Nvidia’s earnings report can best be described by a Shakespeare line in Henry IV: “Heavy is the head that wears the crown.”  When it comes to stocks, being on top is harder than it looks. You are no longer afforded the element of surprise, and particularly important for Nvidia, you must produce new catalysts that can contend with the previous, hard-hitting catalyst that drove the company’s historic growth in previous years. 

The catalysts on the horizon are imperfect, whether it’s custom silicon gaining traction as inference scales or the attach rate on CPUs-to-GPUs shifting as agentic AI requires more orchestration. Nvidia is also seeing heightened supply-related commitments, likely due to steep HBM and NAND pricing. The company is also changed its reporting segments, and below, I discuss why that raises concerns. 

However, on the positive side, Nvidia is a beast on the bottom line. Apple is firmly in the rear-view mirror when comparing profits and cash flows. Nvidia’s operating income is north of $50 billion compared to Apple’s $36 billion, plus cash flows that are nearly 2X Apple’s at $48.6B versus $28.7B. 

To be picky about its growth status, there is a deceleration in QoQ growth as the guide is for 11.5% QoQ growth compared to the previous three quarters, which all reported 19.5% to 22% QoQ growth. If we look a bit further ahead, analyst estimates are calling for flat-ish QoQ growth in the September quarter and then a leveling off to 7% and 6% QoQ growth. All of that will clear up when Rubin ships in volume, although as you’ll see below, the company left an out for themselves on the timing. 

I always enjoy covering Nvidia, because although the stock is well-covered, I think the I/O Fund continues to surface key details you don’t typically hear elsewhere. My strong Nvidia streak is slowing somewhat as I turn my focus to other opportunities across the AI trade, yet understanding the juggernaut is non-negotiable for all AI investors. 

Below, I take a closer look at what was communicated last night. 

The AI Demand Signal is Extraordinary 

As you’ll see below, Nvidia’s positioning is becoming more challenged by both custom silicon and CPUs. However, before we go into those details, the 10,000-foot view is fairly clear – which is that AI demand is parabolic. Collette Kress, the CFO, pointed out in the opening remarks that analysts expect hyperscale capex to exceed $1 trillion in 2027 with AI infrastructure reaching $3T to $4T by the end of the decade.  

This year, analysts are expecting capex to grow between 90% and 100% with Nvidia exceeding this growth rate with data center growth of 120%. As pointed out on the call, this is because Nvidia serves two major customer groups; the first being well-known hyperscalers and the second being neoclouds and enterprises. Whether it’s Nvidia specifically or the AI infrastructure market more broadly, the point Huang made on the earnings call is that far more than just 7 companies will support this market over the next few years: 

“The second category is all of the AI native clouds. They're regional, they're all over the place, they're start-ups all over the world, supporting those companies. They're enterprise, 250,000 enterprise companies around the world, many of them will have to build or want to build AI factories for themselves to operate. Many industrial companies, there's no choice but to put the computer where the context is, where the action is, you can't put that in the cloud. It has to respond reliably, quickly every single time, can't imagine a chip plant, a chip fab being connected to a cloud service provider, doesn't make any sense. And so the second category and the sovereign AI clouds. And so there's a whole category of data centers that semi-custom chips just don't apply because these data centers want to buy systems, they want to operate systems, they don't want to design, they don't want to build it themselves.” 

The point being made is two-fold. On one hand, it helps investors to see the diversity of customer base driving the AI market. On the other hand, it’s self-serving as Nvidia is likely preparing the market for a time when hyperscaler capex is more concentrated in custom silicon. We covered this in the free article recently, stating: “Counterpoint Research believes that by 2028, custom silicon will cross the 15-million mark to surpass GPU shipments as the top 10 hyperscalers will have deployed 40 million AI server compute ASIC chips cumulatively during 2024-2028.” 

Nvidia Changes Segment Reporting to Breakout Hyperscalers from Neoclouds/Enterprises 

If we operate under the assumption that custom silicon will become a fierce contender to GPUs within hyperscaler budgets, then the appropriate defensive move for Nvidia is to breakout the line item that shows they have a diverse set of customers.  

Pictured above: NVIDIA is transitioning to a new reporting framework that better reflects its current and future growth drivers. NVIDIA will have two market platforms — Data Center and Edge Computing. Within Data Center, NVIDIA will report two sub-markets, Hyperscale and ACIE, which incorporates AI Clouds, Industrial and Enterprise. 

The strong revenue trajectory in ACIE is communicating that Nvidia has other avenues for growth should we see a reduction of GPU-related capex spending. As stated above, Nvidia believes there is about 250,000 companies as the potential SAM for that segment, with the following granularity offered: “The second segment is AI natives, enterprise on-prems, industrial on-prems and that — and sovereign AI. That segment is growing incredibly fast because everybody needs AI, and we're going to see AI being adopted by every industry, every country, every company. And so everybody wants to build it in a different way. And the fact that we provide the entire solution, it makes it much easier, makes it possible at all for people to be able to build these things. And then, of course, the robotic edge today.” 

Overall, investors should be prepared for an expanding AI market to meet decreasing market share for Nvidia during the inference phase. Which brings me to my next two points. 

Nvidia’s Market Share in Question for the Inference Market 

Over the years, I’ve grown to have a keen ear for the commentary on earnings calls, and I do believe there was a mis-step last night with one Q&A exchange.

An analyst asked Huang: “How do you see Vera Rubin in your extreme co-engineering impacting your share of the inference market as we look into late '26, '27?”

In my opinion, management did not directly address the question. Instead, it redirected toward Nvidia’s growth in inference deployments, while mixing up an important distinction between growth versus market share.  

Of course, Nvidia is growing inference revenue and inference capacity. In the excerpt below, companies like Anthropic, Azure, AWS and CoreWeave are cited as evidence of this growing “share.” That wasn’t the question though. The question the whole market is wondering is whether Nvidia is gaining market share relative to the inference market.  

As you know, the market’s biggest concern is whether custom silicon will take a larger share of inference as workloads become more repetitive and cost-sensitive. The answer did not resolve that concern, and to me, it appeared the word “share” was being redirected to descriptions around “growth.”  

For example, Anthropic is deploying gigawatt-level inference workloads with custom silicon providers like Amazon/Trainium and Google/TPUs, therefore, to cite that Anthropic was “largely 0 until recently” does not translate to gaining share on Anthropic workloads as it actually means Nvidia is under-indexed or lagging on this very large inference customer. That’s one of a few inconsistencies in this critical Q&A exchange, with the main one being interchanging the words “share” with “growth” 

“Jen-Hsun Huang, Co-Founder, CEO & Director: 

Well, we are growing share in inference, and we're growing share in inference very, very quickly. And the reason for that is this year, the number of frontier model companies grew. And so there's Cursor and Perplexity and there's some new model companies, TML and Reflection, and the list goes on. And so the number of frontier model companies has grown, and we added Anthropic to our partnership this year. They're expanding incredibly fast. We've partnered with them to secure computing capacity across Azure, AWS, CoreWeave. I forget who else we've already announced, but there's a whole list of others that we are bringing online for them. And so the amount of capacity that we're going to bring online for Anthropic this year and next year is going to be quite significant, very significant.   

And so we're growing and our coverage of Anthropic has been largely 0 until just recently. And so we're gaining share tremendously fast in inference. Vera Rubin is going to be even more successful than Grace Blackwell at this point. Every single, I can't think of one. Every single frontier model company will jump on Vera Rubin from the get-go, and that wasn't true before on Blackwell. And so Vera Rubin is off to a tremendous start and it will surely be more successful than even Grace Blackwell.   

So I think the end of your answer, C.J., is that we're gaining share in inference. Let me go back again to the question that Ben was asking. Remember, so far, everything that I've just explained in the inference question is really focused on hyperscale. Remember, there's a whole second category of AI data centers that we serve almost uniquely. Now this segment is very fragmented, requires a fairly integrated — a really well-integrated platform solution and a very large go-to-market. And that segment, all of the inference, 100% of that — the vast majority of that is NVIDIA.” 

CPU-to-GPU Attach Rate is Increasing; What that Means for GPUs 

We covered the rising importance of CPUs in a recent analysis on Arm, stating: “In agentic workflows, the GPU still handles inference, but between each inference call, the CPU is doing the orchestration – which are best described as handling tool calls, API requests and memory tasks. AI agents are surfacing this new constraint, which is how to prevent latency and underutilized GPUs following the exponential growth of orchestration needs. 

For investors, what matters is that CPUs account for 50% to 90% of total latency in workflows, which means the CPU-to-GPU ratio in AI clusters will need to increase. Earlier this year, both AMD and Intel saw analyst upgrades based on the outstripped supply of CPUs leading to higher average sales prices of roughly 10% to 15%. Reuters also reported that Intel’s unfulfilled orders are reaching longer than six months while AMD delivery times are believed to be eight to 10 weeks.” 

According to TrendForce and commentary from Arm, it’s expected the CPU-to-GPU attach rate increases from 1:8 to a ratio of 1:2 or even 1:1. 

This helps explain why Nvidia spent a decent amount of time last night focusing on its Vera CPU designs: “Vera CPU opens a brand-new $200 billion TAM for NVIDIA, a market we have never addressed before, and every major hyperscale and system maker is partnering with us to get it deployed. We have visibility to nearly $20 billion in total CPU revenue this year, setting us up to become the world's leading CPU supplier.” 

However, this inevitably raises the question as to what this means for GPUs (i.e., will more AI compute spend be diverted from largely being GPUs to now include higher CPU content). One analyst went so far as to state it could cannibalize GPUs, of which Nvidia’s management team pushed back on.  The answer was long-winded so I am keeping only the excerpts that pertain to the concern.  

For more information on the topic, you can read our AMD post-earnings analysis here and Arm post-earnings analysis here.  

Vivek Arya 
BofA Securities, Research Division 

Jensen, there's a lot of excitement around CPU for agentic applications and just a lot of noise around the number of CPUs actually exceeding the number of GPUs. And I was just hoping that you could kind of give your perspective that, first of all, is this an incremental workload? Is this kind of cannibalizing what the GPU would have done otherwise? And then secondly, the $20 billion number that you gave, is that for stand-alone Vera CPUs? Or is that kind of already included in that Vera, as part of Vera Rubin? So just if you could educate us on the role of CPU versus GPU, is it cannibalistic? Is it incremental? And then the $20 billion number, how to kind of put that in context with what you sell, right, which is usually the CPU as part of the GPU? 

Jen-Hsun Huang 
Co-Founder, CEO & Director 

The $20 billion is for stand-alone CPU. And remember, we have Vera, is used in 3 ways. As a stand-alone — 4 ways — let me just start with the one that you already know. The first way is Vera Rubin. And we'll sell millions of Rubins, and every 2 of them is connected to a Vera. And of course, we price those 2 and they're properly priced. And so that's #1 use case.  

The second use case is Vera stand-alone CPU. The third is Vera with CX-9 and the software stack for storage. And then Vera in a — with CX-9 with a software stack for security and compute isolation and confidential computing. Okay, so each one of those use cases is built on Vera. And my sense is that we'll be supply constrained throughout the entire life of Vera Rubin. There are 4 different use cases of it. And — but anyhow, the answer to your question is — of the $20 billion is a stand-alone [,,,] And so — but the large length, every one of those agents are going to spin off subagents. And every time they spin these off, you're going to need to do inference. That's where the thinking happens. All of the thinking happens on GPUs, all of the orchestration essentially runs on CPUs. And the subagents when they're spun off, they — when they're thinking they use GPUs.  

[…] So we're going to need a lot more CPUs, and Vera was designed to be an agentic CPU. The CPUs of the past were designed to have many cores so that it could be easily rentable. People rented cores. Well, agents don't rent cores. They just want the work to be done fast. The economics of the past was dollars per core. That's the economics of cloud computing of the past. The economics of the AI of the future is tokens per dollar or dollars per token. And so what we need to do in the future is to generate tokens, process tokens as fast as possible, and that's what Vera does incredibly well […] 

Supply-Related Commitments Surge to $119B; Good or Bad? 

Nvidia’s total supply-related commitments surged once again in Q1, as Nvidia continues to secure supply and capacity to meet demand, yet there may also be a hidden signal that this is driven by materially higher memory component costs that could weigh on margins. 

Total supply commitments reached $119 billion, up nearly $90 billion YoY and $24 billion QoQ. As stated last quarter, we believe this serves as a key sign that the current accelerated QoQ data center growth will persist as Blackwell and Rubin ramp, as Nvidia is putting the pieces together across the supply chain to meet its $1 trillion in forecasted cumulative revenue through 2027.  

As seen above, this is the largest two-quarter step-up in supply commitments Nvidia has seen at nearly $70 billion, and as it stands, this also is more than 2X its reported cash and equivalents, the first time exceeding this level since Hopper’s breakout quarter. Supply commitments are also substantially higher heading into Rubin’s ramp than prior generations – early FY24 ramped into the teens, before stepping up to the ~$30 billion level for Blackwell.

Nvidia expects $95 billion of these commitments to be paid in the remainder of FY27, and the sheer increase over the past two quarters could imply that there may be some margin headwinds with the Rubin ramp if the bulk of these commitments stems from memory costs.  

Considering that Blackwell Ultra and Rubin contain 60% more HBM content versus Blackwell and with memory prices up ~6X since September, it’s entirely possible that securing HBM and auxiliary memory account for the bulk of this increase. For example, Morgan Stanley estimates that Nvidia’s bill of materials on memory for Rubin has reached $2 million per rack, up 435% from the GB300’s $374,000. Putting this a different way, memory could account for 25% of the total BOM for Rubin, versus <10% for the GB300; when translating this to a $500 billion SKU, this is quite a substantial uplift in memory costs that Nvidia must offset via higher prices to avoid operating margin contraction. 

It’s clear that supply-related commitments are surging above and beyond what is normal for previous GPU generations – which could indicate either a very strong pipeline or incoming margin pressure from higher memory costs/commitments. 

Rubin Remains the Next Major Catalyst, But Timing Risk Remains

There were mixed signals provided on Rubin’s timing. The headline statements seemed to confirm shipments would begin in Q3 (if so … no biggie), but then statements during the Q&A section seemed to point toward the stronger ramp not occurring until Q4-Q1. 

Overall, the commentary left it open on when Rubin will make an impact, with my takeaway being somewhere between Q4 and Q1. Keep in mind that Q4 is end of January for Nvidia, so it could be about 8 months out before there is any material Rubin revenue and about 11 months before a bigger impact. 

“Joshua Buchalter 
TD Cowen, Research Division 

And congrats on the great results. Colette, I believe, in your prepared remarks, you mentioned GB300 is sort of the fastest ramp in the company's history. How should we think about Vera Rubin against this benchmark? It's obviously a new architecture at the silicon level, but in similar rack. Does that mean we should expect a similar slope to the Vera Rubin ramp as the GB300? Or should it be a bit more gradual given the new silicon? 

Colette Kress 
Executive VP & CFO 

Yes. Well, we've indicated for a while that we will be launching Vera Rubin in the second half. We will start in Q3. That will be our initial pieces together. And then once we get to Q4, we're probably going to start to see our ramping continue. It's hard to say at this point what will be a faster ramp. But again, we have demand already planned, we've got POs. We've got almost all of our major customers ready to go, and these are very complex systems that we need to put together. So I think it's just about the timing that it's going to take for us to get that into market. Nothing else other than getting from production of all of the different systems that we have ready for order.  

So a little early to say. But yes, we're going to start in Q3 and continue to ramp into Q4. And Q1 of next year certainly is going to be very big as well.” 

Financials 

Revenue Accelerates 12 Points in Q1, Guided to Persist in Q2 

Nvidia reported $81.62 billion in revenue in Q1, beating its own guidance for $78 billion and marking a fresh record for sequential dollar growth at nearly $13.5 billion (versus $11.1 billion last quarter).  

Revenue growth accelerated 12 points from 73.2% YoY in Q4 to 85.2% YoY in Q1, while QoQ growth was steady at 19.8% QoQ, an impressive growth rate considering the sheer scale of Nvidia’s revenue.

For Q2, Nvidia guided for revenue to be $91 billion, +/- 2%, implying YoY growth accelerating further to 94.7% while QoQ growth would moderate to 11.9%. However, dollar growth would remain rather strong sequentially at $9.4 billion guided. This was notably $4 billion ahead of consensus for $86.95 billion.  

For FY27, current consensus estimates sat at $373 billion (up 72.7% YoY) heading into earnings, $43 billion higher than the $330 billion estimate from late February due to Nvidia’s comments about $1 trillion in cumulative revenue for Blackwell and Rubin through 2027. However, considering Q1’s beat and Q2’s raise over estimates, it’s likely that FY27 revenue estimates will have to move a minimum of $10 billion higher.  

