, a 33X increase.
High-Bandwidth Flash Could Boost GPU Memory by 21X
HBF is being proposed as a future alternative/replacement for HBM memory on GPUs, stacking up to 16 3D NAND BiCS8 dies using through-silicon vias (TSV) to deliver up to a 21X capacity boost with similar bandwidth and cost as HBM.
For example, SanDisk’s first-gen HBF could pack capacity of 512GB per stack with HBM-like bandwidth. At partial HBM replacement, such as in six out of the eight dies on the GPU package, HBF could boost total GPU memory to ~3,120 GB per GPU, a nearly 17X increase versus individual Blackwell GPUs featuring 186GB HBM per GPU. In full replacement of HBM, HBF could provide 4,096 GB of total memory per GPU.
The significance of this is that it could allow frontier models to be stored entirely within a single GPU, rather than needed to be partitioned across a rack. SanDisk explains, “Think about a frontier large language model, let's say, something like GPT-4. GPT-4 has 1.8 trillion parameters with 16-bit rates. And that model alone would take 3.6 terabytes or 3,600 gigabytes of memory space. I can put that entire model on a single GPU now. I don't need to shuffle around data any longer.”
In terms of potential commercialization of HBF, Hynix’s aforementioned AI-N B line is built on HBF and expected to be developed in collaboration with SanDisk. The first ‘alpha’ version sample could be released as early as January with the first proof-of-concept samples in 2027, followed by full-scale evaluation afterwards.
SanDisk stated that the “gating item indeed is going to be enabling the ecosystem, aligning with the customers at their system level, integrate it and then bring it to the market,” but the product is highly executable as it is based on its existing NAND architecture. CEO David Goeckeler clarified in Q3 that the company “announced a time line last quarter of having the memory later in '26 and then having the controller for that in '27 we're still working towards that time line.”
However, timelines for HBF are still unclear given the newness of the technology, with some estimates suggesting 2027 to 2028 as a possibility. It is far too early to tell whether HBF will be commercially viable or successful.
Training and Inference are Long-Term SSD Demand Drivers
This section includes a brief excerpt from our 6,000+ word thematic deep dive into the current AI memory boom recently published for our Pro subscribers:
AI training and inference are two main long-term drivers for SSD demand, which is projected to rise ~6X from 2024 to 2030, from 181 exabytes (EB, or equal to 181,000,000 TB) to 1,078 EB, under McKinsey’s base case scenario. Training demand projected to rise at a 62% CAGR to from 7 EB in 2024 to 127 EB by 2030. On the flipside, demand from AI inference is expected to grow at a 105% CAGR from 6 EB to 447 EB by 2030, giving inference a 41% share of demand versus less than 12% for training. base case scenario. Training demand projected to rise at a 62% CAGR to from 7 EB in 2024 to 127 EB by 2030. On the flipside, demand from AI inference is expected to grow at a 105% CAGR from 6 EB to 447 EB by 2030, giving inference a 41% share of demand versus less than 12% for training.
This is not only driven by development of more LLMs, but also the increasing size and complexity of frontier models, where training data sets and context windows for inference are getting increasingly large.
For example, EpochAI estimates that training data set sizes are rising 3.7X per year on average, or nearly doubling every six months, though there are some models that are scaling much quicker. For example, Meta’s Llama2-70B from 2023 was trained on 2 trillion tokens, while Llama3-70B, from 2024, was trained on 15 trillion tokens, a 7.5X increase. Multi-modal models, those integrating audio, video, image or more, are also likely to require significantly more SSD storage, with McKinsey estimating in the hundreds of TBs depending on the mix of data needing to be stored. estimates that training data set sizes are rising 3.7X per year on average, or nearly doubling every six months, though there are some models that are scaling much quicker. For example, Meta’s Llama2-70B from 2023 was trained on 2 trillion tokens, while Llama3-70B, from 2024, was trained on 15 trillion tokens, a 7.5X increase. Multi-modal models, those integrating audio, video, image or more, are also likely to require significantly more SSD storage, with McKinsey estimating in the hundreds of TBs depending on the mix of data needing to be stored.
The increasing size and complexity of models also ties directly to a major pain point when it comes to inference: “As models grow in complexity and require longer contexts, their memory footprint expands beyond what a single GPU can handle. This results in inefficiencies where GPUs are memory-starved, causing significant bottlenecks in AI token generation.” inference: “As models grow in complexity and require longer contexts, their memory footprint expands beyond what a single GPU can handle. This results in inefficiencies where GPUs are memory-starved, causing significant bottlenecks in AI token generation.”
This is exactly what Nvidia is addressing with Rubin and ICMSP, creating a new storage tier within the cluster fabric that is designed to extend GPU memory and facilitate high-speed KV cache distribution among racks.
There are also tailwinds to SSD growth from increasing cluster sizes, with compute-focused eSSDs seeing a 1:1 attach rate per GPU. For example, SanDisk says that a real-world 32,256 GPU cluster (or eight pods of 252 16-GPU racks) would require 4,032 compute eSSDs such as its SN861 product. This could create a strong tailwind for SSD growth as clusters scale to 100K+ GPUs towards 1 million, assuming the correlation for compute eSSDs to GPU remains 1:1.
Financials
Revenue
SanDisk reported a strong sequential revenue acceleration in its fiscal Q1, driven by NAND demand outpacing supply and increasing demand in its data center, edge and consumer end markets. Q1 revenue increased 22.6% YoY and 21.4% QoQ to $2.31 billion, accelerating from 8% YoY and 12.2% QoQ growth in fiscal Q4. Higher-than-expected bit growth drove the outperformance in the quarter relative to guidance of $2.1-2.2 billion, per management.
SanDisk’s Edge segment was the primary growth driver in Q1 with revenue up 30% YoY and 26% QoQ to $1.39 billion, driven by increasing NAND content in PCs and smartphones and a positive PC refresh cycle. Consumer revenue rose 27% YoY and 11% QoQ to $652 million, while data center revenue was down (10%) YoY but up 26% QoQ to $269 million.
Q2 revenue was guided to be $2.55 to $2.65 billion, up 38.6% YoY and 12.6% QoQ at midpoint. CFO Luis Visoso clarified that “the key message is most of the growth in revenue will be pricing driven in the quarter.”
Revenue growth is then expected to accelerate further to 55% YoY in fiscal Q3 (even with a seasonal slowdown in consumer products following the holidays) and then decelerate slightly to 51% in Q4. Pricing tailwinds could strengthen significantly in FQ3 on reports that NAND prices for enterprise SSDs could rise ~100% QoQ in the March quarter, according to supply chain checks by Nomura. Citi estimates SSD prices will rise ~32% QoQ in the March quarter following a 21% QoQ increase in the December quarter.
For fiscal 2026, SanDisk is currently expected to generate revenue of $10.6 billion, up 44.1% YoY. SanDisk sees demand outpacing supply through the entire year, currently estimating supply to support mid-teens demand growth, and potentially lead to strong pricing tailwinds from this tight/tightening environment:
“We saw supply growth in calendar year '25 of about 8%. We see it at about 17% in '26. We see demand — constrained demand around 14% [mid-teens] because that's all that's out there from a supply point of view. But unconstrained demand is in the — literally, a couple of weeks ago, we thought it was 20%, it's probably mid-20s by now. So we see the supply pretty much being able to service that kind of mid-teens level demand for '26.”
This tightening environment comes despite fabs running at 100% utilization, with management adding that they do not plan on adding capacity to any end market, but rather remain prepared with the optionality to shift capacity as visibility into product mix strengthens.
AI Segment Growth
SanDisk’s data center revenue, as mentioned above, declined (10%) YoY but rose 26% QoQ to $269 million, driven by increasing demand for its ‘Stargate’ enterprise SSD product line. However, revenue contribution remains small, at less than 12% of revenue.
SanDisk did not provide a numerical guide for Q2 for data center, but management noted that they are expecting sequential growth throughout fiscal 2026 with faster growth in the back half, driven by the current hyperscaler qualifications planned and underway. SanDisk did clarify that they are “working with 5 major hyperscale customers through active sales and strategic engagements” across its data center portfolio.
Data center growth is supported by solid visibility, with management explaining that they are either “striking deals that are multi-quarters, let's say, through the first half of next calendar year” from customers looking to lock in supply, or working with customers with demand visibility through 2027 to align supply with those demand forecasts. Management also sees undersupply conditions extending potentially into 2027 now, supporting strong pricing in deal negotiations.
Management also increased their forecast for data center exabyte growth, explaining that last quarter, exabyte growth expectations were in the mid-20% range, but now are in the mid-40% range. As a result, data center is expected to be the largest market in NAND on an exabyte basis in 2026, surpassing mobile.
Earnings
SanDisk stands out for its strong expected earnings growth through fiscal 2026 and fiscal 2027, with adjusted EPS expected to reach more than $21 by then, or >7X higher than the $2.99 it earned in fiscal 2025.
Q1 GAAP EPS was $0.75, a strong improvement from a ($0.16) loss in Q4, though this was down (49%) YoY from $1.46 in the year ago quarter as margins remained lower YoY. Adjusted EPS was $1.22, up 321% QoQ but down (33%) YoY.
For Q2, SanDisk guided for adjusted EPS of $3.00 to $3.40, up more than 162% QoQ. Adjusted EPS is expected to further increase to $3.78 in fiscal Q3 and $4.82 in fiscal Q4.
For fiscal 2026, SanDisk is expected to generate $13.29 in adjusted EPS, up 344.6% YoY, while GAAP EPS is projected to be $11.53, up from ($11.32) in FY25 due to the spin off. Fiscal 2027 is expected to see earnings power surpass $21, with GAAP EPS estimated to be up 86% to $21.47 and adjusted EPS up nearly 62% to $21.50.
Margins
Margins are lower YoY compared to pre-spinoff margins, but Q1 saw strong sequential margin expansion that is expected to accelerate in Q2.
Q1 GAAP gross margin was 29.8%, down 8.8 points YoY but up 3.6 points QoQ. Adjusted gross margin was 29.9%, down 9 points YoY but up 3.5 points QoQ.
GAAP operating margin was 8.3%, down 8.3 points YoY but up 5.6 points QoQ. Adjusted operating margin was 10.6%, down 8.2 points YoY but up 5.3 points QoQ.
GAAP net margin was 4.9%, down 6.3 points YoY but up 2.7 points QoQ, and adjusted net margin was 7.8%.
For Q2, SanDisk guided adjusted gross margin to be 41-43%, or up just over 12 points QoQ at midpoint on higher pricing and cost reduction tailwinds, while adjusted operating margin is implied to be 24.2% at the midpoint of opex guidance, or up 13.6 points QoQ. Fab startup costs are expected to transition from headwinds to tailwinds during the quarter, potentially aiding more margin expansion into fiscal Q3 and Q4.
Cash
SanDisk noted that in Q1 it reached a net cash position, six months ahead of schedule, though debt is still almost equivalent to its cash on hand. Cash flows were quite strong, and adjusted FCF margin showed strong expansion.
Operating cash flow was $488 million in Q1 for a 21.1% margin, up from a (7%) margin in the year ago quarter and a 4.9% margin in Q4.
Adjusted free cash flow was $438 million in Q1 for a 19% margin, up from a (10.5%) margin in the year ago quarter and 2.6% in Q4.
SanDisk’s total gross capex to support the JV was $387 million in Q1, though its cash capex spend was only $40 million (1.7% of revenue) as the remainder was funded through external sources such as subsidies or tool depreciation recorded in COGS.
Cash and equivalents totaled $1.44 billion while debt totaled $1.35 billion.
Valuation
SanDisk’s valuation is somewhat hard to pin down given the company’s limited history on the public markets after its February spinoff, and its rapid 362% ascent since the end of August.
SanDisk trades at 5.3x forward PS, surpassing its prior peak at 4x in November and a substantial re-rating higher from 0.6x in the summer. For comparison, this is now on par with former parent Western Digital at 5.1x forward PS, though the two are focused on different memory market segments with WDC primarily in hard disk drives.
For forward PE, SanDisk currently trades at an 28.7x multiple, nearly double its 15.8x average from the second half of fiscal 2025 prior to its fiscal year readjustment in June. Shares traded as low as 3x in July and August due to the sharp earnings increase expected in fiscal 2026.
Notable Risks
The NAND flash market has historically been quite volatile, and is shifting from significant oversupply in 2023 to expectations for substantial supply shortages through 2026 and potentially into 2027. However, if NAND capacity begins to come online quickly through next year, or if demand for PCs and smartphones falters due to rising memory prices, the NAND cycle could reverse and lead to pricing pressures cutting into revenue growth and margins.
Competition is also quite stiff in enterprise SSDs, and SanDisk is a small player with <4% market share, versus Samsung at >35%, SK Hynix/Solidigm at nearly 27%, and Micron and Kioxia in the 14% range. Jefferies analysts also warned that there is “no idea” what market share China’s YMTC could take as it ramps up output.
There’s also the risk of having a limited viewpoint on where normalized earnings could land if/when the cycle peaks and reverses, as SanDisk is currently benefiting from strong pricing and a tightening supply-demand environment. Combined with the sharp 1,000%+ rally since its summer lows and peak valuation multiples, there could be a higher degree of risk if the supply-demand imbalance and pricing revert sooner than expected.
Conclusion
SanDisk has a multi-faceted growth opportunity ahead over the next few quarters, with supply-demand imbalances widening with strong enterprise SSD demand, a potential doubling of prices in the March quarter supporting more upside for revenue, and multiple hyperscaler qualifications on deck.
Nvidia’s CES keynote discussion around the context window becoming the next bottleneck could have positive implications for the SSD market from Nvidia’s ICMS platform utilizing NVMe SSDs to significantly boost KV cache memory and increase throughput for inference applications. HBF is also a potential long-term opportunity later in the decade as it could dramatically boost total GPU memory to allow frontier models to run on a single GPU, though it is too early to tell if it will be viable.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund do not own shares in SNDK at the time of writing and may own stocks pictured in the charts.
Since launching the Discovery tier less than a year ago, The I/O Fund’s internal process for identifying new winners has greatly improved. Based on the research we produced from this tier, we added stocks like Bloom Energy, Core Scientific and Oklo to our portfolio, and these new additions became some of our biggest wins in 2025.
As we continue to build out the Discovery library, we’d like to make it as easy as possible for our readers to follow along. We want to cast a wide net, explore thoroughly, and leave no stone unturned. Therefore, we anticipate our coverage to include dozens of stocks over time. However, we also want to make sure we get down to brass tacks by providing you a clear takeaway by ranking what we’ve dug up every quarter. The ranking will also help clarify which ones we are eyeing an entry for and what setups Knox will cover in his Discovery Tier webinars. This will be called the I/O Fund’s Top 10 New Ideas List.
The ranking we provide is an estimate, which means all 10 stocks are of interest. Given the nature of momentum stocks, the ranking could shift quickly. Please check back to the Discovery Ranking list provided on the Dashboard for any changes in the interim.
Access the updated Discovery Ranking list here: Top 10 Watchlist
Themes:
We have covered the trends below on our Q4 Top 15 AI Stocks Report. Our Discovery list offers names centered in these trends, thus we are repeating some of the information for easy reference, but surfacing several new stocks that stand out this quarter.
AI Networking:
Networking is at the heart of the new architecture that Nvidia is shipping now as the increased bandwidth is instrumental in driving higher performance. For example, the NVL72 systems will deliver 4X faster training and 30X faster inference compared to HGX systems. Notably, this is accomplished with many more GPUs from eight to 72 per system.
To support the new systems, the NVLink domain moves from supporting eight GPUs to 72 GPUs and 36 CPUs with a speed of 1.8 TB/s with 18 NVSwitch ASICs, up from four in the HGX/DGX systems due to increasing the number of GPUs but also due to doubling the per-GPU links. The 5th generation of NVLink supports up to 576 GPUs compared to the fourth generation of up to eight GPUs.
Scale-up refers to increasing the number of GPUs in an AI system. Proprietary NVLink remains the highest performance option for scale-up interconnects, although PCIe and scale-up Ethernet are also used. For the cabling, copper is used for intra-rack scale-up with up to 5,184 cables per system. Future generations of NVLink are likely to integrate optical I/O so that GPUs can communicate across racks without requiring costly retimers.
GB200 NVL72 with 72 GPUs and 36 CPUs has 18 NVSwitch chips and 72 InfiniBand NICs for scale-out networking and 36 Bluefield-3 Ethernet NICs for front-end networking. Compare this to the HGX systems with 8 GPUs and the DGX systems with 8 GPUs and 2 CPUs has 4 NVSwitch chips and 8 InfiniBand NICs.
