Month: June 2026

AMD, Nvidia, Arm, Intel: Inside the $120 Billion CPU Gold Rush

Posted on by io-fund

Three months ago, GPUs were all the rage, and the idea that CPUs could challenge GPUs when it comes to AI budgets was unfathomable. The shift in this perception is evident not only in management commentary, but also in CPU design companies and OEMs raising forecasts that are now 2X+ higher, as many of the largest players have stated they did not foresee the magnitude of the surge in CPU demand from agentic AI. 

In just six months, AMD has issued a massive increase to its server CPU market forecast, nearly doubling its expected CAGR to 35%—estimating that the market will eclipse $120 billion by 2030. Arm made a similar announcement in March, projecting that the total addressable market (TAM) for data center CPUs will grow to over $100 billion by its fiscal year 2031 (roughly calendar year 2030). This would represent a more than 4X increase over its current TAM estimate of $24 billion, equating to a 33% CAGR. 

An important shift is driving these forecasts as the AI market transitions away from chatbots, which saw a CPU-to-GPU ratio that was heavily weighted toward GPUs from 2023-2025. As we move into agentic AI, an Intel and Georgia Tech paper has stated that “tool-dominated agentic AI workloads are significantly bottle-necked" with CPUs consuming up to 88% of the end-to-end latency. The paper further concludes that “with better quality GPUs, the bottleneck can swiftly shift more towards CPUs.” 

What Intel and Georgia Tech are referring to, is that to scale agentic AI efficiently, CPU orchestration capacity will need to catch up to GPU reasoning capacity to minimize latency and prevent GPU underutilization. The answer to this problem is increasing the CPU-to-GPU ratio in AI clusters to keep token costs down.  

Below, I break down why CPUs are positioned to take a larger share of AI cluster bill of materials (BOM) and the explosion in demand we are already seeing. I examine server CPU forecasts that indicate this market will continue to grow rapidly over the coming years. Lastly, I look at the competitive dynamics and key players in this space, and how Nvidia is playing both sides of the CPU-GPU equation, and what front runners Intel and AMD are doing to maintain their lead.  

Ultimately, CPUs have gone from an afterthought to becoming the AI trade’s next great bottleneck – and with AMD, Nvidia, Arm and Intel circling a market that is doubling nearly overnight, the only question left is which company walks away with the lion’s share.

Why Agentic AI Is Driving a Massive Shift to CPUs 

Agentic workloads are structurally different from non-agentic workloads like chatbot queries, which is what has dominated the AI trade up to this point. Chatbots respond to simple requests and provide an output, moving at the pace of the human on the other side. Agents are far more complex, handling hundreds of concurrent tasks autonomously and reasoning through a problem to reach a conclusion, often with limited direction from humans. 

The Intel and the Georgia Tech paper highlights why CPUs are becoming increasingly important as agentic AI proliferates. Researchers noted that while CPU-GPU systems are needed to serve the diverse responsibilities of agents, the “majority of the external tools responsible for agentic capability either run on or are orchestrated by the CPU.” This is not the case in non-agentic workloads, where GPUs are the workhorses that CPUs feed data to. 

Why CPUs Handle Orchestration in AI Workloads 

The key bottleneck this creates on AI infrastructure is orchestration—or the need to call tools, direct API requests, and coordinate tasks between dozens of independent agents. Orchestration is where CPUs thrive. GPUs continue to handle inference reasoning, but CPUs tell GPUs where, when, and how to allocate their resources. 

As AI progresses over the next few years, inference demand is expected to explode—largely driven by agentic AI. Goldman Sachs estimates that by 2030, agentic AI will drive a 24X increase in total token consumption versus today to 120 quadrillion tokens per month. Its forecast shows agentic workloads accounting for over 80% of token consumption in 2030—dramatically higher than their share today.

mid

TrendForce notes that today, the CPU-to-GPU ratio in AI data centers sits between 1:4 and 1:8. For agentic AI applications, TrendForce sees the CPU-to-GPU ratio moving “to between 1:1 and 1:2, significantly boosting market demand for CPUs.”