Networking Remains Robust at 35% QoQ to Nearly $60B Annualized 

As expected, Data Center momentum remained robust, with revenue up 92% YoY and 21% QoQ to $75.25 billion. This marked a 17 point acceleration from 75% YoY growth in Q4 while QoQ again remained steady with Q4’s 22% growth off a larger base. Nvidia said that growth was driven by the GB300 ramp as well as demand across its Networking portfolio, including InfiniBand, Spectrum-X Ethernet and NVLink. 

Compute revenue was $60.4 billion, accelerating 19 point to 77% YoY with QoQ growth of 18%, roughly maintaining the 19% QoQ growth from Q4. On a dollar basis, growth was $9.1 billion, increasing from Q4’s ~$8.3 billion. Nvidia added that it recorded no China-based Hopper revenue in the quarter. 

Networking growth remained robust, up 199% YoY and 35% QoQ to a record $14.8 billion, or nearly $60 billion annualized, compared to $20 billion annualized last Q1. While YoY growth did technically decelerate 36 points from 235% YoY in Q4, the more impressive feat was the slight QoQ acceleration from 34% in Q4 to 35% QoQ this quarter.  

New Reporting Structure for Key Segments 

It should be noted that Nvidia shook up its segment reporting this quarter, re-categorizing Data Center to two sub-markets: Hyperscale and AI Cloud, Industrial and Enterprise (ACIE), to emphasize what customer cohorts are driving growth. While Nvidia did provide Compute and Networking revenue this quarter, it’s unlikely that we will get another breakdown here moving forward.  

Nvidia’s other segments – Gaming, Automotive, Pro Viz, and OEM and Other – were reclassified into Edge Computing.  

For a quick snapshot of the new segment structure: 

Hyperscale revenue accounted for roughly 50% of Data Center at $37.87 billion, up 115% YoY and 12% QoQ. Revenue from Hyperscale was $17.6 billion a year ago (45% of DC) and $33.8 billion in Q4 (54% of DC). 

AI Cloud, Industrial and Enterprise (ACIE) revenue was the remaining half of Data Center at $37.38 billion, up 74% YoY and 31% QoQ. ACIE revenue was $21.5 billion a year ago and $28.5 billion in Q4. 

Edge Computing revenue was $6.37 billion, up 29% YoY and 10% QoQ, driven by strong demand for Blackwell workstations, offset by slower consumer PC demand.  

Margins Remaining Steady 

While Nvidia continues to grow its topline at increasingly large rates on a dollar basis, margins are remaining steady. There were also tiny signs of operating leverage at this scale, with gross margins in line with guidance and slight outperformance on operating margins. 

GAAP gross margin was 74.9% and adjusted gross margin was 75%, both in line with guidance. Both were up >14 points YoY due to the H20 impacts last Q1, and marginally lower QoQ. For Q2, Nvidia guided for both to be flat QoQ at 74.9% and 75% respectively, representing roughly 2.5 and 2.3 points of expansion YoY.  

GAAP operating margin was 65.6%, coming in above guidance for 65%; this marked a >16 point YoY expansion again from the H20-related impacts, and a slight increase from 65% in Q4. Adjusted operating margin was 65.9% and saw a similar dynamic, up >13 points YoY and expanding from 65.3% in Q4.  

Looking ahead to Q2, guidance implies operating margins to remain flat QoQ at 65.6% and 65.9% respectively. On a YoY basis, this would represent a 4.8 point expansion for GAAP operating margin and a smaller 1.4 point expansion for adjusted operating margin.  

GAAP net margin was 71.5%, as Nvidia benefitted from nearly $16 billion in gains related to its equity investments, more than offsetting its $11.6 billion in income tax payments this quarter. Adjusted net margin was 55.8%, up more than 10 points YoY but down 1.4 points QoQ. 

EPS  

Nvidia’s GAAP EPS benefitted from the equity investment gains, though growth for adjusted EPS was also robust at 140% YoY.  

GAAP EPS was $2.39, up 214% YoY due to the equity gains, which contributed roughly $0.64 to the bottom line. Adjusted EPS was $1.87, up 140% YoY (versus Q1’s new adjusted figure of $0.78, per Q4’s change in reporting to include SBC).  

For Q2, GAAP EPS is projected to be $1.91, up 76.9% YoY, while adjusted EPS is projected to be $1.96, up 86.6% YoY. 

Cash Flows and Balance Sheet 

Cash flows were another strong point in Q1 as operating cash flow margin returned to above 60%.  

Q1 operating cash flow margin was $50.3 billion for a 61.7% margin, down from a 62.2% margin a year ago but a rebound from 53.1% in Q4. Nvidia says OCF was driven by higher revenue and lower cash taxes, projecting higher taxes in Q2 which is likely to weigh on OCF.  

Q1 free cash flow was $48.6 billion for a 59.5% margin, up slightly from 59.3% a year ago and 51.2% in Q4. 

Cash, equivalents and marketable debt securities were $50.3 billion (excluding marketable equity securities which were previously included in Q4). Debt remained steady at $8.47 billion.  

Inventories were $25.8 billion, up $4.4 billion or 20.6% QoQ, while accounts receivable increased more than $2 billion QoQ to $40.7 billion.  

Conclusion: 

Nvidia delivered a near perfect quarter, as revenue accelerated, networking grew 35% QoQ, with elite-level margins and cash flow that are significantly better than even the trademark value-stock Apple.  

With that said, the catalysts are not as clean as prior years during Hopper and Blackwell. The inference market is becoming more competitive, CPU-to-GPU attach rates could divert compute spend, and Nvidia’s supply-related commitments are surging – which could indicate either a very strong pipeline or incoming margin pressure from higher memory costs/commitments. Lastly, the segment change is likely a defensive move ahead of hyperscaler allocating more AI budget to custom silicon.  

Although I am far from bearish on Nvidia, the I/O Fund is a top tier team in AI research. We can do better than hold the most well-known name in the AI trade. As we close up our earnings season soon following Broadcom, we turn our attention to new ideas for a dedicated seven weeks. Keep an eye on your inbox as we revisit the biggest winners from this quarter and surface new stocks you likely haven’t heard of.

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 NVDA at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

• Applied Optoelectronics Q1: Management Guides to 141% YoY Growth; Execution Comes Next
• Arm FQ4: AGI CPU Demand Hits $2B, Revenue Outlook Stays at $1B
• Lumentum FQ3: Firing on All Cylinders Despite Stiff Supply Constraints Across EMLs, Pump Lasers
• Astera Labs: Important QoQ Acceleration, Product Road Map is Loaded

MaxLinear is another under-the-radar optical networking beneficiary, supplying a range of components within optical transceivers, with its most notable being optical DSPs for 400G, 800G and soon 1.6T solutions.  

The company is seeing strong demand emerge for its Keystone DSP family, with management raising its 2026 revenue forecast for Keystone by >40% already, from $100-130 million to $150-170 million. Impressively, this growth is being driven by 400G and 800G modules as the 1.6T product, Rushmore, has yet to ramp. 

The pivot upstream to ramping more heavily on 800G and 1.6T is expected to benefit both revenues and margins from higher ASPs, but it also opens the door for a prolonged revenue runway as 1.6T growth is expected to maintain strong through 2027 with attach rates increasing with larger GPU systems.  

Fundamentally, margins have remained pressured and cash flows are quite thin, but there are some green shoots emerging as MaxLinear is forecasting a return to GAAP operating profitability in Q2 for the first time in three years.  

Brief Product Overview 

MaxLinear supplies a range of key components within optical transceivers, AECs and other connectivity solutions, such as PAM4 DSPs, transimpedance amplifiers (TIAs) and storage accelerator SoCs. While AI data centers (Infrastructure) are rapidly becoming MaxLinear’s main growth outlet, it also serves broadband, wireless, automotive and industrial end markets.   

• Keystone Optical DSPs 

MaxLinear’s Keystone family spans twelve PAM4 DSPs, including both 400G to 800G with integrated drivers for EML/SiPho transceivers, bare die options, and driverless DSPs. MaxLinear also has introduced variants of Keystone to target both AOCs and AECs as well.  

The company says the third-gen DSPs offer ‘best-in-class’ power consumption, enabling 7W 400G optical modules and 13W 800G designs. The Keystone DSPs can also be paired with MaxLinear’s Topanga and Washington TIAs to provide complete solutions for customers.  

Keystone was stated this quarter to be ramping at multiple major hyperscale customers across the US and Asia for both 400G and 800G scale-up and scale-out applications. It also underpins MaxLinear’s optical data center momentum this year into 2027, with management seeing DSPs as their outright #1 TAM.  

Keystone is expected to drive the near-term growth, per the opening remarks: “We also expect a step function data center revenue increase beginning in Q2 with expected strong upside as run rates expand into 2027. At the center of this data center momentum is our Keystone PAM4 DSP optical transceiver platform. Keystone is now ramping at multiple major hyperscale customers across both the U.S. and Asia, supporting 400G and 800G — 800G PAM4 deployments for scale-up and scale-out applications.” 

• Rushmore 200G PAM4 DSP 

MaxLinear unveiled Rushmore last year, its 200G/lane PAM4 SerDes and DPS supporting 1.6T optics and active copper cable deployments. Rushmore is compatible with both Ethernet and InfiniBand, and will enable <25W 1.6T optical interconnects with low latency, and optimized performance across all laser technologies from EMLs to SiPho. MaxLinear says that when paired with its Washington TIA, Rushmore will “provide the best performance on the market.” 

MaxLinear added in Q1 that customer engagement for Rushmore is accelerating faster than expected, with production ramps expected in late 2026 with strong revenue growth continuing into 2027.  

Here is what was stated in the opening remarks: “So moving forward to 1.6T, the critical thing to keep in mind is that there is enormous confidence out there. We're shipping Keystone into major data centers today, and they're ramping very strongly in 2026. And we have now rolled out our 1.6 Terabit Rushmore product and Annapurna family for electrical applications.  

And I think that this level of execution a part and the success with the cloud relationships, module partnerships and the call and interop completion is creating a far more pull for 1.6T participation than I would have guessed at this point in time. So in a sense, we hope that by the end of the year, we'll have called on 1.6T and start transitioning […]” 

• Washington 200G TIA 

Transimpedance amplifiers (TIAs) are current-to-voltage converters, amplifying the current generated by photodetectors within optical modules to voltage signals that can be processed by DSPs. For high-speed optics such as 1.6T transceivers, TIAs play an important role in helping maintain signal integrity at faster rates. 

MaxLinear recently unveiled its four lane/200G Washington TIA at the end of April, designed for 1.6T transceivers and capable of supporting fully retimed, half-retimed, and linear interfaces, from LRO/LPO, NPO and CPO applications.  

Washington interoperates with PAM4 DSPs from all major DSP vendors, simplifying integration into existing system architectures, while also complementing MaxLinear’s portfolio of DSPs and drivers to allow customers to build and customize systems for their specific power, performance and reach requirements. 

As discussed below, the ability for the TIA to be a component in any architecture is key: “But honestly, I mean, the TIA is beyond the TIA for Rushmore, right? If you think of an LPO strategy that the TIA is a fundamental block. If you think about LRO strategy, the TA is a fundamental block.” 

• Annapurna scale-up retimer 

Annapurna is MaxLinear’s high-performance 224G PAM4 scale-up retimer, helping boost reliability of copper connectivity at 224G/lane speeds. We recently covered this challenge in our free newsletter, Inside Nvidia’s $4B Optical Strategy—and Why CPO Changes Everything, where effective length of AECs shortens from 100G to 200G speeds due to signal degradation. 

Annapurna is available in both eight and sixteen lane configurations to support both 1.6T and 3.2T AEC and on-board retimer deployments, and compatible with ESUN, UALink, and Ultra Ethernet protocols. Currently, MaxLinear is playing on the Ethernet side with Annapurna but management explained they are laying the groundwork to play wherever the retimer market moves in the future, such as with NVLink Fusion. 

The CEO stated the retimer opportunity is humungous: “So especially, I know there's a lot of hoopla about AECs because of success of one very successful company on AECs. But if you look at the market size, the opportunity for a silicon player, the AEC, the retimer market electrical for AI scale it inside the compute server is humongous. — as the speeds increase.” 

• Panther storage accelerator SoC 

MaxLinear’s Panther V storage accelerator enables ultra-low latency direct memory access between storage accelerators, NVMe SSDs, and network cards, bypassing host memory to reduce memory bandwidth consumption in AI workloads. It also offloads compute-intensive compression, deduplication or other tasks from the host CPU to accelerate performance. With on-chip SRAM, Panther V enables high-speed data movement across the storage tier, with its 450Gb/s throughput offering 2X the performance of its Panther III SoC.  

MaxLinear is actively sampling Panther V with key customers, with management forecasting storage accelerator revenue to at least double in 2026 over 2025 based on current engagement and design win activity across Tier 1 network and cloud providers: “And this is just the beginning of our Panther product — Panther road map product family. So we expect this year the revenues to double. We have said that before. And hopefully, next year as well, we got very strong growth based on the visibility we have.” 

The following context on Panther was also shared during the Q&A regarding its nuances at the product level, in addition to serving the massive memory market: “So the big benefit of Panther is it's an accelerator, so it reduces latency dramatically and the power efficiency that brings to it. So it enables much more capability than just memory compression, right? So I really feel that the performance part related to low latency, high-bandwidth access enablement that Panther provides is the key differentiator.  

Thus far, our use of Panther has been really at the enterprise appliance level, if you will. But now these enterprise storage appliance are getting increasingly deployed into mainstream cloud centers. So I really feel there's much more to come with Panther V and Panther VI in the future.” 

Optical DSP Demand Accelerating, Keystone Revenue Target Raised 40% 

It should be no surprise that optics demand is accelerating, as evidenced by recent earnings reports from Lumentum, Coherent and Applied Optoelectronics, and this accelerating demand is driving rapid growth for MaxLinear.  

While we are just one quarter into 2026, the strength of customer orders and visibility into optical program ramps at customers has already led MaxLinear to increase its optical data center growth forecast by 40%. MaxLinear now projects optical data center revenue between $150 million to $170 million for 2026, raised from its prior forecast from Q4 for $100 million to $130 million. Though management did not provide a 2025 number, analysts implied 2025 optical revenue of $60-70 million, so this could represent up to ~180% YoY growth.  

When questioned about intra-quarter dynamics driving this steep >40% raise (and rightfully so), management explained that they were being conservative with the initial guide and optimistic for stronger growth in the second half.  

“Now with all the visibility and the lead times that are necessary for providing the product, we have very good visibility and the ramps are setting in very nicely, both across 400-gig and 800-gig solutions. So I just think it's all about timing of the ramps and the success of the calls and our ability to scale up to meet the demand that the surging demand we are seeing now. 

The important readthrough here is that MaxLinear’s main optics growth story this year is still being driven by 400G and 800G, and is not yet benefitting from higher ASPs with 1.6T. This suggests that its optics growth runway is likely to extend and expand as both 800G and 1.6T ramp through 2027 and take a larger share of units and revenue, coinciding with Rushmore’s ramp – more on this next. 

Despite being a small fish in a big sea, MaxLinear has a few advantages on its side – it has design wins across all optics module vendors globally, with CEO Kishore Seendripu explaining that the “success of Keystone makes us an incumbent, right? And the power of incumbency is the ability to have the relationships with the cloud customers, the module makers, the confidence in your ability to supply and the quality of your product.”   

Additionally, MaxLinear has optical DSP growth coming from both hyperscaler-owned designs (direct qualification) as well as merchant solutions from module vendors, with CFO Steven Litchfield saying that growth is coming both from “hyperscaler-owned designs [and] through module vendors providing a merchant solution.” However, MaxLinear has not been upfront about which hyperscalers it is directly engaged/qualifying with and shipping to, rather emphasizing that they believe they are “only halfway there to our end data center diversification across all the hyperscalers.” This implies that hyperscale engagements may be more limited, or that there is still room to land additional hyperscalers in the future. 

This positions MaxLinear quite well for growth even as the industry continues to face tight supply constraints, as exposure across the vendor landscape should mean that it is rather insulated if individual suppliers face headwinds to growth from these constraints. Despite having this broad customer exposure, MaxLinear does expect to remain fairly concentrated on a few end customers through its ramp this year, expanding into 2027.  

Touching on 1.6T Dynamics and (Stiff) Competition 

There are a couple puts and takes for MaxLinear’s participation in 1.6T optical modules. On the positives, management hinted that demand is already shaping up much stronger than expected, aligning with commentary from transceiver vendors, with 1.6T boding well for growth via higher ASPs and margins.  