This means the new architecture that Nvidia is shipping now results in 9X more GPUs, 4.5X more NVSwitches, 9X more InfiniBand NICs and 18X more front-end NICs. Each GPU requires its own InfiniBand link for scale-out whereas NVSwitch components grow faster than GPU count as each GPU must talk to every other GPU, therefore, it has more of an exponential growth.
Although NVLink is proprietary, it acts as a bellwether for the importance of AI networking and lesser-known suppliers. Generally speaking, what we can see from looking more closely at Nvidia’s networking fabric is that networking components are increasing 5X to 9X, and in some cases up to 18X.
Scale-out racks refer to connecting multiple racks across a cluster. InfiniBand switches and Ethernet switches are used for this purpose. As you can see below, Nvidia offers scalable units called SuperPODs that offer tens of thousands of GPUs. For a very large SuperPOD with 16,834 GPUs and 2048 nodes, there would be hundreds of InfiniBand switches required (or a hyperscaler can also use Ethernet switches) and extensive cabling is also required.
Also consider that as the networking and interconnects market matures, there will be new opportunities to participate, for example, co-package optics are expected to be introduced for the Rubin generation of GPUs in 2026-2027.
Regardless of the exact networking-to-compute ratio, as we scale up AI systems, the architecture becomes a networking and interconnect problem that must continually be solved for, and we want to be correctly positioned within the networking supply chain. Therefore, AI networking will remain a key focus for the I/O Fund’s Top 10 New Ideas in the coming quarters and years. You can expect our team to deliver additional names in this trend and to increase our allocation as needed.
AI Energy:
Data center power demand is expected to grow at an accelerated clip through the end of the decade and beyond, with more powerful GPUs and surging growth in inference two main drivers.
McKinsey projects data center capacity will rise ~2.5x to 219 GW by 2030, up from 82 GW in 2025, with AI contributing 70% of that demand. This corresponds to total capacity growth of 137 GW over the next five years, with 112 GW coming from AI.
Goldman Sachs estimated global data center power usage at 55 GW in early 2025, far below BCG’s 82 GW figure. However, GS projects power usage to reach 84 GW in 2027 and increase further to 122 GW by 2030, corresponding to total growth of just 67 GW. However, considering 2025 and 2026 capex spend could support >20 GW of new capacity, this forecast may understate the pace of capacity growth.
As that analysis pointed out, Nvidia’s Blackwell lineup is bringing a significant increase in power consumption, nearly double the H200’s 70 kW at 120 kW for the GB200 NVL72 and 140 kW for the upcoming GB300 racks.
Beyond Blackwell, Nvidia’s future design lineup shows continual increases in power consumption. Its Rubin generation is expected to boost thermal design power (TDP) by 50% over Blackwell at up to 180 kW per rack, with the upgraded Vera Rubin then doubling this to 360 kW per rack by 2027.
In its largest configuration, the Vera Rubin NVL576, dubbed the ‘Kyber’ rack, could draw as much as 600 kW (0.6 MW), or 5x that of the GB200 NVL72 in just a two-year design timeframe. These figures do not include networking, interconnects, cooling and other hardware, which will further boost power draw per rack.
This rapid increase in power consumption per GPU generation is critical, as existing infrastructure is simply unable to meet these escalating power demands. For example, Applied Digital pointed out that nearly 70% of current data centers “contain racks requiring between four and nine kW of power, and less than two percent of data centers have racks with greater than 50kW.”
For comparison, Super Micro’s GB200 NVL72 SuperCluster requires 132kW, while the upcoming Kyber rack could more than quadruple that to 600kW. Because Blackwell-based servers are now 15x to 30x the power density, cooling and power delivery strategies have to be redesigned, as liquid cooling now becomes a necessity.
This sharp rise in power density means current infrastructure may be unable to transition from 4-9kW racks to >130kW racks without incurring significant retrofitting costs, while building new infrastructure bypasses that hurdle and allows for optimization for high-powered racks.
For example, Vantage’s upcoming 1.4 GW campus in Texas for Oracle is designed for ultra-high density racks up to 250kW, yet this will not be enough power to able to host NVL576 racks in just two to three years’ time. Additionally, a former Microsoft Azure AI executive reportedly said he estimated that the “requirements in terms of power for the data center would probably at least double every three years and maybe exponentially so over a period of time,” further reinforcing this.
The push to 600kW racks over the next few years means this is not a transient problem, rather it is one that the industry will continue to face, meaning continuous new construction may be needed to handle surging power demand.
In the latest earnings call with CoreWeave, management agreed with the analysis we have presented to Pro Members, stating: “at the end of the day, right now, it's the powered shells that are the choke point that is causing the struggle to get enough infrastructure online for the demand signals that we are seeing […]”
Takeaway: When building our portfolio, we have to balance many things when it comes to AI energy stocks. Time to power is paramount as some energy infrastructure is 5-10 years out from commencing operations and generating revenue. Secondly, many energy stocks require significant cash to secure energy sources, get government approval and build the underlying infrastructure, which also includes taking into consideration regional differences, transmission capacity and distances from generation sites to metro areas.
AI Data Layer:
The initial years of AI development were compute intensive to where training created a compute hierarchy. We still see evidence of this hierarchy as companies with access to GPUs, networking and energy have an advantage, and the barrier to entry is high in both costs and by the limited supply preventing widespread access.
Nvidia has enjoyed a near-monopoly by being the parallel processing leader decades before AI took over as the primary market that demanded massive compute and data throughput for matrix multiplications and vector math. The lack of supply has afforded the world’s largest companies a head start in training and deploying models while startups and enterprises patiently wait for access. As more frontier LLMs are deployed by R&D labs and Big Tech, the emphasis moving forward will be on inference rather than only on training models.
The inference market is when enterprises and companies sitting on large private data sets will be able to increase the accuracy of open sourced models and licensed models. There will be an important shift to where companies that can offer domain-specific data in various industries, such as finance, healthcare, manufacturing, will do well by optimizing processes to generate more revenue and achieve better margins. It will start with the Fortune 500, the Global Fortune 2000 and well-funded startups. Meanwhile, investors should also not overlook the fact that R&D labs are growing closer to cracking the consumer market, as well, with apps such as OpenAI’s ChatGPT and Sora, or Perplexity’s search.
While training is benefiting those who sell the compute or own the infrastructure (and we will continue to own these stocks), there will also be a shift toward companies that manage the data pipes by sitting across the many database and software services that enterprises use. Think of all the ERP systems, CRMs, legacy databases, etc, where private data is stored. There will be an emphasis toward combining the data, keeping it private, yet utilizing it to increase the quality of inference.
Takeaway: Compute will continue to drive the scale for inference; data will drive the quality of inference. Therefore, a key focus for I/O Fund’s Top 10 New Ideas list over the next few quarters (and years ahead) will be AI data stocks that help private enterprises use their valuable data to feed data-hungry reasoning models. We will also, in tandem with AI networking and AI energy, be looking to build our portfolio with exposure to stocks that will participate in the AI inference market, which spans hardware, software, the data layer, and more.
The stocks below are new ideas and at time of publishing, the I/O Fund does not own these stocks although they are under strong consideration for the portfolio. To find out the stocks the I/O Fund owns, subscribe to our Pro tier for Research or our flagship tier Advanced with additional research, real-time trade alerts, allocations to stocks and weekly webinars.
The difference between Palantir and other AI-enabled database competitors is that Palantir is able to answer questions a model cannot answer. Traditional business intelligence companies require a complete data set whereas Palantir is able to tackle situations where there is not a complete data set. You can think of the competitive advantage as being actionable depth, which Palantir has described as “the reasoning that goes into decision-making, not just data.”
Palantir’s Artificial Intelligence Platform (AIP) integrates generative AI with operational data and workflows, and when combined with Palantir’s other platforms Foundry and Apollo, it provides an AI service mesh that can run hundreds of microservices, scale compute through its Rubix engine and orchestrate updates through Apollo. Additionally, Palantir’s knowledge graph referred to as Ontology is a distinct advantage. The graph offers better context than a large language model would on its own – or as Palantir states, it’s “the reasoning that goes into decision-making.”
In Q3, Palantir delivered one of the most outstanding reports across tech, with revenue accelerating nearly 15 points sequentially to almost 63%, with red-hot growth in key metrics and a 50 point acceleration in US Commercial revenue since the start of the year.
Overall Revenue Growth Overall Revenue Growth
Palantir reported $1.18 billion in revenue in Q3, up an impressive 18% QoQ and beating estimates by 8.4%, driven by unwavering momentum in US Commercial. Commercial revenue rose 21.5% QoQ and 73% YoY to $548 million, a 26 point acceleration from 47% YoY growth last quarter. Government revenue rose 14.5% QoQ and 55% YoY to $633 million, a six point acceleration from 49% YoY in Q2; Government remained Palantir’s largest segment at ~53.6% of revenue.
On a YoY basis, overall revenue growth accelerated 14.8 points to 62.8% YoY, the largest sequential acceleration to date and marking to Palantir’s highest growth rate since going public. Over the last nine quarters, topline growth has accelerated ~50 points, from just 12.7% in Q2 2023, a rare feat to accomplish.
For Q4, Palantir guided for revenue up 60.6% YoY to $1.327 to $1.331 billion, well ahead of estimates for 44.2% growth to $1.19 billion. While this does represent a marginal deceleration at face value, this sequential deceleration is in line with trends from previous quarters.
For the full year, Palantir raised its revenue outlook to $4.396 to $4.400 billion, pointing to YoY growth of 53.5% at midpoint, a sharp acceleration from 29% growth in 2024. To put in perspective the strength of this acceleration, Palantir had initially guided for just 30.9% growth to $3.76 billion in revenue back in Q4 2024; growth is now more than 22 points faster.
AI Segment Growth AI Segment Growth
Palantir’s US Commercial segment is generally seen as the primary vector for its AIP-driven growth, with robust momentum only accelerating further in Q3.
US Commercial revenue grew 29% QoQ and 121% YoY to $397 million in Q3, accelerating from 93% YoY growth in Q2. Since the start of the year, US Commercial growth has accelerated 50 points, and since the start of 2024, growth has accelerated 81 points.
For the full year, Palantir significantly boosted its US Commercial growth outlook to >104% YoY, up from 85% previously. This corresponds to revenue of $1.433 billion, up from $1.302 billion previously.
Key metrics for the segment were very strong: TCV closed (total contract value) surged 342% YoY to a record $1.31 billion, while remaining deal value (RDV) rose 199% YoY and 30% QoQ to $3.63 billion. US Commercial deals closed of >$1 million were up 2X YoY and deals closed of >$5 million were up 5X YoY.
Additionally, other key metrics outside US Commercial were very strong — net retention rate (NRR) expanded six points sequentially to 134%, and over the past two years, NRR has risen an impressive 27 points, with Palantir noting that AIP is continuing to drive existing expansions and new customer conversions. Total remaining deal value rose 91% YoY and 21% QoQ to $8.6 billion, and Palantir also closed its highest ever TCV quarter at $2.8 billion.
Earnings Earnings
Palantir reported $0.18 in GAAP EPS in the quarter, up 200% YoY, while adjusted EPS was $0.21, beating estimates by 25.5% and rising 110% YoY. Palantir did not provide a specific guide for EPS for Q4, though current estimates are pegged at $0.12 in GAAP EPS and $0.22 in adjusted EPS, up 300% YoY and 57% YoY, respectively.
For FY25, Palantir is expected to earn $0.72 in adjusted EPS, up nearly 76% YoY, before slowing to 39% growth to $1.01 in FY26.
Margins Margins
Margins strengthened considerably in the quarter, with adjusted operating margin surpassing 50% with more expansion guided for Q4. Palantir’s Rule of 40 score (revenue growth + adj operating margin) expanded to a wild 114%, up from 94% last quarter and 68% last Q3.
Gross margin was 82% in Q3, up one point QoQ and two points YoY, while adjusted gross margin was 84%, up two points YoY and QoQ.
GAAP operating margin was 33%, an impressive 6 point QoQ and 17 point YoY expansion. Adjusted operating margin was 51%, breaking past 50% for the first time and up 5 points QoQ and 13 points YoY. For Q4, Palantir guided for adjusted operating margin to be 52%, showcasing its ability to drive strong margin expansion alongside swift revenue acceleration. Full year adjusted operating margin guidance was raised from 46% to 49%.
GAAP net margin was 40%, up 7 points QoQ and 20 points YoY. Adjusted net margin was 45%, up 5 points QoQ and 12 points YoY. Palantir is one of the few, if not only, tech companies to have 40% GAAP net margins with revenue growth accelerating to above 60%.
Cash Cash
Cash flows were strong, though cash flow margins dipped on a YoY and QoQ basis. Operating cash flow was $507.7 million for a 43% margin, shrinking from a 54% margin in Q2 and 58% in the year ago quarter.
Adjusted free cash flow was $539.9 million for a 46% margin, down from 57% in Q2 and 60% in the year ago quarter. Palantir raised its adjusted FCF guidance for the year to $1.9 to $2.1 billion, or a 45.5% margin, up from a 42.8% margin previously.
Cash and equivalents totaled $6.4 billion and debt remained zero.
Valuation Valuation
Valuation is the crux for Palantir as the stock trades at 100x forward revenue, nearly triple its five-year average of 36x and in rather uncharted territory for software stocks. On the bottom line, Palantir trades at 256x forward adjusted EPS despite a >40% margin, more than double its average 109x multiple.
Notable Risks Notable Risks
The valuation with Palantir is a gamble as the company is attempting to set a new bar for AI software, with >100x forward sales multiples only achieved for short fashion in 2021 for a handful of prior market darlings, whose stocks have yet to return to those prices. This elevated valuation may also present a risk if/when the company reaches peak revenue growth as comps will quickly get tougher.
Celestica: 800G Switch Demand Accelerating into FY26 with 1.6T Switches Soon Ramping
Celestica is an under-the-radar beneficiary of the AI networking trend, capitalizing on strong demand for 800G and 1.6T Ethernet networking switches and leveraging its deep ties to hyperscalers as an original design manufacturer (ODM).
Celestica guided for one of the most impressive accelerations seen in this last quarter of earnings, underpinned by its 800G switches accelerating next year with 1.6T ramps on deck. For 2026, Celestica expects revenue growth to accelerate around five points to 31% YoY in 2026, whereas consensus had been pegged at just 17% YoY. This strong upside is being driven by networking and custom AI compute platforms with visibility into 2026-2027.
In terms of AI revenue, Celestica’s Cloud and Connectivity Solutions (CCS) segment is guided to generate $9 billion in revenue in 2025, up ~40% YoY, accounting for nearly 74% of total revenue. CCS, which includes AI networking, server, storage and rack-scale system solutions, is Celestica’s main growth driver and is also expected to grow ~40% annually in 2026 and 2027.
Celestica is closely linked to Broadcom’s networking platforms as a key vendor, serving major customers such as Google and Meta, with some of its notable product engagements including Google’s TPU server racks, and Meta’s Minerva ASICs servers, Wedge400 switches and also its next-gen Tomahawk5-based 400G AI fabric switch Minipack3. Additionally, management’s commentary suggests that OpenAI could become a key customer as soon as 2027 – read more on this in our Discovery deep dive. Discovery deep dive.
Revenue: Revenue:
Celestica reported Q3 revenue of $3.19 billion, up 28% YoY and 10.4% QoQ, coming in more than 6% ahead of estimates for $3.00 billion. Revenue from the CCS segment rose 43% YoY to $2.41 billion, driven by an 82% YoY increase in Communications revenue to $1.94 billion offsetting a (24%) decline in Enterprise to $477 million on an AI program transition with a hyperscaler. Celestica’s other segment, Advanced Technology Solutions (ATS, serving aerospace, defense, industrial and semicap equipment markets) saw revenue decline (4%) YoY to $781 million.
For Q4, Celestica guided for revenue of $3.325 billion to $3.575 billion, up 35% YoY at midpoint of $3.45 billion, a seven point acceleration. On a QoQ basis, this correlates to 8% growth sequentially.
As a result, Celestica raised its fiscal 2025 guidance from $11.55 billion to $12.2 billion, a strong 7% raise with just one quarter to go, signaling the strength of demand the company is witnessing in Q3 and Q4. The updated guide points to 26% YoY growth.
Perhaps more important was Celestica’s initial guidance for fiscal 2026, with the company laying out an initial forecast for $16 billion, for 31% YoY growth, a five point acceleration. This was $1.86 billion ahead of estimates for $14.14 billion, a large 13.2% beat.