Other forecasts, like those from Arm, rely on the CPU core count per GW metric. This measures the number of CPU cores per unit of data center power, regardless of the discrete number of CPUs. It is the more accurate way to measure the shift in CPU demand as chip density is increasing, with upcoming generations featuring higher core counts per chip. 

Notably, Arm CEO Rene Haas sees agentic AI driving CPU core demand as much as 4X higher to 120 million cores per GW, compared to around 30 million cores per GW today. Aside from the raw increase in core demand, packing more cores into each chip is a margin expansion opportunity for CPU designers. 

CPU Shortages: Supply Constraints and Pricing Power 

We are already seeing the CPU bottleneck start to play out through worsening CPU server shortages. Reuters reported in February that Intel has a substantial backlog of unfulfilled CPU orders, and that delivery times stretch as long as six months. It also noted delivery times for some AMD products of between eight and ten weeks. KeyBanc issued upgrades on Intel and AMD in January, noting that both firms were nearly sold out of CPU servers for 2026. At the time, KeyBanc noted ASP increases of 10% to 15%. 

Intel and AMD Backlogs and Lead Times 

It appears that the situation has become even more dire since, based on several reports from late May. Reuters now says that TikTok parent company ByteDance is working to accelerate its in-house CPU efforts, as Intel and AMD have raised prices by between 10% and 35% QoQ. ByteDance’s move suggests that it sees a prolonged CPU shortage, leading it to lean into this early-stage initiative. This adds weight to the structural increase in CPU demand implied by AMD’s forecast and shows the pricing power that CPU vendors are exerting.

Electronic equipment distributor Fusion Worldwide says that Intel distributors are only fulfilling around 40% of their yearly backlog allocations. It highlights lead times of 8 to 22 weeks domestically, with Asian customers waiting as long as 8 months. Overall, the firm estimates that Intel is under-shipping real demand by 20% “at best." It notes that AMD’s EPYC CPUs are effectively sold out in 2026, with delivery windows stretching more than 30 weeks. 

The Elec, a South Korean electronics industry trade publication, notes won-denominated price increases as high as 3X for some x86 (Intel and AMD) CPUs. This comes as Intel and AMD prioritize supply for U.S. hyperscalers—leaving little capacity for other customers. The Elec also said that the expected timeline for mass production of Intel’s next-gen Xeon 7 “Diamond Rapids” CPU has been delayed, moving from the second half of 2026 to the middle of 2027.

This data points to a shortage that is intensifying, putting pricing power into the hands of CPU vendors as they seek the highest-margin opportunities. 

AMD Sees Record CPU Server Sales, TAM Estimate Doubles to $120B 

The cause of these shortages is the rapid growth in server CPU demand seen at top players like AMD, and expectations that this market will grow much faster than it traditionally has over the coming years. AMD released its Q1 2026 results in early May, posting its fourth consecutive quarter of record server CPU revenue. Sales rose more than 50% YOY, with both cloud and enterprise end markets up over 50%. AMD expects growth to accelerate significantly in Q2, projecting server CPU revenue growth above 70% YOY, “with robust growth continuing through the second half of 2026 and into 2027.” 

Citing this acceleration in demand and the structural increase on CPU compute requirements that agentic AI is putting on data center infrastructure, AMD has doubled its server CPU TAM estimate. Per CEO Lisa Su, the company anticipates that this will be an over $120 billion by 2030—growing by a 35% CAGR. In November, AMD’s server CPU growth TAM CAGR forecast was just 18%. AMD’s decision to double its market growth forecast and add $60 billion to its TAM in just seven months demonstrates how rapidly current demand signals are translating to long-term confidence among industry leaders. 

Beth Kindig of the I/O Fund discussed in 2024 why AMD would be a winning AI stock and surpass Nvidia's returns over a 3-year time frame. Since then, Nvidia returned 80% and AMD has returned 220%

Thinking about margins going forward, AMD noted at the Bank of America 2026 Global Technology Conference that two-thirds of its server CPU growth in Q1 and expected growth in Q2 are coming from unit increases. Thus, units rather than ASPs are the primary growth driver. Given the worsening supply and demand gap, it’s possible that ASPs could drive an increased share of growth—providing a further lever for margin expansion. 