First on demand, CEO Kishore Seendripu explained that MaxLinear’s cloud and module vendor relationships, along with Rushmore’s interoperability with any DSP is “creating far more [of a] pull for 1.6T participation than I would have guessed at this point in time.” This is expected to “have an uplifting effect on our revenues and gross margins even as our market share expands” as mix shifts to 1.6T due to the higher ASPs, along with potential for more unit growth. MaxLinear hammered this point home by emphasizing further that 1.6T “will actually expand our ability to garner more revenues and more market share.” 

However, 1.6T is not appearing in growth (yet) and MaxLinear was straightforward in noting that they are not the first with 1.6T against their two incumbent competitors, which may present a challenge in catching up during the ramp. These two competitors are most likely to be Marvell and Broadcom. Broadcom will likely pose a more substantial competitive threat as its Taurus DSP platform is scaling to 400G/lane for 1.6T support (and upcoming 3.2T modules), double MaxLinear’s Rushmore and Marvell’s Ara platforms at 200G/lane. While competition against these two incumbents will be challenging, considering MaxLinear’s size, even a tiny gain in market share could translate to substantial revenue growth.  

1.6T revenue may begin appearing later this year, but the largest contributions will likely be tied to Rushmore’s ramp geared for 2027. Though MaxLinear has not provided much insight into the degree of ASP uplift from Keystone to Rushmore, a rough assumption for a ~30% uplift and a similarly-sized but accelerated ramp versus Keystone could see Rushmore quickly ramp to >$200 million in revenue by 2028. For comparison, Keystone launched in early 2023, and is scaling to >$150 million in roughly three years.  

Broader industry dynamics suggest that this shift upstream from 400G to 1.6T could be quite lucrative, looking beyond the simple ASP growth story. This is because MaxLinear will soon be playing in a much larger market of growth, and also because optics attach rates are expected to increase rapidly as Nvidia’s Vera Rubin platform comes online. 

Estimates from Goldman Sachs earlier this month suggest that 400G is likely accounting for just 5-9% of the market, while 800G is around the 20% level in Q1 and Q2, meaning MaxLinear’s 2026 optics growth story is being driven by the smaller third of the market. Shifting to 800G later this year and 1.6T into 2027 would see MaxLinear move to participating in a higher-value, larger market with both speeds expected to account for as much as 60% of the market by late next year.  

Source: Goldman Sachs 

Outside of shifting from a (declining) 400G market into mainstream 800G and 1.6T markets, growing attach rates for optical modules with Rubin further support strong growth. With Blackwell and Blackwell Ultra, GS estimates optics attach rates of roughly 1:2 to 1:3 depending on a two or three-layer networking topology, but sees this doubling to 1:4 to 1:6 with the VR200 rack. This combination of doubling attach rate translating to higher unit volumes and higher optics content per rack alongside ASP growth with 1.6T offers a strong tailwind for growth into 2027.   

MaxLinear Plays in Scale-Up, Scale-Out and CPO 

It’s important to touch briefly upon scale-up and scale-out demand, as optical transceiver demand is more heavily weighted towards scale-out applications due to copper’s physical limitations over longer distances at 200G and faster speeds.  

MaxLinear is seeing Keystone ramp for scale-up and scale-out with strong growth across both, and engagement across the two for Annapurna and Rushmore. These two products, targeting electrical retimers and AECs, is likely to be where MaxLinear’s scale-up growth appears as the majority of the optical transceiver TAM (70%) is weighted towards scale-out.  This scale-out focus is key over the medium-term as optical transceiver content could nearly triple from ~$173,000 in Blackwell Ultra to nearly $500,000 in the Rubin Ultra NVL144. 

Moving to CPO — as the industry shifts towards CPO, first revenues among the optics stack are expected to be realized in scale-out applications, though as we noted in our Lumentum analysis for Premium subscribers, scale-up opportunities could be larger. On this note, MaxLinear is preparing for a range of optical outcomes, with its Washington TIAs underpinning a ‘full platform’ approach for CPO or other packaged-optics solutions, such as LPO or NPO:  

“So the CPO market, if [customer are] going to be bare bones, then the TIA and driver is a natural fit. If they go more sophisticated on the DSP-based one, we already have the platform offering. But the real question comes, as you go towards XPO, CPOs and the various manifestations of it. So the full offering is super important. So Washington is the first step in the direction of a fundamental platform that will have multiple derivatives and incarnations.” 

The catch here is that MaxLinear does not expect CPO “to be a huge part of our revenues” over the next couple of years, stating they think they are “3 years out from determining” how the CPO market plays out. Regardless, it will be something to pay close attention to as CPO ramps are on deck for the primary module vendors later this year with further growth expected in 2027. 

Infrastructure Growth up 35% QoQ 

Putting this all together, MaxLinear is seeing robust growth arise in its Infrastructure segment, with Q1 revenue of $62.8 million up 136% YoY and notching one of the strongest QoQ growth rates in the AI industry this quarter at 35% (though arguably at quite a small scale).  

Additionally, data center growth is expected to see a “step function” increase in Q2 with strong upside expected into 2027, and based on commentary, this is likely tied primarily to Keystone. Currently, this is being modeled above at ~31% QoQ, or an increase to 50% revenue share from 46% in Q1. This would also maintain a similar YoY growth rate as Q1 at 138% YoY.  

Considering that Keystone has multiple customers progressing with ramps, more programs expected to layer in later this year and potential initial contributions from Rushmore as well, there is potential for MaxLinear to sustain strong sequential growth through year-end.  

Management hinted that they “absolutely” expect more upside to that $150-170 million optical forecast as programs reach full run rates — assuming that growth does not moderate following Q2’s step-function increase but instead remains robust at ~$20 million QoQ through Q4, this could project Q4 Infrastructure revenue to be roughly $122.5 million, up 163% YoY. This would represent approximately a $500 million annualized run rate, double its current rate at $250 million. 

Financials 

Revenue Growth Accelerating

MaxLinear’s revenue has since recovered from the stiffer headwinds it had faced in 2024, where it had seen (35%) or larger quarterly declines across all four of its segments. Q1 revenue was $137.2 million, up 43% YoY and roughly flat QoQ, marking a slight deceleration from 48% growth in Q4. The flat QoQ growth highlights the strength of Infrastructure and Keystone, as Broadband revenue saw a sharp (24%) QoQ decline.  

For Q2, MaxLinear guided for revenues between $160 to $170 million, representing a reacceleration to 51.7% YoY at midpoint, with QoQ growth similarly accelerating to 20.3%. This is expected to be primarily driven by Infrastructure revenue where MaxLinear projected a “step-function” increase stemming from strong optical interconnect demand, alongside growth from all four segments.  

Looking at the second half of 2026, consensus estimates currently point to growth moderating to the 30% range, exiting the year at 33% growth, a sharp ~18 point deceleration from Q2’s guide. This comes from consensus pointing to sequential dollar growth of just ~$8 million in both Q3 and Q4, a substantial step down from Q2’s guided $28 million at midpoint.  

However, there are multiple signals that suggest MaxLinear could exceed these estimates and maintain strong sequential growth in the back half of the year. Given the demand signals we have been seeing across the optical transceiver landscape, there should be few reasons that MaxLinear cannot maintain rather robust Infrastructure growth as outlined above, aside from a lack of execution with the 400G and 800G ramp with 1.6T on deck.  

Also layering into growth will be large-scale deployments later this year for single-chip fiber PON and WiFi 7 platforms at a second Tier 1 service provider in North America with additional ramps in Europe.  As such, maintaining ~$20-$30 million sequential dollar growth through 2H (which could come from Infrastructure alone) could see MaxLinear exit the year with quarterly revenues above $200 million. 

For the full year, consensus points to growth of 40.5% to $657 million, but the scenario discussed above for ~$25 million QoQ in 2H at the midpoint would place FY26 revenue at $707 million. 

Consensus estimates also point to revenue growth decelerating rather sharply to 20.7% to $791.3 million in 2027, yet expectations for “strong upside as run rates expand into 2027” within the data center and 1.6T ramping both suggest revenue could land significantly higher. This does not include MaxLinear’s first XGS-PON win with a hyperscale data center which management explained could be “quite a bit of needle mover even in the next year itself in the second half on a run rate basis.”  

Key Segments 

MaxLinear reports in four key segments: Infrastructure, now its largest segment as of Q1, Broadband, its historically largest segment, Connectivity, and Industrial/Multi-Market. 

As noted above, Infrastructure revenue was $62.8 million in Q1, accounting for 46% of revenue. This marked a 136% YoY and 35% QoQ increase, a sharp acceleration from 76% YoY and 15% QoQ in Q4. Assuming a step up to 50% revenue share in Q2 as the main growth driver next quarter, Infrastructure revenue would be roughly estimated at $82.5 million, up 138% YoY and 31% QoQ. 

Broadband revenue was $43.6 million, accounting for 32% of revenue. While revenue did increase 6% YoY for the segment, sequential growth was poor at (24%) QoQ. Management said there was a seasonality component to this QoQ decline, but the segment is expected to start growing in Q2 and into 2027, supported by fiber PON ramps. 

Connectivity revenue was $18.6 million, accounting for 14% of revenue. Growth was rather soft, down (8%) YoY but up 3% QoQ. Management sees wireless infrastructure momentum improving due to increased investments in 5G ran, transport overhaul and backhaul to support cloud and edge AI connectivity.  

Industrial and Multi-Market revenue was $12.2 million, accounting for 9% of revenue, and up 47% YoY but down (13%) QoQ. 

Margins Negative but Signs of Improvement in Q2 

MaxLinear’s GAAP margins have been quite heavily pressured down the line, yet Q2 is showing a notable shift as management guided for GAAP operating margin to jump towards positive territory. This is likely driven by the step function data center growth and margin tailwinds carried by higher speed optical components.  

GAAP gross margin was 57.5% in Q1, up from 56.1% a year ago and roughly flat QoQ, while adjusted gross margin was 59.5%, up less than a point YoY and again roughly flat QoQ.  

For Q2, MaxLinear guided for GAAP gross margin of 56-59%, up 1 point YoY and flat QoQ, and adjusted gross margin of 58-61%. Management noted that there are some headwinds to gross margin related to rising wafer costs and packaging, but they “will certainly continue to see nice benefits on the gross margin side as infrastructure gets to be a larger percentage of our business.” 

GAAP operating margin was (12.5%) in Q1, ticking slightly lower from (10.9%) in Q4 but marking a solid improvement from (48.1%) a year ago. Adjusted operating margin was 15.9%, slightly lower from Q4’s 16.2% but up from (1.7%) a year ago.  

GAAP operating margin is where green shoots are arising in Q2, with management guiding for a thin but positive 0.5% margin at midpoint. This would represent a strong 13 point sequential increase and MaxLinear’s first positive GAAP operating margin in over three years. Adjusted operating margin was guided to be 21%, a five point sequential increase and up nearly 14 points YoY. This large delta between GAAP and adjusted operating margins is primarily due to high SBC and some acquisition and integration-related costs. 

GAAP net margin was (32.9%) in Q1, as MaxLinear recorded a rather large $26.5 million income tax provision in the quarter; this compared to (51.8%) a year ago and just (10.9%) in Q4. Adjusted net margin was 14.2%, up from (4.6%) a year ago and 12.7% in Q4. 

GAAP Profitability Expected in 2H  

Given the guide for operating margin to shift back to positive territory in Q2, earnings are expected to soon follow.  

Driven by the income tax provision, GAAP EPS was a wide ($0.52), missing estimates for ($0.21) and only minimally improving from ($0.58) a year ago. Adjusted EPS was $0.22, beating the $0.18 estimate and increasing from ($0.05) a year ago. 

For Q2, GAAP EPS is expected to inflect towards profitability, with consensus pointing to just ($0.05). MaxLinear is expected to see positive GAAP EPS in both Q3 and Q4, though remaining very thin. Adjusted EPS is expected to be $0.33 in Q2, up 1,547% on a small comp of $0.02.  

For the full year, GAAP EPS is expected to be ($0.53), driven by Q1’s loss, while adjusted EPS is expected to be $1.34, up 331% YoY.  

Cash Flows and Balance Sheet 

Operating cash flow dipped to negative territory after three quarters positive, though MaxLinear had a strong reason for this – substantial prepayments for wafers to support increasing demand for data center products with increasing 2H backlogs.   

Q1 operating cash flow was ($8.9 million) for a (6.5%) margin, up from (11.9%) a year ago but down from 7.6% in Q4.  

Q1 free cash flow was ($11.1 million) for an (8.1%) margin, up from (14%) a year ago but down from 4.9% in Q4. 

Cash and equivalents totaled $62.5 million, while debt was $123.8 million. 

Inventories were $85.8 million, roughly flat YoY but up from $78.1 million in Q4. 

Conclusion 

MaxLinear is forecasting strong optical data center revenue growth from 400G and 800G products via Keystone, with its 1.6T focused Rushmore ramping later this year into 2027. Management has already raised its 2026 optical data center revenue forecast by >40% from $115 million to $160 million at the midpoints, with a step function increase expected next quarter. Ramping 1.6T later in 2026 and into 2027 presents further opportunities for growth to remain strong considering the ASP uplift and potential for increased content and attach rates within upcoming rack-scale systems.  

Fundamentally, while MaxLinear stands out for its 35% sequential growth in data center-driven revenue this quarter, it arguably is much weaker down the income statement than other optical beneficiaries such as Lumentum. Margins have been quite weak and Q1 did show a larger GAAP loss. Q2 is expected to right the ship and put the company potentially on a path to GAAP profitability in 2H for the first time in over three years.

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.

Recommended Reading:

• SiTime: Precision Timing Solutions Increasing in Importance, FY Revenue Growth Guide of >80%
• Seagate FQ3: Data Center Growth Accelerates to 12% QoQ, Mozaic4+ Ramping
• NextEra: Pivoting Beyond a Regulated Utility Player
• Silicon Motion: Strong Consumer SSD Demand, Trying to Move into AI Enterprise Markets for 2027-2028

MaxLinear is another under-the-radar optical networking beneficiary, supplying a range of components within optical transceivers, with its most notable being optical DSPs for 400G, 800G and soon 1.6T solutions.  

The company is seeing strong demand emerge for its Keystone DSP family, with management raising its 2026 revenue forecast for Keystone by >40% already, from $100-130 million to $150-170 million. Impressively, this growth is being driven by 400G and 800G modules as the 1.6T product, Rushmore, has yet to ramp. 

The pivot upstream to ramping more heavily on 800G and 1.6T is expected to benefit both revenues and margins from higher ASPs, but it also opens the door for a prolonged revenue runway as 1.6T growth is expected to maintain strong through 2027 with attach rates increasing with larger GPU systems.  

Fundamentally, margins have remained pressured and cash flows are quite thin, but there are some green shoots emerging as MaxLinear is forecasting a return to GAAP operating profitability in Q2 for the first time in three years.  

Brief Product Overview 

MaxLinear supplies a range of key components within optical transceivers, AECs and other connectivity solutions, such as PAM4 DSPs, transimpedance amplifiers (TIAs) and storage accelerator SoCs. While AI data centers (Infrastructure) are rapidly becoming MaxLinear’s main growth outlet, it also serves broadband, wireless, automotive and industrial end markets.   

• Keystone Optical DSPs 

MaxLinear’s Keystone family spans twelve PAM4 DSPs, including both 400G to 800G with integrated drivers for EML/SiPho transceivers, bare die options, and driverless DSPs. MaxLinear also has introduced variants of Keystone to target both AOCs and AECs as well.  

The company says the third-gen DSPs offer ‘best-in-class’ power consumption, enabling 7W 400G optical modules and 13W 800G designs. The Keystone DSPs can also be paired with MaxLinear’s Topanga and Washington TIAs to provide complete solutions for customers.  

Keystone was stated this quarter to be ramping at multiple major hyperscale customers across the US and Asia for both 400G and 800G scale-up and scale-out applications. It also underpins MaxLinear’s optical data center momentum this year into 2027, with management seeing DSPs as their outright #1 TAM.  

Keystone is expected to drive the near-term growth, per the opening remarks: “We also expect a step function data center revenue increase beginning in Q2 with expected strong upside as run rates expand into 2027. At the center of this data center momentum is our Keystone PAM4 DSP optical transceiver platform. Keystone is now ramping at multiple major hyperscale customers across both the U.S. and Asia, supporting 400G and 800G — 800G PAM4 deployments for scale-up and scale-out applications.” 

• Rushmore 200G PAM4 DSP 

MaxLinear unveiled Rushmore last year, its 200G/lane PAM4 SerDes and DPS supporting 1.6T optics and active copper cable deployments. Rushmore is compatible with both Ethernet and InfiniBand, and will enable <25W 1.6T optical interconnects with low latency, and optimized performance across all laser technologies from EMLs to SiPho. MaxLinear says that when paired with its Washington TIA, Rushmore will “provide the best performance on the market.” 