The impressive fiscal 2026 guide and revisions to consensus estimates for fiscal 2027 have taken Celestica’s forward growth CAGR from FY25 to FY27 to a strong 28%, up from 17% prior to the report.
AI Segment Growth: AI Segment Growth:
In terms of AI revenue, Celestica’s Cloud and Connectivity Solutions (CCS) segment is forecast to generate $9 billion in revenue in 2025, up ~40% YoY, and accounting for nearly 74% of total revenue. CCS, which includes AI networking, server, storage and rack-scale system solutions, is Celestica’s main growth driver and is expected to grow ~40% annually in 2026 and 2027.
In Q3, revenue from CCS segment rose 43% YoY to $2.41 billion, driven by an 82% YoY increase in Communications revenue to $1.94 billion offsetting a (24%) decline in Enterprise to $477 million on an AI program transition with a hyperscaler. Communications revenue notably accelerated to 18% QoQ from 15% QoQ in Q2, on strong demand for 800G switch products and solid demand for optical products.
For Q4, CCS revenue is implied to accelerate nine points to 52% YoY, with Communications growth guided in the high-60s YoY on strong switch demand, and Enterprise guided in the low-20s as the new AI program is set to ramp. Despite the seemingly strong guide in Communications, QoQ growth would be just 1% QoQ, a sharp deceleration from Q3’s 18% QoQ growth.
For 2026, Celestica guided for approximately 40% YoY growth in CCS to ~$12.6 billion, up from $9 billion guided for 2025, supported by views for accelerating 800G demand, early 1.6T ramps and the ramp of its next-gen AI compute platform to full-volume. Management also hinted that they have visibility and confidence in maintaining at least 40% growth for CCS in 2027 – more on this in our deep dive.
Earnings Earnings
Celestica reported GAAP EPS of $2.31 in Q3, beating the $1.38 estimate by 67.6%. Adjusted EPS was $1.58, beating the $1.49 estimate by just 6% and representing growth of 52% YoY.
For Q4, Celestica guided adjusted EPS to be in the range of $1.65 to $1.81, which, at the $1.73 midpoint, is only marginally ahead of estimates for $1.71. This also corresponds to a slight acceleration to 56% growth.
For fiscal 2025, Celestica boosted its adjusted EPS outlook by 7.3% to $5.90, from its previous forecast for $5.50 and pointing to 51% YoY growth. For fiscal 2026, Celestica outlined an initial guide for $8.20 in adjusted EPS, up 39% YoY and well ahead of estimates for $7.22.
Margins Margins
Margins continued to expand in Q3, with some signs of operating leverage arising from strong Communications growth as operating margin expanded by 4.7 points YoY versus a 2.6 point YoY expansion for gross margin.
GAAP gross margin was 13.0% in Q3, up 0.2 points QoQ and 2.6 points YoY.
GAAP operating margin of 10.2%, up 0.8 points QoQ and 4.7 points YoY. Adjusted operating margin was 7.6%, up 0.2 points QoQ and 0.8 points YoY.
GAAP net margin of 8.4%, up 1.1 points QoQ and 4.8 points YoY. However, adjusted net margin was just 5.7%, up just 0.1 points QoQ and 0.7 points YoY due to a $113 million impact from gains on total return swaps.
For fiscal 2025, Celestica guided for adjusted operating margin to be 7.4%, and for 2026, only a slight increase to 7.8% despite the 31% growth on the top-line. This suggests that its positioning primarily as an ODM may limit future upside to operating margins.
Cash Cash
On the other hand, cash flows are rather thin and fell to the lowest level in a year.
Operating cash flow was $126.2 million, or a 4% margin, down from 5.3% in Q2 and 4.9% in the year ago quarter. OCF growth was just 2.4% YoY and was also the lowest cash flow since the year ago quarter.
Adjusted FCF was $89 million for a 2.8% margin in Q3, up 15.6% YoY but also the lowest level since the year ago quarter. Adjusted FCF margin was down from 4.1% in Q2 and 3% in the year ago quarter.
For fiscal 2025, Celestica raised its adjusted FCF guidance slightly to $425 million, from $400 million prior, for a 3.5% margin, while capex is guided to $200 million, or 1.6% of revenue. Fiscal 2026 adjusted FCF was guided at $500 million, up 18% YoY and for a 3.1% margin, with the margin decline driven by higher capex, guided to rise 50-100% YoY to $300 to $400 million, or 2.2-2.5% of revenue.
Cash and equivalents totaled $305 million while debt totaled $728 million in term loans. Including an undrawn revolver, total liquidity is approximately $1.1 billion. Celestica’s gross debt to TTM adjusted EBITDA was 0.8x, improving by 0.1 points sequentially and 0.3 points from last year.
Valuation Valuation
Celestica is trading close to peak multiples on the top and bottom line. Forward PS is currently at 2.8x, well above the five-year average of 0.75x and 40% above the 2x multiple it commanded at the start of September. Even on the fiscal 2026 guide, shares are at an elevated 2.1x multiple, just below peaks at 2.5x.
On a forward PE basis, shares are trading at 51x fiscal 2025 adjusted EPS and 41.3x fiscal 2026, well above the five-year average of 15.4x and prior resistance at 25x in late 2024 and early 2025. This is slightly below current peaks around the 60x level from October and November.
Notable Risks Notable Risks
Valuation is the primary risk, and while it could be argued that the company is deserving of a material re-rating higher on strong AI-driven growth and a shift to higher-margin, custom rack solutions come 2027, margins remain thin with operating margin only just crossing into double-digit territory. Additionally, its ODM positioning also presents a risk as even a shift to higher complexity, higher value products may be unable to produce continuous margin expansion into the low-teens.
Celestica’s Communications growth within CCS also presents another key risk for Q4, as the high-60s YoY growth guide would imply QoQ growth of around 1% next quarter. This would mark Communications’ lowest sequential growth in the last two years, and its first time reporting single-digit sequential growth in the last seven quarters, raising a potential red flag considering Communications is primarily driven by networking/800G switches. However, a likely explanation of this could be the strong outperformance in Communications in Q3 – guidance was for low-60s YoY growth, which Celestica beat by ~20 points. As a result, QoQ growth was likely expected to be ~4%, but came in at 18%, possibly representing a much stronger-than-expected ramp of 800G platforms in the quarter.
Arm: Data Center Royalties Double YoY, Targeting 50% Data Center CPU Share
Thematic: 9/10
Fundamentals: 9/10
Valuation: 3/10
Brief Overview Brief Overview
AI’s need for high-performance, energy-efficient chips creates a long-term tailwind for Arm, as the company’s heterogenous CPU architectures are seeing rapid adoption in data center applications. This is coming from both next-gen merchant GPU platforms and custom silicon deployments from hyperscalers, with Arm now forecasting its server CPU share to reach 50% in 2025, up from 15% in 2024.
The company’s license and royalty revenue model had centered around its v9 architecture, as it commanded double the royalty of v8, featuring in “virtually all high-end data center chips” and commanded a majority share in smartphones. For example, Arm’s Neoverse V2 (based on v9) powers Nvidia’s Grace CPU on its Grace Hopper and Grace Blackwell platforms, along with Amazon’s Graviton4 CPUs, Google’s Axion CPUs, and more.
Arm is now pushing ahead with its Compute Subsystems (CSS) platform to help accelerate time to market for complex chip designs, such as Microsoft’s newest Azure Cobalt 200 CPU rolling out through 2026. CSS notably carries double the royalty rate as v9, which management placed at roughly 10%. Arm also signed three new CSS licenses this quarter, bringing its total CSS licenses up to 19 across 11 companies; five of these companies are already shipping CSS-based chips.
Revenue Revenue
Arm’s revenue growth accelerated more than 22 points from 12.1% YoY in fiscal Q1 to 34.5% YoY in fiscal Q2 to $1.13 billion, while QoQ growth rebounded from (15.1%) to 7.8% QoQ.
Royalty revenue increased 21% to a record $620 million, driven by growth in smartphones, auto and data center, along with increased v9 penetration and the ramp of CSS platforms. However, this was a slight deceleration from 25% growth in the prior quarter.
Licensing revenue rose 56% YoY to $515 million on normal timing fluctuations, accelerating from (1%) growth in the prior quarter.
For Q3, Arm guided for revenue of $1.225 billion at midpoint, though this represents a deceleration to 24.6% YoY and 7.9% QoQ growth.
AI Segment Growth AI Segment Growth
Arm does not provide specifics into its data center revenue contributions, but noted that data center royalties doubled YoY on continued deployment of Arm-based chips at hyperscalers. Data center Neoverse royalties more than doubled YoY, and Arm expects to reach 50% share in of CPUs deployed by hyperscalers by the end of 2025.
For another view, Arm’s management explained that it is reasonable to assume cloud and networking would reach 15% to 20% share of royalty revenue for the fiscal year, up from ~10% last year. Assuming Q3 and Q4 see royalty revenue rise ~20% YoY, this could project cloud and networking’s contribution between $394 to $525 million.
Earnings Earnings
Arm delivered strong GAAP EPS growth in Q1 as margins expanded down the line, while adjusted EPS growth was more muted but solid nonetheless.
GAAP EPS was $0.22 in Q1, up 120% YoY and more than 66% ahead of estimates for $0.13. Adjusted EPS was $0.39, nearly 18% ahead of estimates for $0.33 and representing growth of 30% YoY.
For Q3, Arm guided for adjusted EPS to be $0.41, +/- $0.04, for YoY growth of just 5%. Q4 is estimated to see growth of just 2.7% YoY to $0.56, before reaccelerating to >29% YoY growth in each quarter of fiscal 2027.
Margins Margins
Arm saw strong margin expansion down the line, with operating and net margin expanding at a much larger degree than gross margin in Q2, signaling that adoption of its higher margin v9 and CSS platforms is translating to bottom line strength.
GAAP gross margin was 97.4% in Q2, up 1.2 points YoY and 0.2 points QoQ.
GAAP operating margin was 14.4%, up 6.8 points YoY and 3.6 points QoQ. Adjusted operating margin was 41.1%, up 2.5 points YoY and 2 points QoQ; for Q3, adjusted operating margin is implied to be ~39.4% at midpoint assuming gross margin is flat QoQ.
GAAP net margin was 21%, up 8.3 points YoY and 8.7 points QoQ.
Cash Cash
Cash flows improved substantially on a YoY basis, and Arm’s balance sheet remains robust and debt-free.
Operating cash flow was $567 million for a 50% margin, up from a 0.7% margin in the year ago quarter and a 31.5% margin in the prior quarter.
Adjusted free cash flow was $411 million for a 36.2% margin, a significant increase from (7.7%) in the year ago quarter and 14.2% in the prior quarter.
Cash and equivalents totaled $3.26 billion and debt was zero.
Valuation Valuation
Unlike many of the other names on this list, Arm is trading more than 20% below its average multiples on the topline, though the company still commands a premium multiple to many of its semiconductor customers. Arm trades at 24.3x forward PS, below its 31.6x average since IPO and well below its peaks at 50x, though it has traded as low as 16x.
On the bottom line, Arm trades at 64.5x forward PE, below its average of 79.5x, though shares have traded as low as 40x and as high as 125x.
Notable Risks Notable Risks
Despite increasing royalty rates by 2X with each new architecture, from 2.5% under v8 to 5% with v9 and 10% with CSS, Arm’s growth may continue to lag that of peers – the company may need to take the leap into design as IP licensing did not move the needle enough from the mobile era.
Fabrinet: Key Nvidia Partner with Revenue Accelerating to 30% YoY
Thematic: 8/10
Fundamentals: 4/10
Valuation: 4/10
Brief Overview Brief Overview
Fabrinet provides advanced optical packaging, high-precision optical and electro-optical manufacturing services to OEMs, with revenue primarily stemming from transceivers, active optical cables, optical subsystems for high-speed networking, and data center interconnect.
Fabrinet is also a key optical partner for Nvidia, with its contributions said to be for short-reach active optical cables with 800G transceivers for Nvidia’s InfiniBand platforms, and optical engine packaging. Fabrinet was also stated as key partner for Nvidia’s upcoming silicon photonics CPO switch platforms during GTC 2025.
Nvidia accounted for 27.6% of Fabrinet’s revenue in fiscal 2025, or ~$943.7 million; however, this was down nearly (7%) YoY, potentially due to the Blackwell delays from early 2025. Cisco is Fabrinet’s second largest customer, said to be for optical transceivers, accounting for 18.2% of revenue in fiscal 2025, or $622.3 million, up 61% YoY.
However, like peers positioned similarly in the contract manufacturing space, margins are quite thin and did not show any expansion in the past quarter, and Fabrinet may be unable to produce substantial margin expansion moving forward.
Revenue Revenue
Fabrinet reported record fiscal Q1 revenue of $978.1 million, up 7.5% QoQ and 21.6% YoY. This marked a slight acceleration from 20.7% YoY and 4.3% QoQ growth in FQ4.
Fabrinet’s Optical Communications (OC) business remained the key driver, accounting for 76.4% of revenue. OC revenue rose 8.4% QoQ and 19.3% YoY to $747 million in Q1, accelerating 15.5% YoY and 4.8% QoQ in Q4. Non-optical Communications revenue, or auto, industrial laser and other end markets, was $231.2 million, up 4.6% QoQ and 30% YoY.
For fiscal Q2, Fabrinet guided for revenue of $1.05 to $1.10 billion, accelerating more than seven points to 29% YoY at midpoint, while QoQ growth would accelerate 2.4 points to 9.9%. Revenue is expected to accelerate to 30-31% YoY in both Q3 and Q4 to $1.13 billion and $1.19 billion respectively.
For fiscal 2026, Fabrinet has not provided a set guide, though analysts expect revenue to rise 28% YoY to $4.39 billion, accelerating from 18.6% in fiscal 2025.
AI Segment Growth AI Segment Growth
Fabrinet offers a few distinct breakdowns within Optical Communications revenue, highlighting datacom and telecom, and within telecom, data center interconnect (DCI).
Datacom revenue was $273.1 million in Q1, down (17%) YoY and (1%) QoQ, which was a smaller decline than expected due to a smaller QoQ decline at Fabrinet’s largest datacom customer and larger contributions from other datacom customers.
DCI revenue was $138.1 million, up 92.3% YoY and 29% QoQ, marking a sharp acceleration from 44.8% YoY and 3.5% QoQ growth in the prior quarter. DCI now accounts for ~14% of total revenue.
Fabrinet also introduced HPC as a reportable segment under Non-optical Communications in Q1, as the company qualified and started to ramp its first HPC program for AWS, which contributed $15.4 million in revenue. Management believes HPC “will scale considerably over the coming quarters and become a significant driver to our overall growth.”
Fabrinet did not provide a breakdown on OC revenue by data rate in Q1, though in Q4 (the Aug quarter), revenue from >800G data rates was $312.7 million, up 20.7% YoY and 32.5% QoQ and accounting for 34.4% of total revenue.
Earnings Earnings
Despite a lack of margin expansion, Fabrinet has a rather defensible bottom line and delivered solid EPS growth in Q1. GAAP EPS was $2.66, up 24.9% YoY, while adjusted EPS was a record $2.92, up 22.2% YoY and beating estimates by 3.5%.
For Q2, Fabrinet guided for GAAP EPS of $2.91 to $3.06, up 25.4% YoY at midpoint, while adjusted EPS was guided to be $3.15 to $3.30, up 23.6% at midpoint. The growth rates here lagging revenue by four to five points suggests that margins may feel a bit of pressure next quarter.
Adjusted EPS growth is forecast to accelerate into the end of the fiscal year, however, suggesting margins may rebound quickly if there is softness in Q2. Q3 and Q4 are both forecast to see adjusted EPS growth of 36% to 37% YoY, or six to seven points faster than revenue growth.
For fiscal 2026, Fabrinet is currently expected to generate GAAP EPS of $12.25, up 33.6% YoY, and adjusted EPS of $13.25, up 30.3% YoY.
Margins Margins
As noted above, Fabrinet did not deliver much on the margin front in Q1, with gross margin contracting slightly, and operating margin remaining flat YoY but dipping slightly sequentially. Additionally, its contract manufacturing position may not be able to drive meaningful margin upside over the coming quarters.
Q1 GAAP gross margin was 11.9%, down 0.4 points YoY and 0.3 points QoQ. Adjusted gross margin was 12.3%, down 0.4 points YoY and 0.2 points QoQ.
GAAP operating margin was 9.6%, flat YoY but down 0.2 points QoQ. Adjusted operating margin was 10.6%, down 0.1 points YoY and QoQ,
GAAP net margin was 9.8%, up 0.2 points YoY and QoQ. Adjusted net margin was 10.8%, flat YoY but up 0.3 points QoQ.