MaxLinear added in Q1 that customer engagement for Rushmore is accelerating faster than expected, with production ramps expected in late 2026 with strong revenue growth continuing into 2027.  

Here is what was stated in the opening remarks: “So moving forward to 1.6T, the critical thing to keep in mind is that there is enormous confidence out there. We're shipping Keystone into major data centers today, and they're ramping very strongly in 2026. And we have now rolled out our 1.6 Terabit Rushmore product and Annapurna family for electrical applications.  

And I think that this level of execution a part and the success with the cloud relationships, module partnerships and the call and interop completion is creating a far more pull for 1.6T participation than I would have guessed at this point in time. So in a sense, we hope that by the end of the year, we'll have called on 1.6T and start transitioning […]” 

• Washington 200G TIA 

Transimpedance amplifiers (TIAs) are current-to-voltage converters, amplifying the current generated by photodetectors within optical modules to voltage signals that can be processed by DSPs. For high-speed optics such as 1.6T transceivers, TIAs play an important role in helping maintain signal integrity at faster rates. 

MaxLinear recently unveiled its four lane/200G Washington TIA at the end of April, designed for 1.6T transceivers and capable of supporting fully retimed, half-retimed, and linear interfaces, from LRO/LPO, NPO and CPO applications.  

Washington interoperates with PAM4 DSPs from all major DSP vendors, simplifying integration into existing system architectures, while also complementing MaxLinear’s portfolio of DSPs and drivers to allow customers to build and customize systems for their specific power, performance and reach requirements. 

As discussed below, the ability for the TIA to be a component in any architecture is key: “But honestly, I mean, the TIA is beyond the TIA for Rushmore, right? If you think of an LPO strategy that the TIA is a fundamental block. If you think about LRO strategy, the TA is a fundamental block.” 

• Annapurna scale-up retimer 

Annapurna is MaxLinear’s high-performance 224G PAM4 scale-up retimer, helping boost reliability of copper connectivity at 224G/lane speeds. We recently covered this challenge in our free newsletter, Inside Nvidia’s $4B Optical Strategy—and Why CPO Changes Everything, where effective length of AECs shortens from 100G to 200G speeds due to signal degradation. 

Annapurna is available in both eight and sixteen lane configurations to support both 1.6T and 3.2T AEC and on-board retimer deployments, and compatible with ESUN, UALink, and Ultra Ethernet protocols. Currently, MaxLinear is playing on the Ethernet side with Annapurna but management explained they are laying the groundwork to play wherever the retimer market moves in the future, such as with NVLink Fusion. 

The CEO stated the retimer opportunity is humungous: “So especially, I know there's a lot of hoopla about AECs because of success of one very successful company on AECs. But if you look at the market size, the opportunity for a silicon player, the AEC, the retimer market electrical for AI scale it inside the compute server is humongous. — as the speeds increase.” 

• Panther storage accelerator SoC 

MaxLinear’s Panther V storage accelerator enables ultra-low latency direct memory access between storage accelerators, NVMe SSDs, and network cards, bypassing host memory to reduce memory bandwidth consumption in AI workloads. It also offloads compute-intensive compression, deduplication or other tasks from the host CPU to accelerate performance. With on-chip SRAM, Panther V enables high-speed data movement across the storage tier, with its 450Gb/s throughput offering 2X the performance of its Panther III SoC.  

MaxLinear is actively sampling Panther V with key customers, with management forecasting storage accelerator revenue to at least double in 2026 over 2025 based on current engagement and design win activity across Tier 1 network and cloud providers: “And this is just the beginning of our Panther product — Panther road map product family. So we expect this year the revenues to double. We have said that before. And hopefully, next year as well, we got very strong growth based on the visibility we have.” 

The following context on Panther was also shared during the Q&A regarding its nuances at the product level, in addition to serving the massive memory market: “So the big benefit of Panther is it's an accelerator, so it reduces latency dramatically and the power efficiency that brings to it. So it enables much more capability than just memory compression, right? So I really feel that the performance part related to low latency, high-bandwidth access enablement that Panther provides is the key differentiator.  

Thus far, our use of Panther has been really at the enterprise appliance level, if you will. But now these enterprise storage appliance are getting increasingly deployed into mainstream cloud centers. So I really feel there's much more to come with Panther V and Panther VI in the future.” 

Optical DSP Demand Accelerating, Keystone Revenue Target Raised 40% 

It should be no surprise that optics demand is accelerating, as evidenced by recent earnings reports from Lumentum, Coherent and Applied Optoelectronics, and this accelerating demand is driving rapid growth for MaxLinear.  

While we are just one quarter into 2026, the strength of customer orders and visibility into optical program ramps at customers has already led MaxLinear to increase its optical data center growth forecast by 40%. MaxLinear now projects optical data center revenue between $150 million to $170 million for 2026, raised from its prior forecast from Q4 for $100 million to $130 million. Though management did not provide a 2025 number, analysts implied 2025 optical revenue of $60-70 million, so this could represent up to ~180% YoY growth.  

When questioned about intra-quarter dynamics driving this steep >40% raise (and rightfully so), management explained that they were being conservative with the initial guide and optimistic for stronger growth in the second half.  

“Now with all the visibility and the lead times that are necessary for providing the product, we have very good visibility and the ramps are setting in very nicely, both across 400-gig and 800-gig solutions. So I just think it's all about timing of the ramps and the success of the calls and our ability to scale up to meet the demand that the surging demand we are seeing now. 

The important readthrough here is that MaxLinear’s main optics growth story this year is still being driven by 400G and 800G, and is not yet benefitting from higher ASPs with 1.6T. This suggests that its optics growth runway is likely to extend and expand as both 800G and 1.6T ramp through 2027 and take a larger share of units and revenue, coinciding with Rushmore’s ramp – more on this next. 

Despite being a small fish in a big sea, MaxLinear has a few advantages on its side – it has design wins across all optics module vendors globally, with CEO Kishore Seendripu explaining that the “success of Keystone makes us an incumbent, right? And the power of incumbency is the ability to have the relationships with the cloud customers, the module makers, the confidence in your ability to supply and the quality of your product.”   

Additionally, MaxLinear has optical DSP growth coming from both hyperscaler-owned designs (direct qualification) as well as merchant solutions from module vendors, with CFO Steven Litchfield saying that growth is coming both from “hyperscaler-owned designs [and] through module vendors providing a merchant solution.” However, MaxLinear has not been upfront about which hyperscalers it is directly engaged/qualifying with and shipping to, rather emphasizing that they believe they are “only halfway there to our end data center diversification across all the hyperscalers.” This implies that hyperscale engagements may be more limited, or that there is still room to land additional hyperscalers in the future. 

This positions MaxLinear quite well for growth even as the industry continues to face tight supply constraints, as exposure across the vendor landscape should mean that it is rather insulated if individual suppliers face headwinds to growth from these constraints. Despite having this broad customer exposure, MaxLinear does expect to remain fairly concentrated on a few end customers through its ramp this year, expanding into 2027.  

Touching on 1.6T Dynamics and (Stiff) Competition 

There are a couple puts and takes for MaxLinear’s participation in 1.6T optical modules. On the positives, management hinted that demand is already shaping up much stronger than expected, aligning with commentary from transceiver vendors, with 1.6T boding well for growth via higher ASPs and margins.  

First on demand, CEO Kishore Seendripu explained that MaxLinear’s cloud and module vendor relationships, along with Rushmore’s interoperability with any DSP is “creating far more [of a] pull for 1.6T participation than I would have guessed at this point in time.” This is expected to “have an uplifting effect on our revenues and gross margins even as our market share expands” as mix shifts to 1.6T due to the higher ASPs, along with potential for more unit growth. MaxLinear hammered this point home by emphasizing further that 1.6T “will actually expand our ability to garner more revenues and more market share.” 

However, 1.6T is not appearing in growth (yet) and MaxLinear was straightforward in noting that they are not the first with 1.6T against their two incumbent competitors, which may present a challenge in catching up during the ramp. These two competitors are most likely to be Marvell and Broadcom. Broadcom will likely pose a more substantial competitive threat as its Taurus DSP platform is scaling to 400G/lane for 1.6T support (and upcoming 3.2T modules), double MaxLinear’s Rushmore and Marvell’s Ara platforms at 200G/lane. While competition against these two incumbents will be challenging, considering MaxLinear’s size, even a tiny gain in market share could translate to substantial revenue growth.  

1.6T revenue may begin appearing later this year, but the largest contributions will likely be tied to Rushmore’s ramp geared for 2027. Though MaxLinear has not provided much insight into the degree of ASP uplift from Keystone to Rushmore, a rough assumption for a ~30% uplift and a similarly-sized but accelerated ramp versus Keystone could see Rushmore quickly ramp to >$200 million in revenue by 2028. For comparison, Keystone launched in early 2023, and is scaling to >$150 million in roughly three years.  

Broader industry dynamics suggest that this shift upstream from 400G to 1.6T could be quite lucrative, looking beyond the simple ASP growth story. This is because MaxLinear will soon be playing in a much larger market of growth, and also because optics attach rates are expected to increase rapidly as Nvidia’s Vera Rubin platform comes online. 

Estimates from Goldman Sachs earlier this month suggest that 400G is likely accounting for just 5-9% of the market, while 800G is around the 20% level in Q1 and Q2, meaning MaxLinear’s 2026 optics growth story is being driven by the smaller third of the market. Shifting to 800G later this year and 1.6T into 2027 would see MaxLinear move to participating in a higher-value, larger market with both speeds expected to account for as much as 60% of the market by late next year.  

Source: Goldman Sachs 

Outside of shifting from a (declining) 400G market into mainstream 800G and 1.6T markets, growing attach rates for optical modules with Rubin further support strong growth. With Blackwell and Blackwell Ultra, GS estimates optics attach rates of roughly 1:2 to 1:3 depending on a two or three-layer networking topology, but sees this doubling to 1:4 to 1:6 with the VR200 rack. This combination of doubling attach rate translating to higher unit volumes and higher optics content per rack alongside ASP growth with 1.6T offers a strong tailwind for growth into 2027.   

MaxLinear Plays in Scale-Up, Scale-Out and CPO 

It’s important to touch briefly upon scale-up and scale-out demand, as optical transceiver demand is more heavily weighted towards scale-out applications due to copper’s physical limitations over longer distances at 200G and faster speeds.  

MaxLinear is seeing Keystone ramp for scale-up and scale-out with strong growth across both, and engagement across the two for Annapurna and Rushmore. These two products, targeting electrical retimers and AECs, is likely to be where MaxLinear’s scale-up growth appears as the majority of the optical transceiver TAM (70%) is weighted towards scale-out.  This scale-out focus is key over the medium-term as optical transceiver content could nearly triple from ~$173,000 in Blackwell Ultra to nearly $500,000 in the Rubin Ultra NVL144. 

Moving to CPO — as the industry shifts towards CPO, first revenues among the optics stack are expected to be realized in scale-out applications, though as we noted in our Lumentum analysis for Premium subscribers, scale-up opportunities could be larger. On this note, MaxLinear is preparing for a range of optical outcomes, with its Washington TIAs underpinning a ‘full platform’ approach for CPO or other packaged-optics solutions, such as LPO or NPO:  

“So the CPO market, if [customer are] going to be bare bones, then the TIA and driver is a natural fit. If they go more sophisticated on the DSP-based one, we already have the platform offering. But the real question comes, as you go towards XPO, CPOs and the various manifestations of it. So the full offering is super important. So Washington is the first step in the direction of a fundamental platform that will have multiple derivatives and incarnations.” 

The catch here is that MaxLinear does not expect CPO “to be a huge part of our revenues” over the next couple of years, stating they think they are “3 years out from determining” how the CPO market plays out. Regardless, it will be something to pay close attention to as CPO ramps are on deck for the primary module vendors later this year with further growth expected in 2027. 

Infrastructure Growth up 35% QoQ 

Putting this all together, MaxLinear is seeing robust growth arise in its Infrastructure segment, with Q1 revenue of $62.8 million up 136% YoY and notching one of the strongest QoQ growth rates in the AI industry this quarter at 35% (though arguably at quite a small scale).  

Additionally, data center growth is expected to see a “step function” increase in Q2 with strong upside expected into 2027, and based on commentary, this is likely tied primarily to Keystone. Currently, this is being modeled above at ~31% QoQ, or an increase to 50% revenue share from 46% in Q1. This would also maintain a similar YoY growth rate as Q1 at 138% YoY.  

Considering that Keystone has multiple customers progressing with ramps, more programs expected to layer in later this year and potential initial contributions from Rushmore as well, there is potential for MaxLinear to sustain strong sequential growth through year-end.  

Management hinted that they “absolutely” expect more upside to that $150-170 million optical forecast as programs reach full run rates — assuming that growth does not moderate following Q2’s step-function increase but instead remains robust at ~$20 million QoQ through Q4, this could project Q4 Infrastructure revenue to be roughly $122.5 million, up 163% YoY. This would represent approximately a $500 million annualized run rate, double its current rate at $250 million. 

Financials 

Revenue Growth Accelerating

MaxLinear’s revenue has since recovered from the stiffer headwinds it had faced in 2024, where it had seen (35%) or larger quarterly declines across all four of its segments. Q1 revenue was $137.2 million, up 43% YoY and roughly flat QoQ, marking a slight deceleration from 48% growth in Q4. The flat QoQ growth highlights the strength of Infrastructure and Keystone, as Broadband revenue saw a sharp (24%) QoQ decline.  

For Q2, MaxLinear guided for revenues between $160 to $170 million, representing a reacceleration to 51.7% YoY at midpoint, with QoQ growth similarly accelerating to 20.3%. This is expected to be primarily driven by Infrastructure revenue where MaxLinear projected a “step-function” increase stemming from strong optical interconnect demand, alongside growth from all four segments.  

Looking at the second half of 2026, consensus estimates currently point to growth moderating to the 30% range, exiting the year at 33% growth, a sharp ~18 point deceleration from Q2’s guide. This comes from consensus pointing to sequential dollar growth of just ~$8 million in both Q3 and Q4, a substantial step down from Q2’s guided $28 million at midpoint.  

However, there are multiple signals that suggest MaxLinear could exceed these estimates and maintain strong sequential growth in the back half of the year. Given the demand signals we have been seeing across the optical transceiver landscape, there should be few reasons that MaxLinear cannot maintain rather robust Infrastructure growth as outlined above, aside from a lack of execution with the 400G and 800G ramp with 1.6T on deck.  

Also layering into growth will be large-scale deployments later this year for single-chip fiber PON and WiFi 7 platforms at a second Tier 1 service provider in North America with additional ramps in Europe.  As such, maintaining ~$20-$30 million sequential dollar growth through 2H (which could come from Infrastructure alone) could see MaxLinear exit the year with quarterly revenues above $200 million. 

For the full year, consensus points to growth of 40.5% to $657 million, but the scenario discussed above for ~$25 million QoQ in 2H at the midpoint would place FY26 revenue at $707 million. 

Consensus estimates also point to revenue growth decelerating rather sharply to 20.7% to $791.3 million in 2027, yet expectations for “strong upside as run rates expand into 2027” within the data center and 1.6T ramping both suggest revenue could land significantly higher. This does not include MaxLinear’s first XGS-PON win with a hyperscale data center which management explained could be “quite a bit of needle mover even in the next year itself in the second half on a run rate basis.”  

Key Segments 

MaxLinear reports in four key segments: Infrastructure, now its largest segment as of Q1, Broadband, its historically largest segment, Connectivity, and Industrial/Multi-Market. 

As noted above, Infrastructure revenue was $62.8 million in Q1, accounting for 46% of revenue. This marked a 136% YoY and 35% QoQ increase, a sharp acceleration from 76% YoY and 15% QoQ in Q4. Assuming a step up to 50% revenue share in Q2 as the main growth driver next quarter, Infrastructure revenue would be roughly estimated at $82.5 million, up 138% YoY and 31% QoQ. 

Broadband revenue was $43.6 million, accounting for 32% of revenue. While revenue did increase 6% YoY for the segment, sequential growth was poor at (24%) QoQ. Management said there was a seasonality component to this QoQ decline, but the segment is expected to start growing in Q2 and into 2027, supported by fiber PON ramps. 

Connectivity revenue was $18.6 million, accounting for 14% of revenue. Growth was rather soft, down (8%) YoY but up 3% QoQ. Management sees wireless infrastructure momentum improving due to increased investments in 5G ran, transport overhaul and backhaul to support cloud and edge AI connectivity.  

Industrial and Multi-Market revenue was $12.2 million, accounting for 9% of revenue, and up 47% YoY but down (13%) QoQ. 