Cash Cash
Fabrinet’s cash flow generation was solid in Q1, with operating cash flow margin expanding and free cash flow rebounding after a soft Q4. Fabrinet also has a healthy balance sheet with nearly $1 billion in cash and no debt.
Operating cash flow was $102.6 million in Q1 for a 10.5% margin, up from a 10.3% margin in the year ago quarter and a 6.1% margin in Q4.
Free cash flow was $57.3 million for a 5.9% margin, down from a 7.8% margin in the year ago quarter, but up from 0.5% in Q4. The YoY contraction in FCF margin was primarily due to capex, which was $45.3 million, up 123.5% YoY.
Cash and equivalents totaled $968.8 million and debt was zero.
Valuation Valuation
Fabrinet is trading close to peak multiples on the top and bottom line. On a forward PS basis, Fabrinet trades at a 3.8x multiple, below its 4.3x peak from early December and notably well above its five-year average multiple of 2.5x; to note, Fabrinet has spent much of the last two and a half years trading between 2x to 3x, with the re-rating above 3x taking place since October 2025.
On the bottom line, Fabrinet trades at a 35x forward PE, slightly below its peak of 40x and again well above its five-year average of 23x. Similar to forward PS, Fabrinet has spent the last two and a half years primarily between 18x and 28x forward PE.
Notable Risks Notable Risks
The main risks to Fabrinet relate to its positioning as a contract manufacturer for Nvidia, Cisco and other key clients, as the company may be unable to produce substantial margin expansion beyond its current profile, meaning EPS growth will likely be closely correlated with revenue growth rates. Fabrinet is also quite highly exposed to Israel, which accounted for 29% of revenue in fiscal 2025, and continuation of global conflicts could potentially impact growth.
Fabrinet’s strategic positioning with Nvidia as its key customer may present a risk – while it could see tailwinds from the ramp of the silicon photonics CPO switches in 2026, any delay in Rubin’s ramp could adversely affect shipments and revenue. Analysts have also raised concerns in the past that Fabrinet could face headwinds if hyperscalers ‘unbundle’ from Nvidia’s ecosystem and shift away from Nvidia’s InfiniBand or Ethernet switch options.
Vertiv will not win any hypergrowth stock awards, especially as management has previously offered CAGR guidance of 15% to 17% through 2029. Rather, it’s where Vertiv is positioned as an AI infrastructure partner especially as the trend turns toward modular infrastructure that makes this a stock to watch. Essentially, all roads point toward Vertiv’s power and thermal solutions becoming increasingly important for future generations of rack scale solutions, with the company already preparing 800V DC solutions for Nvidia’s Rubin Ultra platform due in 2027.
Revenue Revenue
Vertiv reported revenue up 29% YoY and 1% QoQ to $2.676 billion, well ahead of its original guidance for 23% growth in the third quarter. This was driven by 43% YoY growth in the Americas on accelerated AI demand and 20% growth in APAC.
For Q4, Vertiv guided for revenue to be $2.81 billion to $2.89 billion, up 6.5% QoQ and 18-22% YoY at the $2.85 billion midpoint. While this was ahead of previous guidance for $2.735 to $2.815 billion, this would still represent a nine point deceleration on the topline at midpoint. Management expects Americas revenue to be up high-30s, APAC up mid-single digits and EMEA down high single digits but up mid-teens QoQ.
The strong outperformance in Q3 also led to Vertiv hiking its FY25 revenue guidance from $10 billion at midpoint to $10.2 billion at midpoint, pointing to organic growth of 26-28% YoY. Management did not provide any direct insight into FY26, though they did say that based on the “substantial backlog and clear visibility of pipeline, we anticipate continued significant organic sales growth in 2026,” with EMEA potentially reaccelerating in 2H 2026.
AI Revenue Metrics AI Revenue Metrics
Vertiv’s backlog rose ~30% YoY and 12% QoQ to $9.5 billion, reaccelerating from 21% YoY growth last quarter. More importantly, the $1 billion sequential increase in backlog was the largest in more than two years. However, one of the stronger metrics was order growth, with Vertiv reporting organic orders up 60% YoY and 20% QoQ in Q3. This drove a ten point rebound in TTM organic order growth to 21% YoY, from 11% in Q2.
However, starting in Q4, Vertiv will no longer report on quarterly orders and backlog information, and instead will report a new metric “projected full year orders.”
The following was stated in Q2: “Beginning on our Q4 and full year 2025 earnings call, we will provide projected full year orders rather than quarterly orders and backlog information. We believe this better aligns with how we run our business. We will provide updates on the full year projections quarterly as we progress through the year and as we deem necessary.” This could create a boost to Vertiv’s stock to remove the lumpiness from quarterly reports and to also be more forward looking in terms of visibility offered to investors.
Earnings Earnings
Vertiv reported adjusted EPS up 63% YoY to $1.25 in the quarter, beating the $0.99 estimate by 25%. GAAP EPS of $1.02 beat estimates by 16.7%. For Q4, adjusted EPS was guided to decelerate to 27% growth to $1.26 at midpoint.
For the full year, Vertiv raised its adjusted EPS forecast to $4.07 to $4.13, up from its prior view for $3.75 to $3.85. At midpoint, this represented a nearly 8% hike, now pointing to 44% YoY growth versus 33% previously.
Margins Margins
Vertiv reported expanding margins across the board in Q3, though Q4 is expected to be approximately flat for adjusted operating margin.
Gross margin was 37.8%, up 1.3 points YoY and 3.8 points QoQ.
GAAP operating margin was 19.3%, up 1.4 points YoY and 2.5 points QoQ. Adjusted operating margin was 22.3%, up 2.2 points YoY and 3.8 points QoQ, driven by tariff mitigation efforts and strong execution addressing operational inefficiencies.
Net margin was 14.9%, up 6.4 points YoY and 2.6 points QoQ.
For Q4, adjusted operating margin was guided to be up 0.9 points YoY and approximately flat QoQ at 22.4%, as “progress addressing operational inefficiencies [is] offset by acceleration in growth investments and negative impact from new tariffs.” This is a rather steep decrease from Q2’s guidance for 23.6%, which would’ve been its best adjusted operating margin print since going public in 2020.
For FY25, Vertiv slightly raised its adjusted operating margin forecast by 0.2 points at midpoint to 20.2%, representing YoY expansion of 0.8 points. This is strong as it comes in the face of “significant headwinds from tariffs and operational inefficiencies driven by supply chain actions to mitigate tariffs.” Tariff impacts are expected to be materially offset exiting Q1 ’26.
Cash Cash
Vertiv reported strong cash flows in Q3, with operating cash flow of $508.7 million, up nearly 36% YoY. OCF margin was 19%, up 1.8 points YoY and 6.8 points QoQ.
Q3 adjusted free cash flow was $462 million, up 32% YoY. Adjusted FCF margin was 17.3%, up 1.1 points YoY and 6.8 points QoQ. Q4 adjusted FCF was guided to be $496 million for a 17.4% margin, up marginally from Q3. Vertiv boosted its adjusted FCF guidance by $100 million, now forecasting $1.5 billion for the year, up from $1.4 billion previously. This corresponds to a 14.7% margin.
Accounts receivable dipped (1%) QoQ to $2.81 billion, while inventories rose less than 2% QoQ to $1.43 billion.
Cash, equivalents and investments totaled $1.94 billion, while debt totaled $2.90 billion.
Valuation Valuation
Vertiv is trading around 15% below peak multiples on the top line, and more than 23% below peak on the bottom line. Vertiv’s forward PS is 6.2x, below its recent peak valuation at 7.2x at the end of October and substantially higher than its April low at 2.2x forward PS.
On the bottom line, Vertiv is trading at 40x forward earnings, slightly below its late October peak at 47x and more than 23% below its late 2024 peaks at 52x.
Notable Risks Notable Risks
Vertiv’s extended valuation is a primary risk as the company contends with a sharper deceleration on the top line heading into Q4, as well as a sharp deceleration in EPS growth from 63% in Q3 to 27% in Q4. Margins are also a line item to watch, considering management had guided for a Q4 adjusted operating margin of 23.6% back in Q2 but then subsequently cut that guide to 22.4% in Q3.
Talen: Q3 Revenue Up 29% QoQ, Operating Margin Strengthens
Thematic: 9/10
Fundamentals: 6/10
Valuation: 2/10
Brief Overview: Brief Overview:
Talen is an independent power producer with more than 10GW of generation capacity with 2.2GW of that being nuclear. The company’s assets are primarily located in Pennsylvania, Maryland and now Ohio, yet data center regions and capacity are growing including a long-term power purchase agreement with Amazon to fuel data centers in Pennsylvania.
Talen is expanding its power production portfolio with recent acquisitions of two combined-cycle gas turbine (CCGT) plants, Freedom Energy Center and Guernsey Power Station, for ~$3.8 billion. The two plants will add 2.8 GW to Talen’s energy assets in the PJM region – both are suitable for hyperscale data center power supply. This comes at a time when data center construction is surging in PJM’s region as its grid faces increasing strain, meaning the plants could be more valuable for meeting near-term hyperscaler power needs.
Revenue Growth Revenue Growth
Talen reported revenue of $812 million in Q3, up 28.9% QoQ and 24.9% YoY, with the quarter including the higher 2025/26 capacity pricing of ~$270/MW-day. Revenue from contracts with customers rose 51.9% YoY to $697 million.
Looking ahead, revenue is expected to accelerate to 43.2% YoY in Q4 and 170.2% in Q1, though these estimates have been revised lower from 53.4% YoY and 184.5% YoY at the start of November. Talen said that for Q4, “things are a bit better given the market move-up, but we are still projecting to be at the lower end of our guidance range as we previously stated at that September Investor Day.”
For 2026, Talen said it is “forwards tick up. Gas is up, sparks are expanding and load continues to be strong, all factors that continue to impact commercial positioning on long-term transactions.” However, 2026’s estimated capacity revenue is now $733 million, down $14 million from an estimated $747 million as of Q2.
AI Revenue AI Revenue
Talen’s deal with Amazon is contributing minimally so far, with just $0.70 in adjusted FCF per share expected in FY25. By mid-2026, Talen expects to deliver 240MW, and expects adjusted FCF to rise ~121% to $1.55 per share, before rising at a ~27% annually to $2.50 per share by FY28 as capacity scales to 480MW.
Margins Margins
Gross margin (operating revenue minus energy expense) improved to 64% in Q3 from 60% in Q2 and 62.3% in the year ago quarter.
Operating margin was 32.3% in Q3, a strong improvement from 10.5% in Q2 and 24.3% in the year ago quarter. This was likely driven by the $116 million increase in capacity revenue and, to a lesser extent, the $99 million increase in energy revenue, as operating expenses were up only marginally YoY.
Net margin was 25.5% in Q3, improving from 11.4% in Q2 but down marginally from 25.8% in the year ago quarter.
Earnings Earnings
Talen reported $4.25 in GAAP EPS in the quarter, up 34.5% YoY. Talen did not provide guidance for Q4, yet current estimates point to a (25.9%) QoQ decline to $3.15 in GAAP EPS. For 2025, Talen is expected to report GAAP EPS of $5.13, before rising nearly 290% to $20.00 in 2026.
Talen also reported adjusted EBITDA of $363 million in Q3, representing a 44.7% margin, and improving substantially from $90 million for a 14.3% margin in Q2. Despite the strong increase, Talen narrowed its adjusted EBITDA guidance for fiscal 2025 to $975 million to $1,000 million, down from $975 million to $1,125 million previously. Management said this was because they are towards the lower end of guidance, due to a lack of price volatility in Q3 and the Susquehanna outage.
For 2026, Talen reaffirmed its adjusted EBITDA guidance for $1,750 million to $2,050 million, representing 75% to 110% YoY growth.
Cash and Balance Sheet Cash and Balance Sheet
Talen’s cash flow generation improved in Q3, with adjusted FCF margin rising to the high-20% level. Talen also improved its balance sheet, though cash to debt remains upside-down for the moment.
Operating cash flow was $359 million for a 44.2% margin, a sharp improvement from a (22.9%) margin in Q2 and 14.8% in the year ago quarter.
Adjusted free cash flow was $223 million for a 27.5% margin, up from (12.4%) in Q2 and 14.9% in the year ago quarter. Talen said Q3’s adjusted FCF also included higher capex related to the extended Susquehanna refueling outage. For fiscal 2025, adjusted FCF is tracking near the middle of Talen’s guided range of $470 million to $490 million (narrowed from $450 million to $540 million). For fiscal 2026, Talen is guiding adjusted free cash flow in the range of $980 million to $1.18 billion, more than doubling YoY.
Cash and equivalents totaled $497 million, a solid improvement from $135 million in Q2. Long-term debt totaled $2.99 billion as of quarter end, though this does not include recent debt associated with Talen’s acquisition financing. Management also shared that they expect the balance sheet to get stronger over time as the AWS deal ramps, and as more contracts get added, it is expected to “further strengthen the balance sheet and provide for visibility to those cash flows.”
Management provided an update on total liquidity including revolvers:
“Our liquidity remains substantial with $1.2 billion of liquidity available for working capital, including approximately $490 million of cash available. Once we close on the Freedom and Guernsey acquisitions, we'll have $200 million more of liquidity as our revolver capacity will increase to $900 million. Excluding the acquisition financing, our leverage ratio is still within our 3.5x net debt to adjusted EBITDA target.” Talen’s current year-end forecast for net leverage ratio is 2.6x, well within its target, though it stands at ~5.7x based on current pro-forma debt (incl acquisition financing). Talen is focusing on debt paydown post-closing to reach its 3.5x target by year-end 2026.
Valuation Valuation
Talen’s valuation is trading close to peak levels, though the company has seen substantial multiple expansion over the past two years. Talen is currently valued at 7.2x forward sales, below its peak at 8.5x but well above its 3.4x multiple from March 2025 and its 1.9x valuation from early 2024.
On the bottom line, Talen trades at 74x forward EPS for FY25, more than double its 34x average, though the strong earnings growth in FY26 to the $20 range brings this down to 19x for next year, well below 2024’s peaks of 38x and in line with its five-year average.
Notable Risks Notable Risks
Though Talen looks to prioritize rapid debt deleveraging following the closing of the Freedom and Guernsey acquisitions, debt to cash remains stretched very thin. The valuation does appear stretched when looking at FY25 multiples, though Talen is expected to grow into its multiples on the bottom line quickly next year.
SanDisk: Shares Up 559% In 2025 On NAND Flash, Enterprise SSD Tailwinds
Thematic: 9/10
Fundamentals: 6/10
Valuation: 3/10
Brief Overview Brief Overview
NAND flash-based data center (enterprise) solid state drives (SSDs) are an often overlooked but equally important memory component when it comes to AI training and inference. This is because data center SSDs offer higher read-write speeds critical for accessing and transferring data rapidly, along with higher performance and energy efficiency, making them vital for larger-scale AI training and inference workloads.
NVMe (Non-Volatile Memory Express) is a protocol designed specifically for NAND-flash based SSDs that optimizes performance by reducing latency and increasing data transfer speeds by utilizing the PCIe bus. This helps provide the high throughput and fast data transfer speeds necessary for AI workloads – NVMe SSDs can increase performance by more than 2X versus SATA SSDs.
Now operating independently after being spun out of Western Digital in February 2025, SanDisk is eyeing growth in the enterprise SSD market with its ‘Stargate’ NVMe SSD products, based on its BiCS8 3D NAND architecture jointly-developed with Kioxia, offering industry-leading capacity, energy efficiency and performance. SanDisk’s Stargate line debuted this year with 64TB and 128TB capacity, but will scale to 512TB by 2027, suitable for managing massive AI datasets and workloads. Data center remains a smaller portion of SanDisk’s revenue, with client (PC/smartphones) and consumer products (SD cards/USB) remaining core to its business.
Revenue Revenue
SanDisk reported a strong sequential revenue acceleration in its fiscal Q1, driven by NAND demand outpacing supply and increasing demand in its data center, edge and consumer end markets. Q1 revenue increased 22.6% YoY and 21.4% QoQ to $2.31 billion, accelerating from 8% YoY and 12.2% QoQ growth in fiscal Q4. Higher-than-expected bit growth drove the outperformance in the quarter relative to guidance of $2.1-2.2 billion.
SanDisk’s Edge segment was the primary growth driver in Q1 with revenue up 30% YoY and 26% QoQ to $1.39 billion, driven by increasing NAND content in PCs and smartphones and a positive PC refresh cycle. Consumer revenue rose 27% YoY and 11% QoQ to $652 million, while data ce3nter revenue was down (10%) YoY but up 26% QoQ to $269 million.
Q2 revenue was guided to be $2.55 to $2.65 billion, up 38.6% YoY and 12.6% QoQ at midpoint. Revenue growth is expected to accelerate further to 53% YoY in fiscal Q3 and then decelerate slightly to 49% in Q4.
For fiscal 2026, SanDisk is currently expected to generate revenue of $10.53 billion, up 43.2% YoY. Management expects demand to outpace supply through 2026, creating stronger tailwinds for pricing and growth through the year.
AI Segment Growth AI Segment Growth
SanDisk’s data center revenue, as mentioned above, declined (10%) YoY but rose 26% QoQ to $269 million, driven by increasing demand for its ‘Stargate’ enterprise SSD product line. However, revenue contribution remains small, at less than 12% of revenue.
SanDisk did not provide a numerical guide for Q2 for data center, but management noted that they are expecting sequential growth in the segment throughout fiscal 2026, with two hyperscaler qualifications underway and qualifications with an additional hyperscaler and major storage OEM planned for calendar 2026.
Management also increased their forecast for data center exabyte growth, explaining that last quarter, exabyte growth expectations were in the mid-20% range, but now are in the mid-40% range. As a result, data center Is expected to be the largest market in NAND on an exabyte basis in 2026, surpassing mobile.
Earnings Earnings
SanDisk stands out for its strong expected earnings growth through fiscal 2026 and fiscal 2027, with adjusted EPS expected to reach almost $21 by then, nearly 7X higher than the $2.99 it earned in fiscal 2025.
Q1 GAAP EPS was $0.75, a strong improvement from a ($0.16) loss in Q4, though this was down (49%) YoY from $1.46 in the year ago quarter as margins remained lower YoY. Adjusted EPS was $1.22, up 321% QoQ but down (33%) YoY.
For Q2, SanDisk guided for adjusted EPS of $3.00 to $3.40, up more than 162% QoQ. Adjusted EPS is expected to further increase to $3.67 in fiscal Q3 and $4.67 in fiscal Q4.
For fiscal 2026, SanDisk is expected to generate $13.02 in adjusted EPS, up almost 336% YoY, while GAAP EPS is projected to be $11.49, up from ($11.32) in FY25 due to the spin off. Fiscal 2027 is expected to see earnings power surpass $20, with GAAP EPS estimated to be up nearly 76% to $20.20 and adjusted EPS up nearly 59% to $20.68.
Margins Margins
Margins are lower YoY compared to pre-spinoff margins, but Q1 saw strong sequential margin expansion that is expected to accelerate in Q2.
Q1 GAAP gross margin was 29.8%, down 8.8 points YoY but up 3.6 points QoQ. Adjusted gross margin was 29.9%, down 9 points YoY but up 3.5 points QoQ.
GAAP operating margin was 8.3%, down 8.3 points YoY but up 5.6 points QoQ. Adjusted operating margin was 10.6%, down 8.2 points YoY but up 5.3 points QoQ.
GAAP net margin was 4.9%, down 6.3 points YoY but up 2.7 points QoQ, and adjusted net margin was 7.8%.
For Q2, SanDisk guided adjusted gross margin to be 41-43%, or up just over 12 points QoQ at midpoint, while adjusted operating margin is implied to be 24.2% at the midpoint of opex guidance, or up 13.6 points QoQ.
Cash Cash
SanDisk noted that in Q1 it reached a net cash position, six months ahead of schedule, though debt is still almost equivalent to its cash on hand. Cash flows were quite strong, and adjusted FCF margin showed strong expansion.
Operating cash flow was $488 million in Q1 for a 21.1% margin, up from a (7%) margin in the year ago quarter and a 4.9% margin in Q4.
Adjusted free cash flow was $438 million in Q1 for a 19% margin, up from a (10.5%) margin in the year ago quarter and 2.6% in Q4.
Cash and equivalents totaled $1.44 billion while debt totaled $1.35 billion.
Valuation Valuation
SanDisk’s valuation is somewhat hard to pin down given the company’s limited history on the public markets after its February spinoff, and its rapid 362% ascent since the end of August.
SanDisk trades at 3.3x forward PS, having peaked at 4x in November and having traded as low as 0.6x in the summer, prior to its sharp rally. For comparison, this is a lower multiple than its former parent Western Digital at 5.1x forward PS, though the two are focused on different memory market segments with WDC primarily in hard disk drives.
For forward PE, SanDisk currently trades at an 18.4x multiple, slightly above its 15.8x average from the second half of fiscal 2025 prior to its fiscal year readjustment in June. Since then, shares have traded as high as 21.7x and as low as 3x due to the sharp earnings increase expected in fiscal 2026.
Notable Risks Notable Risks
The NAND flash market has historically been quite volatile, and is shifting from significant oversupply in 2023 to expectations for substantial supply shortages through 2026. However, if NAND capacity begins to come online quickly through next year, or if demand for PCs and smartphones falters due to rising memory prices, the NAND cycle could reverse and lead to pricing pressures cutting into revenue growth and margins.
Teradyne: Quiet Beneficiary of Growing AI Compute, Memory Demand
Thematic: 8/10
Fundamentals: 6/10
Valuation: 1/10
Brief Overview Brief Overview
Surging demand for AI compute and memory chips is most obvious in the reports of leading chipmakers such as Nvidia and Micron, yet there are numerous behind-the-scenes beneficiaries of this powerful trend, such as Teradyne. Teradyne primarily provides automated test equipment (ATE) for the semiconductor industry, spanning high-performance processors and networking devices, as well as DRAM/HBM and SSD manufacturing.
For example, Teradyne’s UltraFLEXplus test system was architected specifically for high-performance AI processors and networking devices, enabling high-efficiency volume production and reducing time to market by up to 20%. Its Magnum 7H is a multi-generational HBM test platform, serving HBM3e and HBM4 needs with upgradability to service HBM4e and HBM5 when these products arrive.
Teradyne sees strong tailwinds from AI compute and memory through the end of the year, while its robotics revenue remains challenged. Management explained in Q3 that its view for 2H 2025 compute revenue is >50% higher than its expectations just three months prior, while memory test sales more than doubled QoQ in the quarter.
Over the longer term, the increasing complexity of chips, shift to chiplet or multi-chip modules, and increasing die sizes all increase test intensity. For example, the cost of scrapping racks escalates from the NVL72 to the upcoming NVL144 and NVL576 platforms due to the increase in complexity and size, creating long-term tailwinds for Teradyne’s high-performance SoC and memory test products.
Revenue Revenue
Teradyne reported revenue up 4% YoY in Q3 to $769.2 million, accelerating from an (11%) decline in Q2. Sequential growth was quite strong at 18% QoQ, though this was against a slightly soft comp of (5%) from Q2.
The primary driver of growth was Semiconductor Test Equipment, which saw revenue rise 7% YoY and 23% QoQ to $606 million, or nearly 79% of total revenue. This accelerated from a (12%) YoY decline in Q2.
Product Test Equipment revenue was $88 million, up 10% YoY and 4% QoQ, while Robotics revenue was $75 million, down (15%) YoY and flat QoQ.
For Q4, Teradyne guided for revenue to be $920 million to $1 billion, a strong acceleration to 28% YoY and 25% QoQ growth at midpoint. Analysts currently estimate revenue growth to continue accelerating to 41% YoY by fiscal Q2 (June 2026).
AI Segment Growth AI Segment Growth
Teradyne’s strongest AI-driven growth came from Memory Test Equipment (MTE), which rose more than 110% QoQ to $128 million, or ~16.6% of total revenue. However, this was down (15%) YoY. MTE revenue was driven primarily by DRAM (LPDDR and HBM performance test demand), which accounted for roughly 75% of revenue, with the remaining 25% coming from flash, primarily SSDs. As mentioned above, Teradyne’s Magnum 7H product for HBM3e, HBM4/4e and HBM5, began volume shipments in Q3.
System-on-Chip (SOC) revenue saw growth driven by AI compute and networking products, with revenue up 11% QoQ and 12% YoY to $440 million, or more than 57% of total revenue.
Management did not break out Q4’s guidance for these two segments, noting that AI-driven test demand remains robust across compute, memory and networking. Analysts from UBS placed the QoQ growth in Semi Test Equipment at ~$200 million as a whole, with management explaining that the growth will likely be 2/3 compute and networking-driven and 1/3 memory-driven, primarily by HBM.
Earnings Earnings
Teradyne reported strong sequential EPS growth in Q3 that will continue in Q4, though EPS was lower YoY as margins were a couple of points lower compared to last year.
In Q3, Teradyne reported GAAP EPS up 53% QoQ but down (16%) YoY to $0.75. Adjusted EPS was $0.85, up 49% QoQ but down (6%) YoY, and beating estimates by 7.5%.
For Q4, Teradyne guided for GAAP EPS to be $1.12 to $1.39, pointing to QoQ growth of 67% and YoY growth of 39% at midpoint. Adjusted EPS was guided to be $1.20 to $1.46, up more than 56% QoQ and 40% YoY at midpoint. Analysts expected adjusted EPS growth to accelerate sharply over the next two quarters, with current consensus estimates calling for growth of 61% in fiscal Q1 and 112% YoY in fiscal Q2.
Adjusted EPS for fiscal 2025 is projected to be $3.52 for 9.3% YoY growth, before accelerating to nearly 46% growth to $5.13 in fiscal 2026.
Margins Margins
While margins were down YoY, Teradyne reported strong sequential growth and improving operating leverage versus Q2, with the expansion in operating margin at 4X the rate of gross margin. Operating margins are guided to continue expanding at a similar rate in Q4.
GAAP gross margin was down 0.8 points YoY but up 1.2 points QoQ to 58.4%. Adjusted gross margin was down 1.2 points YoY but up 1.2 points QoQ to 58.5%
GAAP operating margin was down 1.7 points YoY but up 5 points QoQ to 18.9%. Adjusted operating margin was down 2 points YoY but up 5.3 points QoQ to 20.4%.
GAAP net margin was down 4.2 points YoY but up 3.5 points QoQ to 15.5%. Adjusted net margin was down 2.3 points YoY but up 3.6 points QoQ to 17.7%.
For Q4, Teradyne guided for adjusted gross margin to be 57-58%, down 1 point at midpoint on onetime supply costs to meet accelerated demand. Adjusted operating margin was guided to be 24-27%, up 3.6 to 6.6 points at midpoint on a much lower opex run rate at 31-33% of revenue, versus 38.1% in Q3.
Cash Cash
Cash flow margins contracted sharply Q3, though this was primarily driven by a large QoQ increase in accounts receivable, providing an extra layer of confidence in the upcoming revenue acceleration in the next couple of quarters.
Operating cash flow was $49.1 million for a 6.4% margin, though OCF margin had been >22% for the past five quarters. The sharp contraction was primarily due to a $161 million sequential increase in accounts receivable.
Free cash flow was $2.4 million for a 0.3% margin, down from 20.2% in the prior quarter due to the jump in AR.
Cash and equivalents were $297.7 million in Q3, down from $367.9 million in Q2, while Teradyne took on new debt of $200 million in the quarter (with this being its only debt).
Valuation Valuation
Teradyne is trading at peak multiples, with shares currently at 9.9x forward PS, well above its five-year average of 6.6x and just shy of its 10.4x peak.
On a forward PE basis, Teradyne trades at 55x forward earnings, more than 50% above its five-year average of 35.7x and just off its peak of 58x. Even looking ahead to fiscal 2026, with the ~36 point acceleration for adjusted EPS, Teradyne is trading at a 38x multiple, still above its average.
Notable Risks Notable Risks
It is still early in this AI-driven inflection in growth for Teradyne, and management has pointed out that SoC test growth in Q3 and Q4 should not be extrapolated into the early part of next year, as timing for Q1 and Q2 remains uncertain and growth may be lumpy. The thin cash flows are not necessarily a red flag yet, but if margins remain depressed, this could be more of a risk to watch moving forward.
Dell: FY25 AI Server Revenue Raised to $25 Billion, Up 150% YoY
Thematic: 8/10
Fundamentals: 6/10
Valuation: 7/10
Brief Overview Brief Overview
Dell’s story is primarily centered around its AI server growth opportunities, benefiting from the ramp of Nvidia’s Blackwell and Blackwell Ultra GPUs, though the company does have other outlets into storage, networking, and commercials PCs and workstations. Dell is working closely with Nvidia to help accelerate enterprise AI adoption via the Dell AI Factory, and notably was the first to deliver Nvidia’s GB300 NVL72 racks to CoreWeave back in July.
Riding Nvidia’s coattails this past quarter, Dell reported strong AI server metrics in Q3, booking record orders and reaching a new record for backlog, while guiding for record shipments next quarter. Dell also raised its AI server revenue forecast by another $5 billion, now seeing $25 billion this year, up 150% YoY and up $10 billion from its initial $15 billion forecast at the start of the year. Nvidia’s strong visibility into Blackwell and Rubin sales through 2026 also hints that this AI server momentum may persist through next year.
However, growth in Dell’s consumer and commercial PC business remains low, and the company must battle rather thin AI server margins in a competitive market. Dell did note that AI server margins improved sequentially and helped drive some margin expansion, but the company must prove that this dynamic can be maintained.
Revenue Revenue
Dell reported a solid Q3 with revenue of $27.0 billion, though this was around 1.1% shy of consensus estimates for $27.3 billion. Revenue was up 10.7% YoY but down (9.7%) QoQ, as Q2 was positively impacted by a sharp ~356% QoQ increase in AI server shipments to $8.2 billion due to timing of fulfillment.
Dell’s Infrastructure Solutions Group (ISG) revenue increased by double-digits YoY for a sixth consecutive quarter, up 24% YoY in Q3 to $14.1 billion. This slowed from 44% growth in Q2 to $16.8 billion, though again this was impacted by the strong QoQ growth on shipment timing. Within ISG, Server and Networking revenue was up 37% YoY to a record $10.1 billion, while Storage revenue was down (1%) YoY to $4 billion.
Dell’s Client Solutions Group (CSG) revenue increased 3% YoY to $12.5 billion, a slight acceleration from 1% YoY growth in Q2, as Dell noted robust international demand and continued growth in Commercial demand. Commercial revenue was up 5% YoY to $10.6 billion, acceleration from 2% YoY in Q2, while Consumer revenue was down (7%) YoY to $1.9 billion.
For Q4, Dell guided for $31-32 billion in revenue, marking a sharp acceleration to 32% YoY at midpoint and 16.7% QoQ, with nearly 30% of this revenue coming from AI servers. Full-year fiscal 2026 revenue was guided to be $111.2-112.2 billion, up 17% YoY at midpoint. This was a solid $4.7 billion raise from prior guidance for 12% growth to $107 billion at midpoint.
AI Segment Growth AI Segment Growth
Dell’s AI server momentum was robust in Q3, with the company reporting record orders and backlog, with a strong QoQ increase in revenue guided for Q4. Dell also raised its full-year AI server revenue forecast to $25 billion, up ~150% YoY.
AI server revenue was $5.6 billion in Q3, up 93% YoY but down nearly (32%) QoQ as Dell fulfilled some larger orders in Q2. For Q4, Dell guided for $9.4 billion in AI server revenue, a fresh record, and representing YoY growth of 348% and QoQ growth of 68%.
AI server orders reached a record $12.3 billion, up ~120% QoQ and also surpassing Q1’s $12.1 billion. Fiscal 2026 YTD orders reached $30 billion, more than 3X higher than the $9.4 billion recorded in the same period last year.
AI server ending backlog rose ~57% QoQ to a record $18.4 billion with a significant shift to the GB300 in the quarter. Dell added that its five-quarter pipeline remains multiples of its backlog and grew sequentially across neoclouds, sovereigns and enterprises.
Earnings Earnings
Dell reported a strong GAAP EPS beat in Q3, with earnings of $2.28, up 39% YoY and 15.7% ahead of estimates for $1.97. Adjusted EPS also beat by 4.5%, rising 17% YoY to $2.59, driven by improved profitability in AI servers and storage.
For Q4, Dell guided for GAAP EPS growth to accelerate to 42% YoY to $3.05 at midpoint, while adjusted EPS was guided to accelerate to 31% YoY to $3.50 at midpoint. This accelerate is poised to continue into Q1, with adjusted EPS forecast to rise 52.4% YoY to $2.36.