Margins Negative but Signs of Improvement in Q2 

MaxLinear’s GAAP margins have been quite heavily pressured down the line, yet Q2 is showing a notable shift as management guided for GAAP operating margin to jump towards positive territory. This is likely driven by the step function data center growth and margin tailwinds carried by higher speed optical components.  

GAAP gross margin was 57.5% in Q1, up from 56.1% a year ago and roughly flat QoQ, while adjusted gross margin was 59.5%, up less than a point YoY and again roughly flat QoQ.  

For Q2, MaxLinear guided for GAAP gross margin of 56-59%, up 1 point YoY and flat QoQ, and adjusted gross margin of 58-61%. Management noted that there are some headwinds to gross margin related to rising wafer costs and packaging, but they “will certainly continue to see nice benefits on the gross margin side as infrastructure gets to be a larger percentage of our business.” 

GAAP operating margin was (12.5%) in Q1, ticking slightly lower from (10.9%) in Q4 but marking a solid improvement from (48.1%) a year ago. Adjusted operating margin was 15.9%, slightly lower from Q4’s 16.2% but up from (1.7%) a year ago.  

GAAP operating margin is where green shoots are arising in Q2, with management guiding for a thin but positive 0.5% margin at midpoint. This would represent a strong 13 point sequential increase and MaxLinear’s first positive GAAP operating margin in over three years. Adjusted operating margin was guided to be 21%, a five point sequential increase and up nearly 14 points YoY. This large delta between GAAP and adjusted operating margins is primarily due to high SBC and some acquisition and integration-related costs. 

GAAP net margin was (32.9%) in Q1, as MaxLinear recorded a rather large $26.5 million income tax provision in the quarter; this compared to (51.8%) a year ago and just (10.9%) in Q4. Adjusted net margin was 14.2%, up from (4.6%) a year ago and 12.7% in Q4. 

GAAP Profitability Expected in 2H  

Given the guide for operating margin to shift back to positive territory in Q2, earnings are expected to soon follow.  

Driven by the income tax provision, GAAP EPS was a wide ($0.52), missing estimates for ($0.21) and only minimally improving from ($0.58) a year ago. Adjusted EPS was $0.22, beating the $0.18 estimate and increasing from ($0.05) a year ago. 

For Q2, GAAP EPS is expected to inflect towards profitability, with consensus pointing to just ($0.05). MaxLinear is expected to see positive GAAP EPS in both Q3 and Q4, though remaining very thin. Adjusted EPS is expected to be $0.33 in Q2, up 1,547% on a small comp of $0.02.  

For the full year, GAAP EPS is expected to be ($0.53), driven by Q1’s loss, while adjusted EPS is expected to be $1.34, up 331% YoY.  

Cash Flows and Balance Sheet 

Operating cash flow dipped to negative territory after three quarters positive, though MaxLinear had a strong reason for this – substantial prepayments for wafers to support increasing demand for data center products with increasing 2H backlogs.   

Q1 operating cash flow was ($8.9 million) for a (6.5%) margin, up from (11.9%) a year ago but down from 7.6% in Q4.  

Q1 free cash flow was ($11.1 million) for an (8.1%) margin, up from (14%) a year ago but down from 4.9% in Q4. 

Cash and equivalents totaled $62.5 million, while debt was $123.8 million. 

Inventories were $85.8 million, roughly flat YoY but up from $78.1 million in Q4. 

Conclusion 

MaxLinear is forecasting strong optical data center revenue growth from 400G and 800G products via Keystone, with its 1.6T focused Rushmore ramping later this year into 2027. Management has already raised its 2026 optical data center revenue forecast by >40% from $115 million to $160 million at the midpoints, with a step function increase expected next quarter. Ramping 1.6T later in 2026 and into 2027 presents further opportunities for growth to remain strong considering the ASP uplift and potential for increased content and attach rates within upcoming rack-scale systems.  

Fundamentally, while MaxLinear stands out for its 35% sequential growth in data center-driven revenue this quarter, it arguably is much weaker down the income statement than other optical beneficiaries such as Lumentum. Margins have been quite weak and Q1 did show a larger GAAP loss. Q2 is expected to right the ship and put the company potentially on a path to GAAP profitability in 2H for the first time in over three years.

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.

Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.

Recommended Reading:

• Arm: Computex Update, CPU Core Demand Hinted at Being Higher
• CoreWeave: Revenue Inflecting in 2H, Margin and Profit Questions
• Core Scientific: Multi-GW Pipeline, New Hyperscaler Interest but Still Tied to CoreWeave
• Dell Fiscal Q1: Agentic AI Creates Tailwind for Traditional Servers and Storage

MEMS timing supplier SiTime is seeing solid tailwinds in AI data centers from the increasing complexity of rack-scale platforms and shift to faster data rates in networking switches and optical transceivers, as these place more emphasis on timing solutions to ensure that all components operate as one cohesive unit with maximum performance and reliability.  

SiTime delivered a strong print this week by dramatically beating Q1 estimates, guiding far above Q2 estimates and raising full-year guidance. Q1 revenue of $113.6 million beat consensus of $103.5 million for a 10% beat, yet EPS grew 5X YoY and reported a 23% beat for $1.44.  

Gross margin was especially strong at 64.5% for a 7-point expansion, but the more pronounced margin expansion was operating margin at 28% for a 25-point expansion. Operating cash flow more than doubled to $31.2 million. 

While Q1’s strength is notable, the more pronounced beat is in the Q2 guide for revenue of $145M at the midpoint for growth of 100% YoY. This compares to analyst estimates for 62% growth in Q2. Management also stated gross margin would be above 65% and operating margin above 30% next quarter.  

This led to raising the full year fiscal year guide, with management stating: “For the full year, we are increasing our revenue growth expectations to at least 80%, well above our prior expectations and our long-term target growth rate of 25%-30%. This step change in growth reflects both the depth of our order book and the confidence customers are signaling in their own demand forecast, particularly in CED. That confidence is translating into improved visibility, reinforcing our expectation for sustained momentum throughout the year.” 

Brief Overview of Key Products 

SiTime offers a range of MEMS-based (micro-electromechanical system) timing solutions that it says offer greater resilience, lower power, higher performance and a smaller size versus traditional quartz solutions.  

• Oscillators 

MEMS oscillators are timing devices embedded on a silicon chip that generate a clock signal (frequency) used to coordinate actions of different components, essentially serving as the ‘heartbeat’ of the device; they do this by combining clocks and resonators in a single system. As it relates to AI server buildouts, increasing rack and cluster sizes means data must move across hundreds to thousands of chips at once, requiring precise synchronization across components and interconnects to minimize latency, prevent data loss and maximize system efficiency.  

SiTime’s high-performance oscillators are prevalent across the compute tray within the GPU and CPU boards, NIC cards, and networking switches, and also within the networking fabric, from top-of-rack and spine switches to optical transceivers and AECs. SiTime says its MEMS oscillators can reduce power consumption by 30–50% versus quartz with similar or better frequency in a more compact footprint. 

• Clock ICs – Generators and Buffers  

Clock buffers take a single clock signal input and then multiply and boost the signal across multiple different output lines, ensuring that the clock signals reach all parts of the circuit. Clock generators are a type of oscillator that produce periodic timing signals for all of the components within the circuit, and can handle different frequencies needed for GPUs, CPUs, memory, PCIe and Ethernet components. Generators also help reduce jitter and improve signal integrity on the board.  

SiTime recently launched its Chorus clock-system-on-a-chip family for AI data center applications in April 2024, which combines clocks, oscillators and resonators into a single integrated chip, offering up to 10X higher performance compared to standalone oscillators and clocks. SiTime executives say the new solution also accelerates development time by up to six weeks, and reduces board area for timing by up to 50%, while addressing issues such as noise and impedance mismatch between resonators and clocks. Back in Q3, the clock funnel was stated to have quadrupled in the past year to $300 million, though revenue was stated as sub-$20 million in December. 

SiTime is also acquiring Renesas’ timing unit, which it says will boost its clock IC portfolio by 10X – more on this below.  

• Resonators 

Resonators are key components within oscillators that vibrate at a stable frequency, essentially setting the frequency of the oscillator, determining frequency accuracy and ensuring stability over a range of temperatures. SiTime recently launched its high-performance Titan resonator family in September, which it says offers improved performance under high shock and vibration, while occupying 4.2x and 12x less PCB space than quartz competitors. The new platform is initially geared towards IoT, wearables and medical device applications.   

Precision Timing Products Accelerate from Inference Market 

SiTime’s Communications, Enterprise and Datacenter segment (CED) grew 158% YoY and 17% QoQ for the eight consecutive quarter of triple-digit growth. According to the earnings call, the primary driver for CED strength is the shift from training to inference with newer XPUs requiring 2X to 4X more content per system than previous training workloads. According to management, utilization rates in inference workloads are running 20% to 40% today yet need to reach 50% to 60% for reasonable ROI on capex. 

Notably, there is increased unit volume combined with higher ASPs on the Elite and Elite 2 product line.  

Here is what was stated on the call: 

“On inference infrastructure built on newer XPU, it needs 2 to 4 times more timing content per system than in training infrastructure. GPU utilization in inference workloads is now 20% to 40% and is targeted to get to 50% to 60%. Here, time synchronization plays a critical role in achieving higher GPU utilization and SiTime benefits from its products being used in this application. This emphasis on synchronization is driving demand for high ASP and high-margin products.  

Elite and Elite RF Super-TCXOs are widely deployed in AI infrastructure, and we have recently exceeded and extended our leadership with the new Elite 2 Super-TCXO family. This newer Elite 2 delivers up to 3 times better synchronization performance compared to Elite, which was already significantly better than quartz oscillators.” 

Additionally, as 1.6T ramps, SiTime foresees additional share gains due to higher frequencies and “tighter resilience requirements” driving demand for advanced oscillators. 

The following was stated: 

“As hyperscalers increase networking bandwidth within the data center, we expect to see meaningful adoption of 1.6T optical modules in 2026. Higher frequencies and the need for more resilient performance are driving demand of our advanced oscillators at a higher price than those used in 800G. At the same time, we expect to see continued strong shipment for oscillators for 400G and 800G for at least the next two years.” 

As stated above, management noted that 400G and 800G will remain strong for the next wo years while 1.6T ramps. However, 1.6T will see higher ASPs than 800G with the CEO stating that SiTime can charge a premium price by being the highest-performing option: “This newer Elite 2 delivers up to 3 times better synchronization performance compared to Elite, which was already significantly better than quartz oscillators.” 

Quantifying Inference System Content Opportunities  

SiTime has offered some clues into holistic dollar content per rack, general content per networking components, as well as commentary on how the above trends are shaping content growth. This provides a bit more insight into where the 2X to 4X increase in content per system with inference deployments could land. 

In terms of the holistic dollar content per rack, management has explained that for training platforms where they have a high penetration across the system and networking topology, content opportunities “can be multiple hundreds of dollars in a fully integrated rack,” with opportunities potentially scaling larger in networking fabrics.  

Translating this ‘multiple hundreds of dollars’ to the 2X to 4X increase with inference roughly estimates that SiTime could see content above $1,000 in fully integrated racks for inference deployments — this was also mentioned by analysts in Q1’s call that content “certainly sounds like it could reach into the $1,000+ range.” 

Additionally, to briefly touch upon networking opportunities moving up the stack, SiTime has previously mentioned that content for some optical modules can vary from $1 to $2 ASPs, but shifting to switches and farther up the stack can drive more meaningful content, along the lines of $7 to $10 ASPs.   

CPO Switches to Drive 3X Timing Content 

We’ve published quite a bit on the CPO opportunity, especially in our Coherent and Lumentum analyses. As data centers migrate from pluggables to NPO/CPO, SiTime can benefit from this shift as optics move inside the switch. The result will be more oscillator sockets per switch and higher performance requirements, resulting in 3X higher timing dollar content. Management stated: “On CPO or co-packaged optics, in our discussion with customers, we see even greater strength. For example, in CPO switches, where timing content can be up to 3 times higher.” 

Regarding supply to serve an influx of demand, SiTime also offers high confidence commentary that they have no bottlenecks with an analyst referencing SiTime having a strong supply chain during 2020, unlike many peers: 

“We see no issues around supply chain in particular. I know some people have said that in the past, other semiconductor companies, so we want to be very clear about that. We see strength in our supply chain, and we don’t see any fundamental issues or macro issues or external issues that can trip us up as of now.” 

Book-to-Bill Accelerating 

SiTime does not typically offer its book-to-bill ratio, but brief commentary from Q1 that book-to-bill is growing with pull-through from CED, taken with Q4’s book-to-bill of >1.5X, suggests this ratio is moving higher.  

When management had provided the book-to-bill in Q4, analysts had questioned on the duration of this backlog, and if it would be six, 12 or 18 months and beyond, to which management said it is typically within 12 months:  

“So in terms of the book-to-bill, I think Rajesh talked about the fact that we are seeing customers maybe book out a little longer, but typically, that's well within 12 months. We see a lot of ordering over the next couple of quarters. But we are seeing some customers book meaningfully in the second half already as well. But I would say definitely weighted to Q1 and Q2 in terms of that.” 

This implies near-term demand is strengthening in Q1, driven by CED, while it provides a further layer of confidence in SiTime’s upbeat annual revenue growth guide of >80%. It also suggests Q2 momentum is likely to remain robust, and could signal a similarly strong strong 2H if orders continue to flow as Nvidia’s Blackwell Ultra and Rubin ramp throughout the year alongside strong potential growth in 1.6T transceiver volumes. 

Telecom Offers Diversity for AI-Driven Demand 

Worth noting is that SiTime sells into the telecom industry, to help diversify its customer base beyond hyperscalers. While telecom has gone through a significant trough in recent years, the industry is expected to see an AI-driven resurgence as workloads run at the edge and in the access network. The key markets that SiTime can benefit from are RAN optimization, edge AI inference at base stations, and Open RAN architectures. Each of these trends require more timing sockets and more precision timing requirements, leading to a 3X uplift in content: 

“Finishing up on the telecom part of CED, we see increasing convergence between AI and advanced telecom infrastructures, especially in 5G RAN or Radio Access Network and demand from new applications such as FWA or fixed wireless access. AI-enabled telecom designs contain 3 times higher timing content, primarily from high ASP oscillators and clocks” 

Acquisition of Renesas’ Timing Unit 

SiTime is acquiring Renesas’ timing unit for ~$1.5 billion, significantly increasing its clocking portfolio by ~10X, adding a range of hyperscaler and leading AI server customers, and providing a substantial boost to SiTime’s CED revenue.  

Most importantly, the timing unit acquisition is expected to significantly increase the scale of SiTime’s CED business. Management had explained in Q4 that the acquisition will nearly double its CED business, adding that in the first 12 months post-close (likely starting Q2 ’26), the timing unit is expected to generate more than $300 million in revenue with ~75% of that from CED, or ~$225 million.  

Moving down the line, the timing unit is expected to accretive to both margins and EPS in the first full year post-close. Management explained that the unit has adjusted gross margins around 70%, or nearly 9 points higher than SiTime’s Q4 adjusted gross margin of 61.2%; it will also help push SiTime towards the upper end of its long-term 60-65% gross margin target model. The acquisition is expected to help drive adjusted operating margins above 30% from increased operating leverage at scale, compared to FY25’s 17.9% margin. 

From the product and customer side, SiTime sees the acquisition taking them to scale in clocking, adding 500 differentiated clock products to its portfolio, boosting it by 10X, and being complementary to its high-performance oscillator suite, which contributes the majority of revenue. Customer breadth and diversity will also increase substantially, as it will now integrate the unit’s 10 hyperscalers, seven AI server leaders, 10 networking and communications vendors and other customers to its roster. Because of the complementary nature of SiTime’s oscillators with the unit’s clocking portfolio, management expects there will be minimal product overlap, which will open the door for new revenue opportunities at shared customers, such as cross-selling or integrated oscillator and clocking solutions. 

In Q4, SiTime’s CEO touched on potential revenue goals post-acquisition and set some mile markers for investors further down the line. The first goal post-acquisition is to create a $1 billion company, which is now just ~12% away after combining implied revenue of $588 million with the $300 million expected in the 12 months following the close of the acquisition. From there, they provided a TAM of $10 billion to $11 billion for the timing business with a longer-term total addressable market of $17 billion to $18 billion. 

Financials 

Revenue Accelerates to 88.3% YoY  

SiTime reported Q1 2026 revenue of $113.57 million, beating consensus estimates by 9.1%. Growth accelerated to 88.3% YoY, up from 66.3% YoY in Q4 2025, marking a re-acceleration in the top line for the second consecutive quarter after deceleration through mid-FY25. On a sequential basis, revenue was essentially flat at +0.2% QoQ, an atypical break from Q1’s seasonal declines in the teens to twenties. 

Looking ahead, management guided Q2 2026 revenue to be $140 million to $150 million, implying YoY growth of 108.6% YoY and 27.7% QoQ growth at the midpoint, beating estimates by a solid 29.1%.  

Management guided full year revenue growth of at least 80%, beating estimates by 21%. Beth Howe, Chief Financial Officer, said in the earnings call, “For the full year, we are increasing our revenue growth expectations to at least 80%, well above our prior expectations and our long-term target growth rate of 25%-30%. This step change in growth reflects both the depth of our order book and the confidence customers are signaling in their own demand forecast, particularly in CED.” 

Key Segments 

CED Dominance; Consumer Faces Headwinds 

The quarter's result was driven by continued momentum in the CED (Communications, Enterprise & Datacenter) segment, which reached $75.7 million — up 158% YoY and 17% QoQ — reinforcing SiTime's positioning as a key beneficiary of AI infrastructure buildout. CED now constitutes 67% of total revenue, up from 57% in Q4 2025.

Auto, Industrial & Aerospace revenue came in at $21.2 million, up 51% YoY but declining (13%) QoQ, reflecting some normalization after a 21% sequential growth in Q4. Within this sector, aerospace and defense were the fastest-growing area with all three subsectors benefiting from the accelerating adoption of precision timing across autonomous systems, defense modernization, and industrial automation. 

Consumer, IoT & Mobile revenue of $16.7 million declined (1%) YoY and (31%) QoQ, reflecting ongoing softness in the consumer end market. 

Margins 

Margins are improving primarily due to favorable product mix, cost controls, and operating leverage.  

Q1 adjusted gross margin improved by 7.1 percentage points YoY to 64.5%. The improvement was driven by two factors. Roughly half of the increase was driven by favorable product mix of higher margin products, reflecting strong CED growth, which carries higher above average gross margin, combined with a lower mix of consumer products. The other half was driven by product cost improvements and leverage. Management guided adjusted gross margin of 65% in the next quarter. 

Q1 operating loss was ($12.3 million) or (10.9%) of revenue compared to ($28.1 million) or (46.6%) of revenue in the same period last year. Q1 adjusted operating income was $31.8 million or 28% of revenue compared to a mere $2.1 million or 3.4% of revenue in the same period last year, reflecting strong operating leverage. Management guided Q2 adjusted operating margin to further improve to 32.9%. The difference between GAAP operating margin and non-GAAP operating margin was due to high stock-based compensation, which was 27.1% of revenue in Q1. 

Q1 adjusted net income was $38.9 million or 34.3% of revenue compared to $6.3 million or 10.5% of revenue in the same period last year.  

Management also offered some more clarity on how margins will evolve through the year, with a higher mix of CED benefitting 1H, before a higher mix of consumer weighs a bit more on 2H:   

“We certainly benefited in Q1 from kind of the double benefit of a stronger mix of CED, which has those higher gross margins and a lower mix of consumer. As we move through the year, we would expect consumer to be a larger portion of the mix in the back half, which might modulate gross margins a bit just based on mix. Overall, we still expect gross margins to be above that 60% level and kind of well into this range. It may modulate a bit, but still, very toward the higher end of our target range.” 

Q1 Adjusted EPS grew by 454% 

Q1 adjusted EPS grew by 453.8% YoY to $1.44, beating estimates by 21.4% primarily due to strong operating leverage.  

Management also provided a strong Q2 adjusted EPS guide of $1.85 to $2.00, implying a YoY growth of 309.6%, beating estimates by a stellar 65.9%. Looking ahead, 2026 full year adjusted EPS is expected to grow by 81.7% YoY to 5.81 and 33.6% YoY to $7.77 in 2027.  

Cash Flows and Balance Sheet 

The company also reported strong cash flows primarily driven by higher profits. 

• Q1 operating cash flows grew by 108% YoY to $31.2 million or 27.5% of revenue compared to 24.9% of revenue in the same period last year. 
• Q1 free cash flow was $17.9 million or 15.7% of revenue compared to ($1.4 million) or (2.3%) of revenue in the same period last year. 
• The company also maintains a strong balance sheet of $788.6 million of cash & short-term investments with no debt at the end of Q1 2026. 
• Inventories increased by 11.6% QoQ to $91.1 million, suggesting demand visibility and preparation for the anticipated Q2 ramp. 

Conclusion 

SiTime is seeing a clear inflection in its CED segment with 158% YoY and 17% QoQ growth in Q1. Management sees strong tailwinds due to a mix of increased unit volume of 3X from inference and higher ASPs, especially as we approach 1.6T. The acquisition of Renesas’s timing unit is expected to boost the company’s presence across the data center with new customer additions, while providing another lever for CED to expand. 

The company also offers a 65% gross margin, 30%+ operating margin and an 80% revenue growth guide for the year – with a healthy supply chain as the cherry on top. The setup in AI networking stocks moves quickly. SiTime is not for the passive investor and will require an active stance.

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.

Recommended Reading:

• Seagate FQ3: Data Center Growth Accelerates to 12% QoQ, Mozaic4+ Ramping
• NextEra: Pivoting Beyond a Regulated Utility Player
• Silicon Motion: Strong Consumer SSD Demand, Trying to Move into AI Enterprise Markets for 2027-2028
• Fabrinet: EML Constraints Easing, Capacity Expansion Adding >$2.6 Billion Revenue Potential

MEMS timing supplier SiTime is seeing solid tailwinds in AI data centers from the increasing complexity of rack-scale platforms and shift to faster data rates in networking switches and optical transceivers, as these place more emphasis on timing solutions to ensure that all components operate as one cohesive unit with maximum performance and reliability.  

SiTime delivered a strong print this week by dramatically beating Q1 estimates, guiding far above Q2 estimates and raising full-year guidance. Q1 revenue of $113.6 million beat consensus of $103.5 million for a 10% beat, yet EPS grew 5X YoY and reported a 23% beat for $1.44.  

Gross margin was especially strong at 64.5% for a 7-point expansion, but the more pronounced margin expansion was operating margin at 28% for a 25-point expansion. Operating cash flow more than doubled to $31.2 million. 

While Q1’s strength is notable, the more pronounced beat is in the Q2 guide for revenue of $145M at the midpoint for growth of 100% YoY. This compares to analyst estimates for 62% growth in Q2. Management also stated gross margin would be above 65% and operating margin above 30% next quarter.  

This led to raising the full year fiscal year guide, with management stating: “For the full year, we are increasing our revenue growth expectations to at least 80%, well above our prior expectations and our long-term target growth rate of 25%-30%. This step change in growth reflects both the depth of our order book and the confidence customers are signaling in their own demand forecast, particularly in CED. That confidence is translating into improved visibility, reinforcing our expectation for sustained momentum throughout the year.” 

Brief Overview of Key Products 

SiTime offers a range of MEMS-based (micro-electromechanical system) timing solutions that it says offer greater resilience, lower power, higher performance and a smaller size versus traditional quartz solutions.  

• Oscillators 

MEMS oscillators are timing devices embedded on a silicon chip that generate a clock signal (frequency) used to coordinate actions of different components, essentially serving as the ‘heartbeat’ of the device; they do this by combining clocks and resonators in a single system. As it relates to AI server buildouts, increasing rack and cluster sizes means data must move across hundreds to thousands of chips at once, requiring precise synchronization across components and interconnects to minimize latency, prevent data loss and maximize system efficiency.  

SiTime’s high-performance oscillators are prevalent across the compute tray within the GPU and CPU boards, NIC cards, and networking switches, and also within the networking fabric, from top-of-rack and spine switches to optical transceivers and AECs. SiTime says its MEMS oscillators can reduce power consumption by 30–50% versus quartz with similar or better frequency in a more compact footprint. 

• Clock ICs – Generators and Buffers  

Clock buffers take a single clock signal input and then multiply and boost the signal across multiple different output lines, ensuring that the clock signals reach all parts of the circuit. Clock generators are a type of oscillator that produce periodic timing signals for all of the components within the circuit, and can handle different frequencies needed for GPUs, CPUs, memory, PCIe and Ethernet components. Generators also help reduce jitter and improve signal integrity on the board.  

SiTime recently launched its Chorus clock-system-on-a-chip family for AI data center applications in April 2024, which combines clocks, oscillators and resonators into a single integrated chip, offering up to 10X higher performance compared to standalone oscillators and clocks. SiTime executives say the new solution also accelerates development time by up to six weeks, and reduces board area for timing by up to 50%, while addressing issues such as noise and impedance mismatch between resonators and clocks. Back in Q3, the clock funnel was stated to have quadrupled in the past year to $300 million, though revenue was stated as sub-$20 million in December. 

SiTime is also acquiring Renesas’ timing unit, which it says will boost its clock IC portfolio by 10X – more on this below.  

• Resonators 

Resonators are key components within oscillators that vibrate at a stable frequency, essentially setting the frequency of the oscillator, determining frequency accuracy and ensuring stability over a range of temperatures. SiTime recently launched its high-performance Titan resonator family in September, which it says offers improved performance under high shock and vibration, while occupying 4.2x and 12x less PCB space than quartz competitors. The new platform is initially geared towards IoT, wearables and medical device applications.   

Precision Timing Products Accelerate from Inference Market 

SiTime’s Communications, Enterprise and Datacenter segment (CED) grew 158% YoY and 17% QoQ for the eight consecutive quarter of triple-digit growth. According to the earnings call, the primary driver for CED strength is the shift from training to inference with newer XPUs requiring 2X to 4X more content per system than previous training workloads. According to management, utilization rates in inference workloads are running 20% to 40% today yet need to reach 50% to 60% for reasonable ROI on capex. 

Notably, there is increased unit volume combined with higher ASPs on the Elite and Elite 2 product line.  

Here is what was stated on the call: 

“On inference infrastructure built on newer XPU, it needs 2 to 4 times more timing content per system than in training infrastructure. GPU utilization in inference workloads is now 20% to 40% and is targeted to get to 50% to 60%. Here, time synchronization plays a critical role in achieving higher GPU utilization and SiTime benefits from its products being used in this application. This emphasis on synchronization is driving demand for high ASP and high-margin products.  

Elite and Elite RF Super-TCXOs are widely deployed in AI infrastructure, and we have recently exceeded and extended our leadership with the new Elite 2 Super-TCXO family. This newer Elite 2 delivers up to 3 times better synchronization performance compared to Elite, which was already significantly better than quartz oscillators.” 

Additionally, as 1.6T ramps, SiTime foresees additional share gains due to higher frequencies and “tighter resilience requirements” driving demand for advanced oscillators. 

The following was stated: 

“As hyperscalers increase networking bandwidth within the data center, we expect to see meaningful adoption of 1.6T optical modules in 2026. Higher frequencies and the need for more resilient performance are driving demand of our advanced oscillators at a higher price than those used in 800G. At the same time, we expect to see continued strong shipment for oscillators for 400G and 800G for at least the next two years.” 

As stated above, management noted that 400G and 800G will remain strong for the next wo years while 1.6T ramps. However, 1.6T will see higher ASPs than 800G with the CEO stating that SiTime can charge a premium price by being the highest-performing option: “This newer Elite 2 delivers up to 3 times better synchronization performance compared to Elite, which was already significantly better than quartz oscillators.” 

Quantifying Inference System Content Opportunities  

SiTime has offered some clues into holistic dollar content per rack, general content per networking components, as well as commentary on how the above trends are shaping content growth. This provides a bit more insight into where the 2X to 4X increase in content per system with inference deployments could land. 

In terms of the holistic dollar content per rack, management has explained that for training platforms where they have a high penetration across the system and networking topology, content opportunities “can be multiple hundreds of dollars in a fully integrated rack,” with opportunities potentially scaling larger in networking fabrics.  

Translating this ‘multiple hundreds of dollars’ to the 2X to 4X increase with inference roughly estimates that SiTime could see content above $1,000 in fully integrated racks for inference deployments — this was also mentioned by analysts in Q1’s call that content “certainly sounds like it could reach into the $1,000+ range.” 

Additionally, to briefly touch upon networking opportunities moving up the stack, SiTime has previously mentioned that content for some optical modules can vary from $1 to $2 ASPs, but shifting to switches and farther up the stack can drive more meaningful content, along the lines of $7 to $10 ASPs.   

CPO Switches to Drive 3X Timing Content 

We’ve published quite a bit on the CPO opportunity, especially in our Coherent and Lumentum analyses. As data centers migrate from pluggables to NPO/CPO, SiTime can benefit from this shift as optics move inside the switch. The result will be more oscillator sockets per switch and higher performance requirements, resulting in 3X higher timing dollar content. Management stated: “On CPO or co-packaged optics, in our discussion with customers, we see even greater strength. For example, in CPO switches, where timing content can be up to 3 times higher.” 

Regarding supply to serve an influx of demand, SiTime also offers high confidence commentary that they have no bottlenecks with an analyst referencing SiTime having a strong supply chain during 2020, unlike many peers: 

“We see no issues around supply chain in particular. I know some people have said that in the past, other semiconductor companies, so we want to be very clear about that. We see strength in our supply chain, and we don’t see any fundamental issues or macro issues or external issues that can trip us up as of now.” 

Book-to-Bill Accelerating 

SiTime does not typically offer its book-to-bill ratio, but brief commentary from Q1 that book-to-bill is growing with pull-through from CED, taken with Q4’s book-to-bill of >1.5X, suggests this ratio is moving higher.  

When management had provided the book-to-bill in Q4, analysts had questioned on the duration of this backlog, and if it would be six, 12 or 18 months and beyond, to which management said it is typically within 12 months:  

“So in terms of the book-to-bill, I think Rajesh talked about the fact that we are seeing customers maybe book out a little longer, but typically, that's well within 12 months. We see a lot of ordering over the next couple of quarters. But we are seeing some customers book meaningfully in the second half already as well. But I would say definitely weighted to Q1 and Q2 in terms of that.” 

This implies near-term demand is strengthening in Q1, driven by CED, while it provides a further layer of confidence in SiTime’s upbeat annual revenue growth guide of >80%. It also suggests Q2 momentum is likely to remain robust, and could signal a similarly strong strong 2H if orders continue to flow as Nvidia’s Blackwell Ultra and Rubin ramp throughout the year alongside strong potential growth in 1.6T transceiver volumes. 

Telecom Offers Diversity for AI-Driven Demand 

Worth noting is that SiTime sells into the telecom industry, to help diversify its customer base beyond hyperscalers. While telecom has gone through a significant trough in recent years, the industry is expected to see an AI-driven resurgence as workloads run at the edge and in the access network. The key markets that SiTime can benefit from are RAN optimization, edge AI inference at base stations, and Open RAN architectures. Each of these trends require more timing sockets and more precision timing requirements, leading to a 3X uplift in content: 

“Finishing up on the telecom part of CED, we see increasing convergence between AI and advanced telecom infrastructures, especially in 5G RAN or Radio Access Network and demand from new applications such as FWA or fixed wireless access. AI-enabled telecom designs contain 3 times higher timing content, primarily from high ASP oscillators and clocks” 

Acquisition of Renesas’ Timing Unit 

SiTime is acquiring Renesas’ timing unit for ~$1.5 billion, significantly increasing its clocking portfolio by ~10X, adding a range of hyperscaler and leading AI server customers, and providing a substantial boost to SiTime’s CED revenue.  

Most importantly, the timing unit acquisition is expected to significantly increase the scale of SiTime’s CED business. Management had explained in Q4 that the acquisition will nearly double its CED business, adding that in the first 12 months post-close (likely starting Q2 ’26), the timing unit is expected to generate more than $300 million in revenue with ~75% of that from CED, or ~$225 million.  

Moving down the line, the timing unit is expected to accretive to both margins and EPS in the first full year post-close. Management explained that the unit has adjusted gross margins around 70%, or nearly 9 points higher than SiTime’s Q4 adjusted gross margin of 61.2%; it will also help push SiTime towards the upper end of its long-term 60-65% gross margin target model. The acquisition is expected to help drive adjusted operating margins above 30% from increased operating leverage at scale, compared to FY25’s 17.9% margin. 

From the product and customer side, SiTime sees the acquisition taking them to scale in clocking, adding 500 differentiated clock products to its portfolio, boosting it by 10X, and being complementary to its high-performance oscillator suite, which contributes the majority of revenue. Customer breadth and diversity will also increase substantially, as it will now integrate the unit’s 10 hyperscalers, seven AI server leaders, 10 networking and communications vendors and other customers to its roster. Because of the complementary nature of SiTime’s oscillators with the unit’s clocking portfolio, management expects there will be minimal product overlap, which will open the door for new revenue opportunities at shared customers, such as cross-selling or integrated oscillator and clocking solutions. 