For fiscal 2026, Dell guided GAAP EPS to be $8.38 at midpoint, up 31% YoY, a $0.40 raise from its Q2 guide of $7.98 for 25% growth. Adjusted EPS was guided to be $9.92 for 22% YoY growth, also a $0.37 raise from its prior guide for $9.55.
Margins Margins
Margins were mixed for Dell in Q3, with gross margin rebounding QoQ but remaining down YoY, and operating margin expanding on both a YoY and QoQ basis.
Q3 GAAP gross margin was 20.7%, down 1.3 points YoY but up 1.4 points QoQ. Adjusted gross margin was 21.1%, down 1.4 points YoY but up 2.4 points QoQ.
Q3 GAAP operating margin was 7.8%, up 0.7 points YoY and 1.8 points QoQ. Adjusted operating margin was 9.3%, up 0.1 points YoY and 1.6 points QoQ.
Q3 GAAP net margin was 5.7%, up 0.9 points YoY and 1.8 points QoQ. Adjusted net margin was 6.5%, flat YoY and up 1.2 points QoQ.
By segment:
ISG adjusted operating margin was 12.4%, down 0.9 points YoY but up 3.6 points QoQ. This was driven by storage and sequential improvement AI server profitability, with management saying AI server margins were mid-single-digit, up from an implied 2.5% last quarter.
CSG operating margin was 6.0%, down 0.2 points YoY and 0.4 points QoQ.
Cash Cash
Operating and free cash flow both declined YoY and QoQ, though adjusted FCF more than doubled YoY due to financing receivables.
Operating cash flow was $1.17 billion, down (25%) YoY. OCF margin was 4.3%, down from 6.4% in the year ago quarter and 8.5% in Q2.
Free cash flow was $506 million, down (45%) YoY. FCF margin was 1.9%, down from 3.8% in the year ago quarter and 6.3% in the prior quarter. Adjusted FCF, however, was up 133% YoY to $1.67 billion for a 6.2% margin, up from 2.9% in the year ago quarter.
Cash and equivalents totaled $9.57 billion, while debt totaled $31.24 billion, with $7.39 billion being current.
Valuation Valuation
Dell trades slightly above its average forward PS multiple at 0.76x currently versus its five-year average of 0.69x, though the company has rarely traded above 1x over the past two years.
On the bottom line, Dell trades at 12.9x forward PE, also slightly above its five-year average of 12x, but well below its October high at 16.6x and previous resistance around the 18x level.
Notable Risks Notable Risks
The main risk to watch for Dell is AI server margins, and whether or not the company can maintain mid-single-digit margins moving through 2026, considering the competitiveness of the industry between Super Micro and Taiwanese ODMs such as Foxconn and Quanta.
Additionally, the sharp surge in memory prices is expected to cause substantial PC price hikes to preserve margins – Dell is said to be raising commercial PC prices by as much as 30% as a result. These price hikes could weigh on demand and cause PC growth to slow next year, with IDC forecasting the industry to decline (5%) in a conservative scenario and as much as (9%).
Micron: HBM, LP Server DRAM Driving Strong Growth, FQ2 to Accelerate to 37% QoQ
Micron is a primary beneficiary of rapidly increasing dollar content of high-bandwidth memory (HBM) chips with each new generation of GPUs. AI training and inference rely heavily on HBM for the massive memory bandwidth that complex models require. AI servers also use more DRAM and NAND than a traditional server. These are reasons that Micron’s cyclical fundamentals could become more secular as the AI economy is built out.
In fiscal 2025, Micron’s HBM, high-capacity dual in-line memory modules (DIMMs) and low-power (LP) server DRAM revenue reached $10 billion, up more than fivefold from the prior year, while HBM alone reached $2 billion in revenue in Q4.
Management expects robust AI server demand, the shift to HBM4 and tight DRAM supply to drive substantial records for revenue, gross margin, EPS and free cash flow for fiscal 2026 with business strengthening throughout the year. Micron also expects these tight market conditions to persist beyond 2026.
What Micron will need to answer is if the cyclical nature of the memory market will smooth out as the dollar content of memory is rapidly increasing. Data center is already proving to be a strong driver of growth for Micron, accounting for 56% of sales in FY25, up from 35% in FY24. On a dollar basis, data center revenue surged 137% YoY to $20.75 billion for the year.
Overall Revenue Growth Overall Revenue Growth
Micron reported record Q1 revenue of $13.64 billion, beating estimates by 5.9% and accelerating 10.7 points to 56.7% YoY growth. Sequentially, growth was 20.6% QoQ, just one point slower than Q4’s 21.7% QoQ growth. DRAM products (within that HBM and LPDDR5X) were the primary driver of Q1’s results, with revenue up 69% YoY and 20% QoQ to $10.8 billion, or 79% of revenue.
Micron also delivered one of the largest beats in AI semiconductors outside of Nvidia’s May 2023 report with its Q2 guidance. Management forecast revenue of $18.7 billion, +/- $0.4 billion for next quarter, more than 31% above estimates for $14.23 billion. This corresponds to a very sharp 75.5 point acceleration to 132.2% YoY growth, while QoQ growth would accelerate to 37.1% at the midpoint of the guide.
Micron has not provided a full-year guide for revenue, but current consensus estimates call for 98% growth to $73.98 billion in revenue.
AI Revenue Growth AI Revenue Growth
In Q1, Micron's data center revenue rose 55% YoY to $7.66 billion (56% of company revenue), with growth primarily driven by DRAM (HBM/LPDDR5X) products and aided by data center SSDs and NAND components.
Micron’s Cloud Memory Business Unit (CMBU), which consists of its HBM, high-capacity dual in-line memory modules (DIMMs), and low-power server DRAM solutions, saw Q1 revenue of $5.28 billion, up 99.5% YoY.
Micron’s other data center unit, its Core Data Center Business Unit (CDBU), consists primarily of data center SSDs and NAND components. This unit saw Q1 revenue grow just 4% YoY to $2.38 billion.
Earnings Earnings
Micron is expected to see robust earnings growth this fiscal year as margins are rapidly expanding on surging prices. In Q1, Micron reported GAAP EPS of $4.60, up 175% YoY; this also is a sharp uptick from $2.83 in Q4.
For Q2, Micron guided for GAAP EPS to be $8.19, +/- $0.20, nearly 74% ahead of estimates for $4.71 and corresponding to YoY growth of almost 481%, a 306 point acceleration. GAAP EPS growth is expected to remain >250% for both Q3 and Q4 to $9.37 and $10.04.
For the full year, Micron is expected to deliver GAAP EPS of $31.17, more than quadrupling from $7.59 in fiscal 2025. Earnings estimates also moved more than 60% higher following Q1’s report and Q2’s blowout guide, moving from $19.42 to the now $31.17 estimate.
Margins Margins
Micron’s margin turnaround story has been impressive, with gross margin up nearly 57 points over the last two years and operating margin up 69 points. Micron also guided for substantial records for gross and operating margin in Q2, on the backs of strong pricing.
GAAP gross margin in Q1 was 56%, up 17.6 points YoY, aided by the strong growth in CMBU which carried a 66% gross margin in the quarter. For Q2, GAAP gross margin was guided to be 67% at midpoint, an 11 point sequential expansion and up 31.2 points YoY.
GAAP operating margin was 45%, up 12.7 points QoQ and 20 points YoY, again aided by CMBU which carried a 55% margin in the quarter. For Q2, Micron implied operating margin to be 58.7%, up 12.7 points QoQ and 36.7 points YoY, signaling strong tailwinds from surging DRAM prices.
GAAP net margin was 38.4% in Q1, up 10.1 points QoQ and nearly 17 points YoY.
Cash Cash
Operating cash flow was $8.41 billion in Q1, up more than 159% YoY and nearly 47% QoQ. OCF margin was 61.7%, up 10 points QoQ and up 24.4 points YoY.
Adjusted free cash flow was $3.91 billion in Q1, up sharply from $803 million in Q4 and $112 million in the year ago quarter. Adjusted FCF margin was 28.6%, up from 7.1% in the prior quarter and 1.3% in the year ago quarter.
Micron reported total cash and equivalents of $12.0 billion and total debt of $11.76 billion.
Valuation Valuation
Despite its recent rally, Micron trades at somewhat reasonable multiples, well below its peak from 2024. On the top line, Micron trades at 4.4x forward revenue, 22% above its average 3.6x multiple and far below its peak of 6.8x from mid 2024.
On the bottom line, Micron trades at 9x forward earnings, though its 40.9x average is skewed higher by 2024’s >100x multiples when margins were razor-thin. Since the start of 2025, Micron has traded as high as 16x forward earnings, and as low as 3x on its fiscal-year reset in September.
Notable Risks Notable Risks
Micron’s growth to this point and beyond has been centered around HBM and data center DRAM products, both on the top and bottom lines. Rising HBM demand in 2026 as next-gen GPU systems ramp and content growth in LP server DRAM are strong tailwinds for growth, yet sharply rising DRAM prices from tight supply could cut into demand for consumer electronics products. This is Micron’s second largest segment and growth driver (Mobile and Client), with nearly $4.3 billion in revenue and a 47% operating margin in Q1, and any demand softness from price hikes could be felt more acutely in 2026.
Conclusion:
We are thrilled about our new tier Discovery as we’ve seen immediate results in 2025 after delivering a separate tier for new ideas. We quickly spotted the limitations around running an active portfolio that does not dedicate a separate effort to new idea generation as the market moves fast with new winners emerging every year. It would be easy to miss up-and-coming momentum stocks without this tier – especially for enthusiastic AI investors, such as ourselves.
We believe pointing out names that are not often spoken of (yet) while also reiterating those that are well-known yet remain strong quarter-after-quarter allows us to go beyond the I/O Fund portfolio to offer maximum value from our research efforts.
As we look at Q4 and beyond, we believe this quarterly analysis combined with an actively managed Top 10 list will become a strong offering in 2026. Our cumulative record proves we are one of the strongest teams in the world on AI stocks and 2025 was no exception. Moving forward, our goal is to use our proven methodology to deliver additional value add as we participate heavily in the once-in-a-lifetime trend of AI.
In the coming weeks, we expect things to shift rapidly as new information is published daily during earnings season. Please reference our Top 10 Watchlist spreadsheet and incoming analysis as critical tools for staying on top of the Must Know stocks in the space. We will also cover these stocks in a Discovery webinar hosted by Knox Ridley early next month.
Stay tuned for frequent updates!
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Every Thursday at 4:30 pm Eastern, the I/O Fund team holds a webinar for premium members to discuss how to navigate the broad market, as well as various stock and crypto entries and exits. Beth Kindig offers weekly deep dives including lesser-known cryptocurrencies and AI stocks, plus the team offers trade alerts. The I/O Fund team is one of the only audited portfolios available to individual investors. If you’d like to subscribe to the Advanced Market Signals plan, email us at premium@io-fund.compremium@io-fund.com.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
The difference between Palantir and other AI-enabled database competitors is that Palantir is able to go further in answering questions that a model would struggle to answer. Traditional business intelligence companies require a complete data set whereas Palantir is able to tackle situations where there is not a complete data set. You can think of the competitive advantage as being actionable depth, which Palantir has described as “the reasoning that goes into decision-making, not just data.”
Palantir’s Artificial Intelligence Platform (AIP) integrates generative AI with operational data and workflows, and when combined with Palantir’s other platforms Foundry and Apollo, it provides an AI service mesh that can run hundreds of microservices, scale compute through its Rubix engine and orchestrate updates through Apollo.
Additionally, Palantir’s knowledge graph referred to as Ontology is a distinct advantage. The graph offers better context than a large language model would on its own – or as Palantir states, it’s “the reasoning that goes into decision-making.” Palantir made some key upgrades to AIP with the introduction of AI forward-deployed engineers (FDEs) and AI Hivemind, and brought Ontology to the edge, allowing it to be deployed on mobile devices.
Fueled once again by AIP, Palantir delivered one of the best reports across tech in the third quarter, with revenue accelerating nearly 15 points sequentially to almost 63%, with strong growth in key metrics and a 50 point acceleration in US Commercial revenue since the start of the year. We break this down and more below.
New Product Upgrades – AIP FDEs, Hivemind and Edge Ontology
Palantir made a handful of upgrades to its AIP and Ontology in Q3, unveiling AIP forward-deployed engineers (FDEs) in beta, AI Hivemind, and Edge Ontology, all aimed at accelerating AI deployment with its customers.
AI FDEs build upon Palantir’s take on software engineers, the forward deployed engineer, which sits at the intersection of software, sales, and platform engineering, embedding within customer teams to closely develop and tailor AI software and solutions directly to their needs. Palantir now brings this to AIP, with the AI FDEs being its AIP-native deployment AI agent that “understands how to connect to data sources, how to integrate and transform data, how to create ontologies and functions and build applications.” AI FDEs function in conversational commands, allowing customers to easily turn requests into autonomously-executed Foundry operations.
Palantir says the AI FDEs are increasing productivity for customers, noting that at one customer, two of its human FDEs utilized AI FDEs to migrate a legacy data warehouse in five days, a task which Palantir says would normally have taken up to two years.
AI Hivemind is a new tool within AIP that Palantir says will orchestrate a “swarm of dynamically generated agents to tackle hard problem solving, idea generation refinement and executable proposal generation that is integrated with Ontology and therefore aware of the context of your enterprise.” Palantir says the tool was developed to help government clients solve extremely complex problems with classified data (such as generating intricate mission plans), but it has already been tested by commercial clients to help “identify bottlenecks to their supply chain, proactively developing possible solutions and then leveraging AI FDE to code that up into an actual solution.”
Edge Ontology is Palantir’s new, lightweight Ontology that allows it to run on mobile devices, letting customers build mobile apps or embedded software for hardware such as robots or drones. Edge Ontology is also fully integrated with AIP. While Palantir was very thin on details, it’s likely tied to its existing partnership with Qualcomm, where the two brought Ontology to edge devices powered by Qualcomm’s Dragonwing processors. This partnership focused on the industrial, auto and manufacturing sectors, and remote and offline environments.
Together, the new product upgrades reflect Palantir’s dedication to continuously improve its platform and help customers consistently solve problems daily. Financially, the three new upgrades can help accelerate deployments and adoption of AIP, help secure more (and potentially larger) contracts, and tap into new markets such as IoT at the edge.
Financials
Revenue Accelerates Nearly 15 Points to 63% YoY, Up 18% QoQ
Palantir reported $1.18 billion in revenue in Q3, up an impressive 18% QoQ and beating estimates by 8.4%, driven by unwavering momentum in US Commercial. On a YoY basis, revenue growth accelerated 14.8 points to 62.8% YoY, the largest sequential acceleration to date and marking Palantir’s highest growth rate since going public. Over the last nine quarters, topline growth has accelerated ~50 points, from just 12.7% in Q2 2023, a rare feat to accomplish.
For Q4, Palantir guided for revenue up 60.6% YoY to $1.327 to $1.331 billion, well ahead of estimates for 44.2% growth to $1.19 billion. While this does represent a marginal deceleration at face value, this sequential deceleration is in line with trends from previous quarters.
For the full year, Palantir raised its revenue outlook to $4.396 to $4.400 billion, pointing to YoY growth of 53.5% at midpoint, a sharp acceleration from 29% growth in 2024. To put in perspective the strength of this acceleration, Palantir had initially guided for just 30.9% growth to $3.76 billion in revenue back in Q4 2024; growth is now more than 22 points faster.
Impressive 28 Point Acceleration in US Commercial to 121% YoY
Palantir’s US Commercial segment is generally seen as the primary vector for its AIP-driven growth, with robust momentum only accelerating further in Q3.
US Commercial revenue grew 29% QoQ and 121% YoY to $397 million in Q3, accelerating from 93% YoY growth in Q2. Since the start of the year, US Commercial growth has accelerated 50 points, and since the start of 2024, growth has accelerated 81 points.
For the full year, Palantir significantly boosted its US Commercial growth outlook to >104% YoY, up from 85% previously. This corresponds to revenue of $1.433 billion, up from $1.302 billion previously.
Key metrics for the segment were very strong: TCV closed (total contract value) surged 342% YoY to a record $1.31 billion, and TTM TCV was $3.8 billion, up 217% YoY. Remaining deal value (RDV) rose 199% YoY and 30% QoQ to $3.63 billion. US Commercial deals closed of >$1 million were up 2X YoY and deals closed of >$5 million were up 5X YoY.
Key Segments – Government and Commercial Revenue
Q3 also marked the first time Commercial growth outpaced Government segment growth since Q2 2024, fueled by the robust momentum and sharp acceleration in US Commercial as discussed above.
Commercial revenue rose 21.5% QoQ and 73% YoY to $548 million, a 26 point acceleration from 47% YoY growth last quarter. International Commercial revenue was ~$151.7 million in Q3, up ~10% YoY and 5% QoQ.
Government revenue rose 14.5% QoQ and 55% YoY to $633 million, a six point acceleration from 49% YoY in Q2. Government remained Palantir’s largest segment at ~53.6% of revenue. US government revenue was up 52% YoY and 14% QoQ to $485.9 million, while International Government revenue was ~$146.8 million, up nearly 66% YoY and 16% QoQ.
Other Key Metrics – NRR Expands, Strong Customer Growth
Palantir had a handful of other strong key metrics that support strong revenue growth continuing despite the sharp acceleration the company delivered in Q3.
Palantir’s net retention rate (NRR) expanded six points sequentially to 134%, and over the past two years, NRR has risen an impressive 27 points, with Palantir noting that AIP is continuing to drive existing expansions and new customer conversions. Palantir also continued to emphasize that NRR does not include revenue from new customers acquired over the last twelve months, and accelerating momentum in quarterly deals closed supports more upside to NRR in 2026.
Moving to deals closed, Palantir closed 201 deals of >$1 million in Q3, up 30% QoQ; this was a sharp acceleration from 13% QoQ growth in Q2. Palantir also signed more deals in all of its cohorts (>$1M, >$5M and >$10M) in Q3 than it had in Q2.
To put deal growth in the context of NRR, Palantir has signed 629 >$1M deals over the last twelve months, up more than 61% from 390 in the same period last year – with none of these new deals appearing yet in NRR.
However, there were a few blemishes within key metrics – billings growth decelerated from 53.5% YoY in Q2 to 48.8% YoY in Q3 to $1.23 billion, with QoQ growth also decelerating from 21.8% QoQ to 11.2% QoQ. RPO growth decelerated from 77% YoY and 27% QoQ in Q2 to 66% YoY and 8% QoQ in Q3 at $2.60 billion.
Margins – Q3’s Rule of 40 of 114%, from 94% in Q2
Margins strengthened considerably in the quarter, with adjusted operating margin surpassing 50% with more expansion guided for Q4. Palantir’s Rule of 40 score (revenue growth + adj operating margin) expanded to a wild 114%, up from 94% last quarter and 68% last Q3.
Gross margin was 82% in Q3, up one point QoQ and two points YoY, while adjusted gross margin was 84%, up two points YoY and QoQ.
GAAP operating margin was 33%, an impressive 6 point QoQ and 17 point YoY expansion. Adjusted operating margin was 51%, breaking past 50% for the first time and up 5 points QoQ and 13 points YoY. For Q4, Palantir guided for adjusted operating margin to be 52%, showcasing its ability to drive strong margin expansion alongside swift revenue acceleration. Full year adjusted operating margin guidance was raised from 46% to 49%.
GAAP net margin was 40%, up 7 points QoQ and 20 points YoY. Adjusted net margin was 45%, up 5 points QoQ and 12 points YoY. Palantir is one of the few, if not only, tech companies to have 40% GAAP net margins with revenue growth accelerating above 60%.
Earnings
Palantir reported $0.18 in GAAP EPS in the quarter, up 200% YoY, while adjusted EPS was $0.21, beating estimates by 25.5% and rising 110% YoY. Palantir did not provide a specific guide for EPS for Q4, though current estimates are pegged at $0.12 in GAAP EPS and $0.22 in adjusted EPS, up 300% YoY and 57% YoY, respectively.
For FY25, Palantir is expected to earn $0.72 in adjusted EPS, up nearly 76% YoY, before slowing to 39% growth to $1.01 in FY26.
Cash
Cash flows were strong, though cash flow margins dipped on a YoY and QoQ basis. Operating cash flow was $507.7 million for a 43% margin, shrinking from a 54% margin in Q2 and 58% in the year ago quarter.
Adjusted free cash flow was $539.9 million for a 46% margin, down from 57% in Q2 and 60% in the year ago quarter. Palantir raised its adjusted FCF guidance for the year to $1.9 to $2.1 billion, or a 45.5% margin, up from a 42.8% margin previously.
Cash and equivalents totaled $6.4 billion and debt remained zero.
Valuation
Valuation is the crux for Palantir as the stock trades at 100x forward revenue, nearly triple its five-year average of 36x and in rather uncharted territory for software stocks. On the bottom line, Palantir trades at 256x forward adjusted EPS despite a >40% margin, more than double its average 109x multiple.
The valuation with Palantir is a gamble as the company is attempting to set a new bar for AI software, with >100x forward sales multiples only achieved for short fashion in 2021 for a handful of prior market darlings, whose stocks have yet to return to those prices. This elevated valuation may also present a risk if/when the company reaches peak revenue growth as comps will quickly get tougher.
Conclusion
Palantir delivered one of the strongest earnings reports in all of tech in Q3, with revenue growth of nearly 63% and a strong 28 point acceleration in its AI-driven US Commercial segment. Since the start of 2025, revenue growth has accelerated more than 23 points, with US Commercial revenue accelerating 50 points. Trends in other key metrics such as NRR and quarterly deals closed remain robust, although billings and RPO growth decelerated in the quarter. However, what Palantir has to contend with is an extended valuation, in uncharted territory even by its own measures, and where the market will ultimately price its shares.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund do not own shares in PLTR at the time of writing and may own stocks pictured in the charts.
As we have discussed for our Discovery and Pro members, AI networking is one of the strongest trends for this year and next, driven by scale-up and scale-out networking to support larger GPU racks and accelerating GPU cluster sizes. Celestica is an under-the-radar beneficiary of this trend, capitalizing on strong demand for 800G and 1.6T Ethernet networking switches and leveraging its deep ties to hyperscalers.
Celestica guided for one of the most impressive accelerations seen in this last quarter of earnings, underpinned by its 800G switches accelerating next year with 1.6T on deck. For 2026, Celestica expects revenue growth to accelerate around five points to 31% YoY in 2026, whereas consensus had been pegged at just 17% YoY. This strong upside is being driven by networking and custom AI compute platforms with visibility into 2026-2027.
In terms of AI revenue, Celestica’s Cloud and Connectivity Solutions (CCS) segment is guided to generate $9 billion in revenue in 2025, up ~40% YoY, accounting for nearly 74% of total revenue. CCS, which includes AI networking, server, storage and rack-scale system solutions, is Celestica’s main growth driver, expected to grow ~40% annually in 2026 and 2027.
Celestica is closely linked to Broadcom’s networking platforms as a key vendor, serving major customers such as Google and Meta, with some of its notable product engagements including Google’s TPU server racks, and Meta’s Minerva ASICs servers, Wedge400 switches and also its next-gen Tomahawk5-based 400G AI fabric switch Minipack3. Additionally, management’s commentary suggests that OpenAI could become a key customer as soon as 2027.
Below, we cover Celestica’s strategic positioning in the high-bandwidth Ethernet market, its engagements with hyperscalers and upcoming platform ramps, its updated growth outlook for FY26 and beyond, and more.
Celestica’s Strategic Positioning in Custom Networking Switches
Celestica is strategically positioned in the AI supply chain as it provides hyperscalers with highly customized data center networking switches, servers and storage platforms, alongside custom rack-scale integration services.
Celestica is also closely aligned with Broadcom, as a preferred provider offering customized high-performance Ethernet switches based on its Tomahawk platform and integrated XPU-based racks and systems. CEO Rob Mionis explained that when Broadcom launches new silicon, such as its newest Tomahawk6, “they’ll work with us to develop products, and those products end up in the major hyperscalers.”
Growth opportunities are primarily centered around its high-bandwidth Ethernet switch portfolio focused on back-end networking, with the company being the leading supplier with 41% share of the >200G switch market through Q2, and with 55% share of the custom switch market (up from 40% in 2024). The back-end networking positioning is important for Celestica as it means the company is exposed to the faster-growing segment of Ethernet switching – the back-end TAM is forecast to grow at a 56% CAGR through 2029 on scale-out, and potentially soon, scale-up demand, whereas front-end (user-facing) is forecast to grow at a 20% CAGR.
Per management, the back-end also sees a much faster refresh rate of every 18-24 months versus >5 years for front-end deployments, and it adopts the newest and fastest bandwidths (800G and soon 1.6T) due to the greater performance and reliability requirements of GPU-to-GPU and rack-to-rack communications.
Celestica’s primary products include scalable top-of-rack switches and high-bandwidth Ethernet switches (>400G). Its 100G and 400G switches are optimized for data center leaf-and-spine deployments.
For 800G switches, Celestica’s DS4100/DS4101 are based on Broadcom’s Tomahawk4 portfolio and the DS5000 is based on the Tomahawk5, targeting high bandwidth data center leaf-spine, and top-of-rack applications. For additional clarity on 800G switch dynamics, management explained that they have seen “tremendous growth in 800G this year to the point where we'll end 2025 with roughly a 50% split between 800G and 400G in terms of the products that we're delivering. As we look into 2026, we're seeing the 800G demand, in particular, accelerating.”
For 1.6T, Celestica will offer the DS6000 and DS6001, based on Broadcom’s upcoming Tomahawk6 offering 102.4Tbps bandwidth, with availability later in 2026. The DS6000 comes in an air-cooled version with linear pluggable optics (LPO) to improve power efficiency, while the DS6001 features hybrid-cooling and the first to integrate direct-to-chip liquid cooling. Both 1.6T switches are optimized for AI back-end networking (scale-out and scale-up), as well as large-scale AI fabrics for AI training and inference for frontier model sizes. Management expects the 1.6T upgrade cycle to emerge in late 2026 but primarily land in 2027, with one customer giving visibility to a back-half 2026 ramp and multiple other ramps occurring through 2027.
Celestica is also already making early investments for 400G SerDes to support 3.2T switch platforms, though it does not expect 3.2T mass production to arrive until 2028. Management is also preparing for co-packaged optics (CPO) and other interconnect types such as co-packaged copper (CPC), and while it sees some potential CPO shipments with 1.6T, it does not expect CPO to emerge in full-force until the 3.2T cycle.
Although these switch products can be highly customized, they support open-source networking stacks such as SONiC, offering hyperscalers flexibility in deployments, facilitating integration into existing software and hardware ecosystems, and letting customers avoid vendor lock-in. This is furthered with Celestica’s extensive Circular Services offering, spanning hardware lifecycle management, remanufacturing, and refurbishment to extend hardware lifetimes and reduce TCO.
ODM Pivot Driving Hyperscaler Growth, with Google and Meta Key Customers (and Soon Likely OpenAI)
This positioning in custom Ethernet switches also pushes Celestica towards more of an ODM (original design manufacturer) model from a traditional EMS contract manufacturer, as it engages more deeply across the design and engineering phase, tailoring products exactly to its hyperscaler customers’ needs. Some of its notable confirmed/implied hyperscaler product engagements include Google’s TPU server racks, and Meta’s Minerva ASICs servers, Wedge400 switches and also its next-gen Tomahawk5-based 400G AI fabric switch Minipack3.
Here is an example of what Celestica’s involvement would look like, such as on Meta’s Wedge400, its top-of-rack networking switch based on Broadcom’s Tomahawk3 from 2021:
Celestica works with Meta on system requirements and finalizes system level architecture.
Celestica is fully responsible for hardware design and complete prototyping, along with functional and reliability tests and diagnostic software development.
After this, Meta reviews Celestica’s engineering design and test reports and moves to production.
Management’s discussion on its first-of-kind rack-scale liquid cooled 1.6T networking win with a hyperscaler (likely to be Meta for its upcoming Santa Barbara racks) also shed light on why Celestica continues to win these engagements:
“The customer required an accelerated road map to allow the solution to be early to market, leveraging Broadcom's Tomahawk 6 SC silicon, making speed to market a key consideration. In addition, the customer required multi-node manufacturing capabilities in Asia and the U.S. to support the delivery of the program. As with many of our key engagements, managing complexity was a defining factor.
Celestica was awarded the program earlier this year based on a strong working relationship with the customer and their confidence in our industry-leading design engineering. They also valued our advanced manufacturing capabilities, specifically our ability to operationalize highly complex production lines for liquid cooled racks at scale and to do this faster and more seamlessly than other potential partners.
After receiving initial Tomahawk 6 samples earlier this year, we quickly stood up an operational prototype for the 1.6T switch and believe we were the first team anywhere to have done so. The program is scheduled to begin mass production next year.”
This ODM pivot and ability to co-design and manufacture highly customized, next-gen networking switch and rack solutions at speed is quite visible in Celestica’s hyperscaler growth trajectory, as hyperscalers are expected to account for ~$6.93 billion of revenue in 2025, up from $2.19 billion in 2022, an incredible ~47% CAGR. However, the ODM positioning also presents a risk as even a shift to higher complexity, higher value products may be unable to produce continuous margin expansion into the low-teens.
Also aiding this hyperscaler growth is a high level of stickiness and deep customer engagement across data rate upgrades. Management explained that they have been able to consistently upgrade all customer engagements each cycle:
“When you look at where we carved out this industry-leading position in networking, it started in 400-gig, and we were able to translate all of those engagements into 800-gig. And those engagements have been expanding incrementally to new opportunities, and we fully plan to translate all of our 800-gig engagements into 1.6T as well, and we're on track to do that.” Management quietly dropped that they currently have ten programs underway with 1.6T.
Digital Native Customer Win and 2027 Ramp (Hint: It’s Likely OpenAI)
Back in January, Celestica announced a “a significant new HPS win with a leading digital native company, who is a pioneer in the commercialization of AI applications,” collaborating with this company to “deliver a full rack, which is an optimized AI system solution built around the customer's custom ASIC accelerator.” The rack-scale solution also will include Celestica’s 1.6T switches and rack-level cooling and connectivity. At the time, Celestica said that “production for this program is expected to begin ramping in the latter part of 2026.”
In Q3, Celestica provided an update, saying that the “design work for this program is well underway, and we expect to receive initial XPU deliveries in the second half of 2026 to support early test deployments with full-scale production expected to commence in 2027.” Management did add that 2026’s $16 billion revenue outlook does not include any contribution from this customer, but if the custom silicon “is available sooner for mass production, then we may be able to produce sooner.”
Broadcom’s discussion around its 10GW commitment from OpenAI likely confirms that OpenAI is Celestica’s new digital native customer, with Broadcom saying that 2026 contribution from OpenAI is expected to be minimal with the 10GW deployments concentrated in 2027 through 2029.
This would potentially be a landmark deal for Celestica in the long-run, as it is an entirely new customer with the potential to add several billion in annual revenue; management said in the original announcement that “demand from this customer at scale could achieve a level similar to those of our largest hyperscaler customers today.” For context, this would compare to Celestica’s largest hyperscaler contributing 28% of revenue in 2024, or ~$2.7 billion, implying OpenAI’s revenue could match that once it ramps.
Financials
Celestica’s financials are somewhat mixed – on one hand, the company expects strong switch demand to drive its Cloud and Connectivity Solutions (CCS) segment revenue up ~40% YoY in both 2026 and 2027, yet gross and operating margins are quite low compared to its key supplier Broadcom.
CCS Revenue to Grow ~40% Annually Through 2027
Celestica was also one of the few companies to really provide a solid long-term growth outlook in its AI-related segment this past quarter, with management confident in maintaining ~40% annual growth in CCS through 2027. While this is already reflected by consensus estimates (meaning Celestica will need substantial upside as growth expectations are already being baked in to shares), there are five main factors that could push growth above and beyond this guide – more on this below.
For 2026, Celestica guided for approximately 40% YoY growth in CCS to ~$12.6 billion, up from $9 billion guided for 2025, supported by views for accelerating 800G demand, early 1.6T ramps and the ramp of its next-gen AI compute platform to full-volume.
Driven by this growth in CCS, Celestica guided for initial revenue of $16 billion in 2026 during Q3’s report, nearly 18% ahead of the consensus estimate for $13.6 billion. This would correspond to ~31.1% YoY growth, a five point acceleration from FY25.Supporting this, management says they “currently have about 12 to 15 months of real solid forecast inputs and demand inputs from our customers,” and in many cases, visibility extends beyond that and longer into 2027. For example, some customers have a “certain amount of ASICs, for example, that they may have committed to, and it gives us some assurance as to the longevity and the size of the overall program.”
Also backing up the guidance is capacity, with Celestica explaining in Q2 that it can support “$3 billion to $4 billion of additional revenue” across its footprint in Thailand, Malaysia, Mexico and the US; Celestica is also aiming to expand production of 800G switches and add capacity for thousands of advanced AI racks annually by 2027.