In Q4, SiTime’s CEO touched on potential revenue goals post-acquisition and set some mile markers for investors further down the line. The first goal post-acquisition is to create a $1 billion company, which is now just ~12% away after combining implied revenue of $588 million with the $300 million expected in the 12 months following the close of the acquisition. From there, they provided a TAM of $10 billion to $11 billion for the timing business with a longer-term total addressable market of $17 billion to $18 billion. 

Financials 

Revenue Accelerates to 88.3% YoY  

SiTime reported Q1 2026 revenue of $113.57 million, beating consensus estimates by 9.1%. Growth accelerated to 88.3% YoY, up from 66.3% YoY in Q4 2025, marking a re-acceleration in the top line for the second consecutive quarter after deceleration through mid-FY25. On a sequential basis, revenue was essentially flat at +0.2% QoQ, an atypical break from Q1’s seasonal declines in the teens to twenties. 

Looking ahead, management guided Q2 2026 revenue to be $140 million to $150 million, implying YoY growth of 108.6% YoY and 27.7% QoQ growth at the midpoint, beating estimates by a solid 29.1%.  

Management guided full year revenue growth of at least 80%, beating estimates by 21%. Beth Howe, Chief Financial Officer, said in the earnings call, “For the full year, we are increasing our revenue growth expectations to at least 80%, well above our prior expectations and our long-term target growth rate of 25%-30%. This step change in growth reflects both the depth of our order book and the confidence customers are signaling in their own demand forecast, particularly in CED.” 

Key Segments 

CED Dominance; Consumer Faces Headwinds 

The quarter's result was driven by continued momentum in the CED (Communications, Enterprise & Datacenter) segment, which reached $75.7 million — up 158% YoY and 17% QoQ — reinforcing SiTime's positioning as a key beneficiary of AI infrastructure buildout. CED now constitutes 67% of total revenue, up from 57% in Q4 2025.

Auto, Industrial & Aerospace revenue came in at $21.2 million, up 51% YoY but declining (13%) QoQ, reflecting some normalization after a 21% sequential growth in Q4. Within this sector, aerospace and defense were the fastest-growing area with all three subsectors benefiting from the accelerating adoption of precision timing across autonomous systems, defense modernization, and industrial automation. 

Consumer, IoT & Mobile revenue of $16.7 million declined (1%) YoY and (31%) QoQ, reflecting ongoing softness in the consumer end market. 

Margins 

Margins are improving primarily due to favorable product mix, cost controls, and operating leverage.  

Q1 adjusted gross margin improved by 7.1 percentage points YoY to 64.5%. The improvement was driven by two factors. Roughly half of the increase was driven by favorable product mix of higher margin products, reflecting strong CED growth, which carries higher above average gross margin, combined with a lower mix of consumer products. The other half was driven by product cost improvements and leverage. Management guided adjusted gross margin of 65% in the next quarter. 

Q1 operating loss was ($12.3 million) or (10.9%) of revenue compared to ($28.1 million) or (46.6%) of revenue in the same period last year. Q1 adjusted operating income was $31.8 million or 28% of revenue compared to a mere $2.1 million or 3.4% of revenue in the same period last year, reflecting strong operating leverage. Management guided Q2 adjusted operating margin to further improve to 32.9%. The difference between GAAP operating margin and non-GAAP operating margin was due to high stock-based compensation, which was 27.1% of revenue in Q1. 

Q1 adjusted net income was $38.9 million or 34.3% of revenue compared to $6.3 million or 10.5% of revenue in the same period last year.  

Management also offered some more clarity on how margins will evolve through the year, with a higher mix of CED benefitting 1H, before a higher mix of consumer weighs a bit more on 2H:   

“We certainly benefited in Q1 from kind of the double benefit of a stronger mix of CED, which has those higher gross margins and a lower mix of consumer. As we move through the year, we would expect consumer to be a larger portion of the mix in the back half, which might modulate gross margins a bit just based on mix. Overall, we still expect gross margins to be above that 60% level and kind of well into this range. It may modulate a bit, but still, very toward the higher end of our target range.” 

Q1 Adjusted EPS grew by 454% 

Q1 adjusted EPS grew by 453.8% YoY to $1.44, beating estimates by 21.4% primarily due to strong operating leverage.  

Management also provided a strong Q2 adjusted EPS guide of $1.85 to $2.00, implying a YoY growth of 309.6%, beating estimates by a stellar 65.9%. Looking ahead, 2026 full year adjusted EPS is expected to grow by 81.7% YoY to 5.81 and 33.6% YoY to $7.77 in 2027.  

Cash Flows and Balance Sheet 

The company also reported strong cash flows primarily driven by higher profits. 

• Q1 operating cash flows grew by 108% YoY to $31.2 million or 27.5% of revenue compared to 24.9% of revenue in the same period last year. 
• Q1 free cash flow was $17.9 million or 15.7% of revenue compared to ($1.4 million) or (2.3%) of revenue in the same period last year. 
• The company also maintains a strong balance sheet of $788.6 million of cash & short-term investments with no debt at the end of Q1 2026. 
• Inventories increased by 11.6% QoQ to $91.1 million, suggesting demand visibility and preparation for the anticipated Q2 ramp. 

Conclusion 

SiTime is seeing a clear inflection in its CED segment with 158% YoY and 17% QoQ growth in Q1. Management sees strong tailwinds due to a mix of increased unit volume of 3X from inference and higher ASPs, especially as we approach 1.6T. The acquisition of Renesas’s timing unit is expected to boost the company’s presence across the data center with new customer additions, while providing another lever for CED to expand. 

The company also offers a 65% gross margin, 30%+ operating margin and an 80% revenue growth guide for the year – with a healthy supply chain as the cherry on top. The setup in AI networking stocks moves quickly. SiTime is not for the passive investor and will require an active stance.

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.

Recommended Reading:

• Arm: Computex Update, CPU Core Demand Hinted at Being Higher
• CoreWeave: Revenue Inflecting in 2H, Margin and Profit Questions
• Core Scientific: Multi-GW Pipeline, New Hyperscaler Interest but Still Tied to CoreWeave
• MaxLinear: Optical Data Center Demand Accelerating, Margins to Improve in Q2

Applied Optoelectronics (Nasdaq: AAOI) has been one of our portfolio's biggest winners, up 350% year-to-date and 450% over the past six months. It hasn't been an easy name to accumulate, given the stock's significant volatility, driven in part by lumpy timing of AAOI's revenue.

Q1 results came in roughly in line with other AI networking companies that have reported this week, with revenue of $151.1M up 51% year-over-year and 13% sequentially. The more material development was the forward guide: management now expects 2026 revenue above $1.1B, well above the prior consensus of $962M and compared to $455.7M in 2025. The updated guide implies 141% year-over-year growth, and with Q2 guided to $180M-$198M, the math points to a heavily back-weighted year as clearly the bulk of the revenue is expected to arrive in the second half. 

Significant Ramp Expected Q3 2026 through Q2 2027 

AAOI joined a growing list of AI networking management teams talking about a very strong 2H. It’s unlikely the Street rewards forward-looking guidance for two quarters out especially with a tricky supply chain environment. However, it's worth a minute to look at the acceleration that AOI is guiding to, with the model below showing about 70% QoQ growth between Q3 and Q4.

Management sees an even further ramp into 2027 with an indication they could see $471M in revenue per month at full utilization.  The revenue ramp aligns with what we covered last quarter, which was a roughly 5X increase in monthly transceiver output. 

However, analysts on the call were cautious as they pointed toward limited InP capacity as an industry-wide issue that AAOI has to overcome to reach these targets. Typically, for an aggressive forecast, analysts also want to see progress in the current quarter, but Q1 offered 13% QoQ growth (a solid print but not helpful for the forecasted numbers that are much higher). Overall, management has some serious execution milestones to reach in the near future, and that was the overall tone on the call.  

Here is a sample of what was stated on the call regarding the well-known InP shortages: “Jarren, we see a shortage of indium phosphide laser manufacturing capacity across the industry right now, and we think that's going to persist and even get more acute with the advent of ELSs, as Thompson mentioned. That's why we see this need to really expand our — phosphide fabrication capability pretty dramatically over the next 12 to 18 months.” 

We've covered indium phosphide at length for a few years, primarily in our Lumentum and Coherent analyses. InP is the substrate that produces high-speed lasers for optical transceivers, and is a major constraint in the networking industry right now. According to management, they have almost 1 year of inventory on hand and are in discussions with “a good line of sight” to avoid a shortage. Here was the tone on the call, which I’d characterize as rather vague: 

“Chih-Hsiang Lin, CEO 

Right now, we record — supplier with some kind of discussion — sorry, not much we can say. But 4 of them are outside of China. So I would say right now, we should have enough inventory minimal almost 1 year. But since the volume, we increased surface, we are making calls with all the suppliers. 

Stefan Murry, CFO 

I would say we've got good line of sight into how we think we can not see the shortage there. But we can't say too much about it specifically at this point because a lot of it is under discussions.” 

In another exchange, AAOI offered an important point, which is the company has been around since 1990 and is vertically integrated with very little dependency on external suppliers. As stated, they’ve made lasers in-house for a few decades, and they call this out as the single biggest reason they have avoided shortages compared to competitors. The company also stated competitors face lead times of 21 to 24 months, offering AAOI an important incumbency.  

“Stefan Murry, CFO: 

Right. Great question. So as I said earlier, I think indium phosphide capacity is critical right now. The fact that we have our own in-house laser manufacturing capability is one of our key advantages. Certainly, when you talk to customers, that's one of the big things that they like about us, especially now that we're seeing shortages across the industry. So our fab expansion is well underway. 

As Thompson mentioned, we've got a number of critical pieces of equipment and coding machines and others that are in various stages of either being delivered or being qualified. It does take a pretty extended period of time to qualify a new piece of laser manufacturing equipment, as you can imagine, you don't want to take a risk of having an unknown quality issue there. 

So a lot of that is already here and already undergoing qualification or it's very close to being here. And that's why we can be pretty confident that our capacity is going to be where we need it to be.” 

An area of confusion that management cleared up on the call is that quoted capacity numbers take about 1-2 quarters to recognize in revenue. That is why management discussed 100K units per month yet will not see that revenue until Q3. 

800G and 1.6T Product Mix 

When we examine further how AAOI can guide for 70% growth QoQ at the midpoint in 2H, it helps to look at current product mix versus anticipated product mix.  

In the current quarter, only $4.6M was from 800G for 5.6% of data center revenue and 0% of revenue was 1.6T. The far majority of the data center was still 400G, which grew 10X year-over-year.  

Over the 12-18 months, management is building toward 46% of revenue driven by 800G and 35% of revenue driven by 1.6T. In other words, two product categories that currently represent 3% of revenue will represent 80% of revenue. 

Here is what that ramp looks like in terms of transceiver numbers: 

“Exiting Q1, our total manufacturing capacity approached 100,000 units per month of 800G and 1.6 terabit capacity. Looking ahead, we expect to continue to rapidly expand our production capacity to approach 150,000 per month of 800G and 1.6 terabit this quarter. As a reminder, we expect by the end of this year that we will be capable of producing over 650,000 pieces of 800G and 1.6 terabit products per month with about 30% of that output coming from Texas as we expand into additional facility space and bring new production online. 

By the end of next year, 2027, we expect to grow our production capacity to be able to produce over 930,000 pieces of 800G and 1.6 terabit products per month, with over half of that output coming from Texas. These investments reflect measured scaling of our footprint while aligning with our strong and growing customer demand and qualification progress across both 800G and 1.6 terabit products.” 

Last quarter, management announced its first 1.6T volume order from a hyperscaler, and added another 1.6T hyperscaler plus two 800G customers in Q1. Also important to note, management stated that 1.6T will run on the same production lines as 800G and even 400G – which means they do not have to build a new factory, and can offer speed to market for a concurrent ramp: “As I mentioned, this automation platform is also highly flexible, enabling us to produce across multiple generations from 400G to 800G to 1.6 terabit using many of the same techniques and equipment.” 

ELS Lasers for the CPO Opportunity 

The ELS thesis is that CPO will drive sustained demand for high-power lasers, which is exactly the product category where AAOI's vertical integration is most differentiated. To meet that demand, the company plans to expand laser fabrication capacity in Texas by roughly 350% by the end of 2027. 

“We believe that in the future, CPO will continue to drive increased demand for high-power lasers, and we plan to continue to expand our laser manufacturing capacity in Texas in order to accommodate these future growth drivers. We expect to further expand our laser fabrication capacity by around 350% by the end of 2027.” 

The specific bet is on External Laser Source (ELS) modules, which is the remote laser packages that provide light into co-packaged optical engines. ELS is a smaller market than transceivers today, but is expected to scale with CPO adoption. 

Financials: 

By Royston Roche 

Q1 Revenue grew by 51% 

AAOI Q1 revenue grew by 51.4% YoY and 12.6% QoQ to $151.1 million. However, missed estimates by 1.8%. Revenue growth accelerated by 17.5 percentage points from 33.9% YoY growth and 13.2% QoQ growth in the previous quarter. 

Management guided Q2 revenue in the range of $180 million to $198 million, implying a YoY growth of 83.6% and 25% QoQ at the midpoint. Missed estimates by 1.9% as growth pushed to 2H. The more material development was that the management raised full-year 2026 revenue guidance to over $1.1 billion, up from the prior guidance of over $1.0 billion issued during Q4 results, implying 141.4% YoY growth for the full year. 

Key Segments 

Data Center Revenue Growth of 154%  

Q1 Data Center revenue grew by 154% YoY and 8.7% QoQ to $81.4 million. 100G products revenue increased by 36% YoY, while sales for the 400G products increased tenfold YoY. In the first quarter, 41.9% of data center revenue was from 100G products; 46.7% was from 200G and 400G products, 800G transceiver products accounted for 5.6% of revenue, and 5.6% was from 10G and 40G transceiver products. Management expects a sequential increase in Data Center revenue in the next quarter. 

CATV Revenue 

CATV revenue grew by 3.6% YoY and 23.8% QoQ to $66.8 million. The revenue came close to the higher end range of the guidance range of $61 million to $67 million. Looking ahead, management expects CATV revenue to be between $75 million and $80 million for Q2, implying a YoY growth of 38.3% and 15.9% QoQ at the midpoint. Looking further ahead, management expects to generate over $325 million annually in CATV. The vast majority of the CATV revenue this year is expected to be amplifiers, and they do anticipate generating some revenue from the software solutions this year. 

Telecom/Other Revenue 

Telecom revenue was down (12.9%) YoY and (49.9%) QoQ to $2.6 million. While the other revenue, which is negligible, was down (8.6%) YoY and up 19.3% QoQ to $0.34 million. 

Margins 

Q1 adjusted gross margin was down 150 basis points YoY to 29.2% due to higher data center revenue and missed the guidance of 30%. Management remains committed to its long-term goal of returning 40% adjusted gross margin and the CFO, Stefan Murry, said in the earnings call,  “While we do expect continued gradual improvement in gross margins, we continue to expect that the revenue mix in data center in the short term will be a slight headwind. We remain committed to our long-term objective of returning non-GAAP gross margins to around 40% and believe that this goal is achievable as our mix shifts towards higher-margin products and as we capture additional efficiencies across our operations.” 

Q1 adjusted operating loss was ($7.3 million) or (4.8%) of revenue compared to ($4.8 million) or (4.8%) of revenue in the same period last year. 

Q1 adjusted net loss was ($4.9 million) or (3.3%) of revenue compared to ($0.9 million) or (0.9%) of revenue in the same period last year. Management has guided Q2 adjusted net income in the range of a loss of ($2.5 million) to income of $2.8 million, with a midpoint of $0.20 million or 0.1% of revenue.

EPS 

The company reported Q1 adjusted EPS of ($0.07) and missed the estimates of ($0.05) primarily due to higher data center revenue mix. Management has guided Q2 adjusted EPS in the range of ($0.03) to $0.03, the midpoint implies breakeven in the next quarter. 

Cash Flow and Balance Sheet 

The company’s cash flows were weak in Q1 due to high working capital and capex to support future growth. 

• Q1 operating cash outflow was ($85.4 million) or (56.5%) of revenue compared to ($50.9 million) or (51%) of revenue in the same period last year. 
• Q1 free cash outflow was (143.6 million) or (95%) of revenue compared to ($87.2 million) or (87.3%) of revenue in the same period last year. Capex increased 105.1% YoY to $58.2 million. 
• The company had cash & short-term investments of $449.4 million compared to debt of $206.5 million. The company issued shares worth $382.5 million in Q1. 
• Inventories grew by 12.6% QoQ to $206.2 million to support future growth.

Conclusion: 

My bull case for AAOI is that hyperscaler demand for high-speed optics is intense, and has outpaced the industry’s ability to make them. AAOI offers vertical integration on lasers and is publicly committing to a significant step-up in revenue come 2H with over 141% growth this year.  

However, the stock was down after hours because Q1 offered average growth for an AI networking stock, forcing investors to wait for the next blowout quarter. Beyond that, any small slip in InP capacity or equipment delivery could trigger an adjustment to expectations. 

Or the opposite could happen and AAOI could execute flawlessly on all fronts.

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 AAOI at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

• Arm FQ4: AGI CPU Demand Hits $2B, Revenue Outlook Stays at $1B
• Coherent FQ3: InP Capacity Doubling to Drive CY26 Inflection
• Lumentum FQ3: Firing on All Cylinders Despite Stiff Supply Constraints Across EMLs, Pump Lasers
• Astera Labs: Important QoQ Acceleration, Product Road Map is Loaded

Our main takeaway from Lumentum’s Q2 was “capacity constrained and loving it”, and that theme was just as evident, if not more so, this quarter. Supply-demand imbalances for EMLs widened, transceivers face a similarly large imbalance, but the largest supply constraint arose in an unexpected area – pump lasers for DCI.  

For OCS and CPO, revenue remains modest for both, though Lumentum expects to satisfy its $400 million OCS target in calendar 2H 2026 and realize the first significant scale-out CPO revenue in calendar Q4 (FQ2 27). More importantly, the scale-up CPO opportunity was discussed as multiple times larger than scale-out, with output significantly larger than the company’s new fab capacity, with expectations it will add >$5 billion in incremental revenue capacity as it ramps in 2028. Management stated CPO will be a significant driver for meeting their $2B per quarter revenue goal that was announced at the OFC event. 

Fundamentally, Lumentum is firing on all cylinders, with YoY growth forecast to accelerate to 105% YoY and sequential growth guided to maintain >20% QoQ for a third straight quarter in FQ4. Margins showed strong expansion, with GAAP gross margin up more than 15 points YoY to 44.2% and GAAP operating margin up more than 30 points YoY to 21.6%.   

200G EML Revenue Doubled QoQ, Supply-Demand Imbalance Widening 

As we had noted in last quarter’s write-up, Lumentum is benefiting from outsized demand for its EML lasers, reaching a quarterly company record in EML laser shipments with 200G ramping faster than expected. Lumentum reached another quarterly record for EML shipments in FQ3, driven by 100G but with 200G EML revenue more than doubling QoQ.  

While Lumentum remains capacity constrained in EML, the company is working quickly to expand in Japan, noting that it expects to achieve >50% YoY growth in EML units by the December 2026 quarter versus the December 2025 baseline. Layering in higher-ASP 200G EML units in the back half of 2026 is likely to drive revenue at a higher rate than the >50% YoY growth in capacity. However, the primary challenge with this capacity expansion is that it may not be coming soon enough to alleviate widening supply shortages:  

“The supply-demand imbalance is probably even higher than we reported in our last call, somewhere greater than 30%. I think last time we gave a metric of 25% to 30%. We still seem to be behind significantly. We had conversations today with customers, significant customers looking to really up their demand and get output from us, and we simply can't service that.” 

There are a couple puts and takes here – on a positive note, the fact that the supply-demand gap is widening suggests pricing power can drive sequential margin expansion as new capacity comes online in 2H. Additionally, management believes that increasing supply in the near-term is largely within its own control and not reliant on external factors, based on the amount of InP substrates it has secured within the supply chain.  

However, the main challenge is that 2027 EML output represents “a massive step-up just given the scale-out and scale-up demands,” meaning that future capex for capacity expansion, and some InP substrate procurement, will likely be required to help close this supply-demand imbalance. This may pressure free cash flows as Lumentum is already investing heavily to ease pump laser supply tightness.  

Lumentum’s acquisition of the Greensboro fab from Qorvo earlier this year will serve as its fifth InP fab, capable of supporting >$5 billion in annual run rate capacity, though first material contributions from this fab are not expected until 2028. 

Scale-Across Components Grow 120% and 80% YoY 

While the EML constraints are rather widely known at this point in time, it’s important to touch upon pump and narrow linewidth lasers serving scale-across applications. Not only is Lumentum effectively sold out of both for the foreseeable future, but pump lasers were highlighted as an “unanticipated” constraint this quarter.  

Both products serve scale-across applications and witnessed robust growth in Q3, with narrow linewidth lasers recording a ninth consecutive quarter of growth, up 120% YoY, and pump lasers up 80% YoY. However, Lumentum detailed in Q3 that pump lasers are even more constrained than EMLs, with this hitting suddenly: 

“These components remain effectively sold out for the foreseeable future, and we are actively working to secure long-term agreements that will help offset anticipated capital expenditures.” 

For more color on the output front, Lumentum explained that near-term output should rise rather substantially as there are less constraints on the fab front for pumps versus EMLs. This near-term uplift in capacity is crucial in minimizing the supply imbalance, as well as helping meet elevated demand in the near term.  

Because capex is high, Lumentum explained that they are “talking to the major customers around trying to help, right, and put some skin in the game around the CapEx that we're going to try to lay out. One, that can entail prepayment, that can entail take-or-pay, that can entail price increases,” minimizing capital and risk associated with expansion, or in the case of price increases, aiding growth and margins. 

This connects over to what we have discussed regarding a long-haul networking stock for our Discovery members. Narrow linewidth lasers support high-bandwidth, low-power 800G/1.6T coherent pluggables for scale-across and data center interconnect (DCI) applications. Pump lasers are key components in amplifying signal strength over four, eight or 16 fiber pairs simultaneously, essential for DCI, long-haul or subsea links and more so for multi-rail optical systems. For example, by scaling from one rail per module to four, multi-rail systems could deliver up to 32X the density of current single-rail solutions. 

To learn more about this networking stock, the robust demand it is seeing for scale-across applications and upcoming catalysts for 2027, sign up for Discovery here or click to email us at premium@io-fund.com.To learn more about this networking stock, the robust demand it is seeing for scale-across applications and upcoming catalysts for 2027, sign up for Discovery here or click to email us at premium@io-fund.com.sign up for Discovery here or click to email us at premium@io-fund.com.

For Lumentum, solving or at least easing this unanticipated constraint in pump lasers by late 2026 is essential as multi-rail platform content is higher, due to needing more lasers per system. Being able to meet higher levels of multi-rail demand would likely act as a stronger revenue growth and margin lever next year, as management was explicit in pointing out both as significant gross margin drivers.   

1.6T Transceivers Ramping in Q4, Insourcing CW Lasers 

Another bright spot for Lumentum was its transceiver business, accounting for the majority of growth in its Systems segment, which was up 121% YoY and 24% QoQ to $275.1 million, or 34% of revenue. Cloud transceivers grew more than 40% QoQ with record shipments, with this likely largely driven by 800G as the ramp of 1.6T transceivers is slated for FQ4.  

As should be expected by now, Lumentum said that the “the supply-demand imbalance on our own transceivers was somewhere in that ZIP code” of EMLs at >30%. Management said that they could have actually shipped quite a bit more in Q3 and in Q4’s guide had supply constraints for electrical components or laser diodes not been this tight, and that its pricing power suggests the supply-demand imbalance “isn't going to be solved for a while,” shooting down concerns over laser oversupply.  

To help alleviate some of the external laser supply constraints, Lumentum began insourcing CW lasers in Q3, a quarter earlier than originally expected. Insourced supply is expected to scale further in Q4, accounting for ~20% of transceiver modules in the quarter. This pivot is expected to augment transceiver margins as 1.6T ramps, alongside better yields and lower scrap rates.  

Pricing power can act as an important lever in Q4 — having stronger pricing power on 800G while leaning into the 1.6T ramp in Q4 should help further improve margins, as 1.6T already carries higher margins versus 800G.  

OCS Supply Considerably Tighter, Scale-Up CPO Opportunity Multiples Larger 

As we discussed in Q1, OCS and CPO are expected to emerge as strong growth contributors in fiscal 2027, with revenue contribution at the moment remaining modest. OCS is expected to begin contributing more heavily in calendar 2H, with scale-out CPO arising in calendar Q4 (FQ2 27).  

There were a handful of key insights this quarter for both. OCS is now seeing considerable supply tightness. For CPO, Lumentum projects scale-up opportunity to be substantially larger than $5 billion, and believes it could see a faster path-to-market from vertical integration.  

For OCS, the ramp remains largely on track, with management confident in meeting its $400 million target in calendar 2H 2026 and ramping to >$1 billion in calendar 2027. The pace of this ramp will be determined by the supply chain, with Lumentum “experiencing considerable tightness” in OCS due to a substantial step-up in requested output, tied to both new OCS opportunities and likely Google’s upcoming TPU v8 chips (as its key OCS customer).  

This could create some volatility or lumpiness in the ramp phase if the supply chain tightness fails to resolve easily. However, the ramp of TPU v8 later this year could provide additional upside as there is incremental OCS content growth versus TPU v7.  

Moving to CPO, Lumentum noted that its ultra-high-power (UHP) laser ramp is progressing to plan, driving sequential growth in Q3. Meaningful revenue is slated for calendar Q4 (FQ2 27), with Lumentum on track to satisfy its multi-hundred million dollar purchase order in the first half of calendar 2027.  

These near-term opportunities for CPO are primarily for scale-out applications, yet Lumentum foresees the opportunities in scale-up CPO to be multiples larger. And if you weren’t tired of hearing this by now, Lumentum expects a massive supply-demand imbalance with CPO due to scale-up:  

“We will have a massive supply-demand imbalance on CPO. It's going to be very, very significant. We've seen multibillion-dollar orders that we've characterized on previous calls come in mostly on scale-out. 

We expect to scale-up to be significantly more than that in terms of revenue opportunity. I think it's going to be somewhere greater than $5 billion of incremental revenue that we can add [with the new Greensboro facility] if we execute properly.” 

First scale-up CPO shipments are not expected until late 2027, per Lumentum’s OFC briefing, though commentary here suggests that scale-up CPO demand could materialize as Lumentum’s largest revenue driver come 2028 and beyond.  

Financials 

Revenue Accelerates to 90.1% YoY in FQ3 

Lumentum's Q3 FY2026 ending March revenue came in at $808.4 million, missed estimates marginally by (0.2%), but represents a strong reacceleration on a YoY basis from the previous quarter. Revenue grew 90.1% YoY and 21.5% QoQ and accelerated 24.6 percentage points from 65.5% on a YoY basis although decel’d slightly from 24.7% QoQ growth in the previous quarter. 

Sequential dollar growth of $142.9 million reflects the scale of Lumentum's ramp, with the company now approaching the $1 billion quarterly revenue threshold. Management issued a strong guide for Q4 FY2026 of $960 million to $1.01 billion, implying a YoY growth of 104.9% YoY and 21.8% QoQ at the midpoint.  

While this beat estimates by 7.4% and signals that the hyperscaler-driven demand cycle remains firmly intact, the more impressive part is that sequential dollar growth was guided to be higher next quarter despite worsening supply constraints. At the midpoint, Q4’s guide implies nearly $177 million in QoQ dollar growth, driven primarily by transceivers, EMLs, scale-across components (narrow linewidth and pump lasers), and incremental OCS revenue.  

During the OFC conference held in March management also provided the $2.0 billion revenue target to be achieved in the 18 to 24 months period. Management remains confident in reaching this target, leveraging EMLs, scale-across, and upcoming OCS and CPO ramps, with consensus currently expecting Lumentum to reach its $2 billion quarter in December 2027. 

Key Segments 

Components Revenue grew by 77% 

Components revenue grew by 77.3% YoY and 20.2% QoQ to $533.3 million. However, was below the guidance of $536.7 million. Revenue growth accelerated from 68.3% YoY and 17% QoQ growth in the previous quarter.  

As noted above, shipments of the narrow linewidth laser assemblies grew for the ninth consecutive quarter, rising over 120% YoY, while pump laser shipments grew 80% YoY. EML shipments reached another quarterly record led by 100G, while 200G EML revenue more than doubled QoQ. 

Lumentum also shipped twice the number of laser chips compared to the same period last year and on track to achieve more than 50% growth in EML units by the December quarter of 2026 as compared to the same period last year. 

Systems Revenue grew by 121% 

Systems revenue grew by 121.1% YoY and 24% QoQ to $275.1 million. The strong growth was primarily due to the cloud transceivers revenue that grew by over 40% sequentially as the company successfully leverage the expanded manufacturing footprint in Thailand. The supply constraints on critical components are keeping the shipments well below customer demand. 

The company is poised to ramp poised to ramp 1.6T-speed transceiver shipments in FQ4 with a portion of this volume leveraging the company’s own CW lasers. Management highlighted that they are improving transceiver profitability through better yields and lower scrap rates. 

Looking ahead to Q4, Lumentum expects more than half of Q4’s sequential growth to be driven by Components, and the remainder from Systems. 

Margins Showing Pronounced Expansion

One of the most compelling aspects of Q3 FY2026's report is the continued margin expansion primarily driven by better manufacturing utilization, favorable product mix, and operating leverage.  

However, management admits their margins are not as strong as peers due to the transceiver business – although as noted, should improve with 1.6T: “I think we are underperforming peers. We have room to grow. We're getting better. I think we are — we've certainly gotten the lead in terms of design. And now in terms of margin, I think we're improving. We still trail.” 

• FQ3 adjusted gross margin improved by 12.7 percentage points YoY to 47.9% primarily due to better manufacturing utilization, increased pricing on certain products, and favorable product mix. GAAP gross margin was 44.2%. 
• FQ3 adjusted operating margin improved by 21.4 percentage points YoY to 32.2% primarily due to operating leverage along with product mix and improving factory utilization. GAAP operating margin was 21.6%. 
• FQ3 adjusted net income grew by 184.8% YoY to $225.7 million with an adjusted net margin of 27.9% compared to 9.6% in the same period last year.   
• Adjusted EBITDA margin also improved significantly by 19.6 percentage points YoY to 36.3% primarily due to strong operating leverage.  

For FQ4, management expects the adjusted operating margin to further improve to 35.5%, up more than 3 points QoQ and more than 20 points YoY, despite growth being driven by transceivers. Insourcing CW lasers is expected to help improve gross margins on that product line in Q4 as 1.6T layers in, alongside growth in narrow linewidth and pump lasers. 

Looking further ahead, Lumentum has other strings to pull for margin expansion, with management discussing that they will turn to contract manufacturers (like Fabrinet) to improve margins: “The margins that we pay to those contract manufacturers are more than offset by the efficiency and cost benefit that they can drive on common components. So that ends up being a lever for us.” 

EPS Showing Strong Growth Trajectory Ahead 

FQ3 adjusted EPS grew by 315.8% YoY to $2.37, beating estimates by 4.6% reflecting favorable product mix and operating leverage. Management also provided a strong adjusted EPS guide of $2.85 to $3.05 for the next quarter, implying a YoY growth of 235.2% at the midpoint and beat estimates by 9.7%. 

Looking ahead, analysts expect adjusted EPS to grow 200.1% YoY to $3.30 in FQ1 and 138.2% YoY to $3.98 in FQ2. 

Cash Flows and Balance Sheet 

The company’s cash flows improved significantly, driven by higher profits.  

• FQ3 operating cash flow was $203.8 million or 25.2% of revenue compared to an operating cash outflow of ($1.6 million) or (0.4%) of revenue in the same period last year.  
• FQ3 free cash flow was $79.1 million or 9.8% of revenue compared to a free cash outflow of ($64.4 million) or (15.1%) of revenue in the same period last year. 
• The company had cash and short-term investments of $3.17 billion compared to convertible notes of $3.28 billion at the end of the quarter. Cash and short-term investments increased from $1.16 billion at the end of FQ2 primarily due to the $2.0 billion investment by Nvidia in March. 
• Inventories grew by 10.9% QoQ to $632.8 million to support strong growth. 

Conclusion 

Lumentum is firing on all cylinders with revenue growth accelerating more than 25 points sequentially to 90% YoY alongside substantial margin expansion in Q3. The more impressive piece was Q4’s guidance for substantially higher sequential dollar growth for revenue despite supply constraints tightening in EMLs and unexpectedly arising in pump lasers.  

OCS and CPO remain bright spots for future growth, with management expecting both to begin layering in more materially in calendar 2H and calendar Q4, before ramping more significantly in 2027. The scale-up CPO opportunity, while still six to seven quarters away, will be one we’re watching with anticipation as it is expected to be perhaps the largest single upcoming opportunity ahead for Lumentum.

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 LITE at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

• The I/O Fund’s Top 15 Stocks for Q2 2026
• Astera Labs: Important QoQ Acceleration, Product Road Map is Loaded
• AMD Q1: Doubled CPU TAM, Helios Incoming for Q4
• Palantir Q1: Strong Headline Numbers; TCV to be Watched