For 2027, management explained that it is around 12 months too early to provide concrete numbers, but “right now, we think at least 40% [CCS growth] into 2027 is what we have visibility to,” at least 40% [CCS growth] into 2027 is what we have visibility to,” with opportunities to potentially accelerate that growth. Underscoring this strong outlook includes solid visibility into significant new program ramps starting in 2027 (multiple 1.6T ramps with hyperscalers), scale-up engagements translating to production and revenue, a next-gen custom ASIC platform, and mass-production of the rack-scale custom AI system with the new digital-native customer (likely OpenAI).
Thus, assuming ~40% growth in CCS and comments for high-single digit (6-8%) YoY growth in its Advanced Technology Solutions segment (ATS/focused on aerospace, industrial and semicap equipment), a reasonable initial estimate for Celestica’s 2027 revenue would be ~$21.3 billion. This would mark a slight two point acceleration to 33% YoY.
Five Factors Supporting Growth Accelerating Beyond 31-33%
There are five main factors that support Celestica’s revenue growth being materially faster than the initial 31% and 33% implied growth for 2026 and 2027, as each of these five factors all exhibit growth rates in excess of Celestica’s guidance.
While there is no guarantee that Celestica can match or exceed some of the growth rates in the opportunities below, these five factors provide ample evidence of market conditions that can support higher growth.
1) Ethernet Switch Demand Growing for Back-End Networking
As we had discussed in our Top 10 New Ideas report, networking is at the heart of the new architecture that Nvidia is shipping now as the increased bandwidth is instrumental in driving higher performance. The majority of growth is expected to be driven by back-end networking — scale-up and scale-out networks.
Scale-out is where the near-term growth opportunities for Ethernet switches lie, despite Ethernet adoption and revenue share historically lagging InfiniBand by a wide margin. High-bandwidth Ethernet switches are seeing strong demand in recent quarters as hyperscalers pivot away from Nvidia’s lock-in ecosystem of GPU + InfiniBand. Arista has said that momentum for Ethernet “has really shifted in the last year” while Nvidia touted that its new Spectrum-X Ethernet is annualizing at $10 billion in revenue. This is also validated by some of the largest operational GPU clusters of today, such as xAI’s Colossus, utilizing Ethernet for the back-end fabric (Nvidia’s Spectrum-X).
These proof points support bullish growth forecasts for the Ethernet switching market over the next few years. Through 2025 to 2029, the high-bandwidth Ethernet switch TAM is projected to rise at a 30% CAGR, driven by >800G rates rising at a 54% CAGR. In dollar terms, the market is expanding from ~$18 billion to nearly $50 billion over the period.
It is still quite early for the scale-up opportunity, as Broadcom and others just introduced the ESUN consortium (Ethernet for Scale-Up Networking) a few months ago. Scale-up is inherently linked to Broadcom’s 102.4T Tomahawk6 platform, which, as we had discussed in our recent newsletter, Broadcom Stock: The Silent Winner in the AI Monetization Supercycle, paves the way for >100K to 1 million accelerator clusters by allowing larger leaf-spine fabrics to be constructed, while drawing less power and keeping latency low.
Broadcom’s management points toward the flattening of the AI cluster as an important catalyst for this product, stating: “[…] Tomahawk 6 enables clusters of more than 100,000 AI accelerators to be deployed in just two tiers instead of three … this flattening of the AI cluster is huge because it enables much better performance in training next-generation frontier models through a lower latency, higher bandwidth and lower power.”
Broadcom already sees multiple >100K accelerator deployments using Tomahawk 6 for both scale-out and scale-up interconnect, with bookings at record rates. As such, Celestica sees scale-up as an “emerging multibillion-dollar new market opportunity,” having already secured some program wins for its first scale-up solutions leveraging Tomahawk6.
Main takeaway: A majority of Celestica’s switch deployments already go to back-end networking, with >800G growth expected to rise at a 54% CAGR through 2030, 14 points faster than CCS revenue. Main takeaway: A majority of Celestica’s switch deployments already go to back-end networking, with >800G growth expected to rise at a 54% CAGR through 2030, 14 points faster than CCS revenue.
2) Broadcom’s $73 Billion Backlog and AI Revenue CAGR
Broadcom provided an update on its backlog in early December with its fiscal Q4 results, saying that its total AI semiconductor backlog was >$73 billion, with AI switch backlog exceeding $10 billion. Tomahawk6 was booking at record rates, with management later clarifying that TH6 is one of the “fastest-growing products in terms of deployment that we've ever seen of any switch products.”
Broadcom expects the $73 billion backlog to be delivered over the next six quarters, and this is also expected to be a baseline, with CEO Hock Tan explaining that “we fully expect more bookings to come in over that period of time.”
Broadcom also provided a strong AI revenue guide for FQ1 of $8.2 billion, implying a 100% YoY and 26% QoQ growth, primarily driven by custom AI accelerators and Ethernet switches. For 2026 and 2027 AI revenue, BofA analysts are already laying the tracks for $50 billion and $100 billion, implying growth at a 122% CAGR if this pans out. This is notably 2X faster than Broadcom’s previously-laid-out 60% serviceable addressable market CAGR of 60%.
Analysts from RBC also believe that Broadcom’s AI semiconductor revenue acceleration “bodes well” for Celestica’s Q4 as the numbers were “’directionally positive’ for Celestica’s near-term business momentum” in CCS.
Main takeaway: Celestica’s close ties to Broadcom suggest that its accelerating XPU and networking driven momentum and backlog growth could drive a stronger acceleration for CLS.Main takeaway: Celestica’s close ties to Broadcom suggest that its accelerating XPU and networking driven momentum and backlog growth could drive a stronger acceleration for CLS.
3) ASICs CAGR Forecasts
The ASICs market is forecast to see substantial growth over the next few years, as a handful of major hyperscalers pursue ASICs-based AI platform roadmaps. Marvell has projected the ASICs market to rise at a 47% CAGR from 2023 through 2028, rising from $6 billion to $40.8 billion, or nearly 7X growth in five years.
On the other hand, 650 Group has a slightly more bullish forecast, projecting the ASICs TAM to rise at a 54% CAGR from $18 billion in 2025 to $104 billion in 2029, or a roughly 6X increase over the next four years. Broadcom’s rising backlog and increasingly large ASICs orders — $10 billion and $11 billion with Anthropic and the 10GW commitment from OpenAI – support strong growth as Google and Meta continue to build expand their ASICs platforms.
Main takeaway: Rising demand for ASICs over the next few years can drive stronger growth for Celestica as its solutions are almost exclusively focused on ASICs platforms. Main takeaway: Rising demand for ASICs over the next few years can drive stronger growth for Celestica as its solutions are almost exclusively focused on ASICs platforms.
4) Google’s TPU Shipments Accelerating
While the market continues to debate the TPU vs GPU story, there are some reports from analysts that see TPU volumes accelerating over the next few years. For example, Morgan Stanley projects Google’s TPU shipments to be 1.75 million in 2025, with initial contribution of ~0.5 million from TPU v7 Ironwood.
For 2026, the firm projected Ironwood shipments to rise 5X to 2.5 million, driving total TPU shipments up ~83% YoY to 3.2 million. For 2027, Morgan Stanley boosted its shipment forecast by ~67%, from 3 million to 5 million, driven mostly by TPU v8 and future generations, while its 2028 forecast was boosted 120% from ~3.2 million to 7 million. This would imply 2027 and 2028 growth of ~56% and ~40%, up from (6%) and 19% previously.
There are some unsubstantiated claims that Celestica could generate $500 million in revenue per 1 million TPUs shipped, and while this is unverified, if Celestica can capture that amount ($500) per chip, the TPU linked opportunity could be increasingly large over the next few years.
Main takeaway: Celestica could see solid revenue tailwinds linked to Google’s TPUs if shipments accelerate per some analyst estimates, assuming it remains engaged on the platform.Main takeaway: Celestica could see solid revenue tailwinds linked to Google’s TPUs if shipments accelerate per some analyst estimates, assuming it remains engaged on the platform.
5) Meta’s Capex
Meta’s capex strategy is to “aggressively front-load building capacity” to prepare for the most optimistic cases on when AI superintelligence will arrive, with the company outlining substantial capex growth in 2026. Meta aims to meet these capacity needs by “both building our own infrastructure and contracting with third party cloud providers.”
As a result, Meta expects capex dollar growth to be “notably larger in 2026 than 2025,” implying 2026 capex of at least $103 billion, as current guidance for 2025 at $70-72 billion implies a minimum of ~$32 billion YoY growth. However, considering management’s comments for notably larger dollar growth, there is potential for capex to come in at or above $110 billion, up ~55% YoY, above current estimates for $107.9 billion.
Main takeaway: Meta’s capex is expected to grow a minimum of >45% in 2026 as the company spends heavily on AI data center infrastructure, potentially driving faster growth for Celestica as it has worked closely with Meta on prior products (and potentially the upcoming next-gen AI rack ramp in Q4). Main takeaway: Meta’s capex is expected to grow a minimum of >45% in 2026 as the company spends heavily on AI data center infrastructure, potentially driving faster growth for Celestica as it has worked closely with Meta on prior products (and potentially the upcoming next-gen AI rack ramp in Q4).
Celestica’s CCS segment is broken down further into two other subsegments – Communications (networking) and Enterprise (servers and storage). This breakdown highlights a key risk moving to Q4, as guidance implies Communications revenue will decelerate sharply QoQ, an odd print considering the strong ramp in switching products.
In Q3, revenue from CCS segment rose 43% YoY to $2.41 billion, driven by an 82% YoY increase in Communications revenue to $1.94 billion offsetting a (24%) decline in Enterprise to $477 million on an AI program transition with a hyperscaler.
For Q4, CCS revenue is implied to accelerate nine points to 52% YoY, with Communications growth guided in the high-60s YoY and Enterprise guided in the low-20s as the new AI program is set to ramp. Despite the seemingly strong guide in Communications, QoQ growth would be just 1% QoQ, a sharp deceleration from Q3’s 18% QoQ growth.
This would mark Communications’ lowest sequential growth in the last two years, and its first time reporting single-digit sequential growth in the last seven quarters, raising a potential red flag considering Communications is primarily driven by networking/800G switches.
A likely explanation of this could be the strong outperformance in Communications in Q3 – guidance was for low-60s YoY growth, which Celestica beat by ~20 points. As a result, QoQ growth was likely expected to be ~4%, but came in at 18%, possibly representing a much stronger-than-expected ramp of 800G platforms in the quarter.
Margins
Margins continued to expand in Q3, with some signs of operating leverage arising from strong Communications growth as operating margin expanded by 4.7 points YoY versus a 2.6 point YoY expansion for gross margin.
GAAP gross margin was 13.0% in Q3, up 0.2 points QoQ and 2.6 points YoY.
GAAP operating margin of 10.2%, up 0.8 points QoQ and 4.7 points YoY. Adjusted operating margin was 7.6%, up 0.2 points QoQ and 0.8 points YoY.
GAAP net margin of 8.4%, up 1.1 points QoQ and 4.8 points YoY. However, adjusted net margin was just 5.7%, up just 0.1 points QoQ and 0.7 points YoY due to a $113 million impact from gains on total return swaps.
For a segment breakdown:
CCS adjusted gross margin was 12.1% in Q3, up from 11.9% a year ago. CCS adjusted operating margin was 8.3%, up from 7.6% a year ago. Positioning as an ODM may not find much more margin upside even as higher-margin products such as advanced AI rack systems ramp.
ATS adjusted gross margin was 10.6%, up from 8.4% a year ago. ATS adjusted operating margin was 5.5%, up from 4.9% a year ago.
For fiscal 2025, Celestica guided for adjusted operating margin to be 7.4%, and for 2026, only a slight 0.4 point increase to 7.8% despite the 31% growth on the top-line. This suggests that its positioning primarily as an ODM may limit future upside to operating margins even as hyperscaler-linked revenue and higher-margin, higher-complexity designs increase in its mix:
“So we continue to see the benefits of both operating leverage as well as positive mix in our numbers, on track for about 7.4% operating margin at the company level for 2025, and we're guiding that, that can expand now going into 2026. We do continue to believe that there's opportunities for even more margin expansion. But again, I'm not giving formal numbers for '27 at this point.
When you look at our ATS business, the business has done very well on doing some selective pruning in order to really focus on the highest value engagement. And so we're really happy with the margin expansion that we've seen in ATS already. And we think that there's opportunities to continue to expand and get it above 6%, hopefully in the near to medium term.
On the CCS side, which is operating in the low 8s right now, what's working to our favor is the fact that we will continue to be seeing growth in networking, which are primarily our HPS products. And our HPS products are accretive to the company and accretive to CCS. And so as we see growth in that area, we will continue to see some margin upside.
That being said, we do continue to evaluate how we can support our customers on multiple areas such as doing complex rack integration work. And so sometimes that will be margin dilutive.”
Earnings
Celestica reported GAAP EPS of $2.31 in Q3, beating the $1.38 estimate by 67.6%. Adjusted EPS was $1.58, beating the $1.49 estimate by just 6% and representing growth of 52% YoY.
For Q4, Celestica guided adjusted EPS to be in the range of $1.65 to $1.81, which, at the $1.73 midpoint, is only marginally ahead of estimates for $1.71. This also corresponds to a slight acceleration to 55.9% growth. Looking ahead to Q1 and Q2, estimates point to 52.3% growth and 41.5% growth, decelerating in both quarters, likely driven by margin expansion slowing.
For fiscal 2025, Celestica boosted its adjusted EPS outlook by 7.3% to $5.90, from its previous forecast for $5.50 and pointing to 51% YoY growth. For fiscal 2026, Celestica outlined an initial guide for $8.20 in adjusted EPS, up 39% YoY and well ahead of estimates for $7.22.
Cash
On the other hand, cash flows are rather thin and fell to the lowest level in a year.
Operating cash flow was $126.2 million, or a 4% margin, down from 5.3% in Q2 and 4.9% in the year ago quarter. OCF growth was just 2.4% YoY and was also the lowest cash flow since the year ago quarter.
Adjusted FCF was $89 million for a 2.8% margin in Q3, up 15.6% YoY but also the lowest level since the year ago quarter. Adjusted FCF margin was down from 4.1% in Q2 and 3% in the year ago quarter.
For fiscal 2025, Celestica raised its adjusted FCF guidance slightly to $425 million, from $400 million prior, for a 3.5% margin, while capex is guided to $200 million, or 1.6% of revenue. Fiscal 2026 adjusted FCF was guided at $500 million, up 18% YoY and for a 3.1% margin, with the margin decline driven by higher capex, guided to rise 50-100% YoY to $300 to $400 million, or 2.2-2.5% of revenue.
Cash and equivalents totaled $305 million while debt totaled $728 million in term loans. Including an undrawn revolver, total liquidity is approximately $1.1 billion. Celestica’s gross debt to TTM adjusted EBITDA was 0.8x, improving by 0.1 points sequentially and 0.3 points from last year.
Inventories were $2.05 billion, up nearly 7% QoQ, while accounts receivable totaled $2.44 billion, also up nearly 7% QoQ.
Valuation
Celestica is trading just off peak multiples on the top and bottom line following this recent pullback after Broadcom’s earnings. Celestica’s forward PS is currently at 2.5x, well above the five-year average of 0.75x and 25% above the 2x multiple it commanded at the start of September. Even on the fiscal 2026 guide, shares are at a 1.9x multiple.
On a forward PE basis, shares are trading at 45.7x fiscal 2025 adjusted EPS and 32.3x fiscal 2026, well above its five-year average forward PE of 15.4x and prior resistance at 25x in late 2024 and early 2025. The company has been seeing a re-rating higher as it captures AI-growth tailwinds, but any hint of softness in growth could easily see Celestica re-rated lower given growth through 2027 is visible and may already be priced in.
Conclusion
Celestica is benefiting from strong market demand for 800G switches with its Tomahawk6-based 1.6T switches on deck for availability later in 2026. The company guided for an impressive five-point acceleration in 2026 during Q3’s report, outlining more than 31% growth to $16 billion in revenue. 2027 was implied to accelerate slightly to around 33% YoY to surpass $21 billion in revenue, again on strong switch demand, the ramp of 1.6T programs, and a new custom rack-scale solution with a digital native customer entering the picture, likely OpenAI.
However, Celestica’s valuation remains quite stretched, with the company sitting well above its five-year average multiples on the top and bottom line as shares are being re-rated higher for its visible topline acceleration and strong 40% growth momentum in CCS – this extended valuation will need to be watched considering growth expectations could be getting priced in already given the high level of visibility into 2027.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