Server CPU Market Growth Forecasts Surging TAM 

Notably, server CPU TAM forecasts among several Wall Street banks line up with AMD’s forecast. For reference, AMD’s forecast implies a 2025 TAM of just under $27 billion. 

UBS projects that the market will grow from $31 billion in 2025 to $170 billion in 2030, or a 40.6% CAGR. It sees AI CPUs driving the vast majority of this growth, with the TAM increasing from $7 billion to $125 billion, or an 88% CAGR. Their forecast also includes a 56% increase in AI CPU ASPs over this period—implying a significant margin expansion opportunity. 

CPU TAM Revised Higher by Analysts 

Bank of America forecasts a TAM expansion from $43 billion in 2026 to $125 billion in 2030, or a CAGR of 30.6%, recently raising its 2030 estimate from $110 billion. While BofA’s growth rate is lower than AMD’s, this is likely because it accounts for the particularly high growth rates already being seen in 2026. 

Citi breaks down its forecast into three buckets: general purpose CPUs, AI head nodes, and agentic CPUs. It sees the overall market growing from $29.3 billion in 2025 to $132 billion in 2030, or a 35% CAGR. Within this, general purpose CPUs grow by a 20% CAGR to $50.9 billion, and AI head nodes grow by a 21% CAGR to $21.1 billion. Citi estimates that agentic CPU growth will drastically outpace the rest of the market, hitting $59.4 billion in 2030 for a massive 185% CAGR. Overall, the estimates from these three banks circle around the 35% CAGR that AMD outlined. 

Why Growth Rates Are Unprecedented for CPUs 

These very high CAGR forecasts highlight why server CPU shortages are escalating. This market has historically experienced single-digit annual growth rates. Thus, the supply chain was not necessarily prepared for a scenario where customers suddenly look to procure CPUs at a drastically higher pace, and long-term expected growth rates soar in a matter of months. 

AMD’s Goal: 50% Server CPU Market Share 

As AMD looks to increase its share of the CPU market to over 50% by 2030, it is targeting all three of the CPU categories Citi described. This will come through its Venice family of EPYC CPUs, including Verano, its first EPYC CPU purpose-built for AI infrastructure. AMD has begun to ramp production of Venice, while it plans to launch Verano in 2027. 

With this, AMD clearly expects CPUs to be a core growth driver over the coming years. If AMD achieves a 50% market share in the server CPU market, it would imply $60 billion in annual revenue. With server CPUs representing around half of data center revenue, this side of AMD’s business generated approximately $2.9 billion in revenue last quarter, or nearly a $12 billion run rate. Thus, hitting its $60 billion target would require a 5X increase in server CPU sales by 2030—an ambitious goal.

AMD vs Intel: x86 Market Share Dynamics 

There are two ways to think about market share in server CPUs. Mercury Research is one of the key authorities that estimates share in this space, with their estimates often centered around the x86 market. 

AMD is already very much in range of a 50% market share within x86. At its Investor Day, AMD noted that based on metrics from Mercury Research, its share of the server CPU market was around 40%. This lines up with Mercury’s estimate of AMD x86 market share of 41% at the time. Since then, AMD has gained considerable ground on Intel. Mercury Research estimates that AMD controlled 46.2% of x86 server CPU revenue share in Q1 2026 to Intel’s 53.8%. At this pace, AMD is well on its way to achieving a 50% market share in x86.

AMD data center revenue growth and server CPU market share approaching 40 percent shown at Investor Day 2025

At Investor Day 2025, Lisa Su said AMD has a clear path to capturing more than 50% of server revenue market share, up from 40% today, alongside a 50% data-center CAGR and a goal of 40% PC revenue share.

AMD, Intel and Arm Market Share Dynamics 

Arm estimates that in terms of chip value, it held 20% of the cloud compute market share at the end of its fiscal year 2025, which ended in March 2025. Considering Mercury’s estimates on x86, or the 80% of the market that is not Arm-based, these figures imply overall market shares of 43% for Intel, 37% for AMD, and 20% for Arm. However, note the figures from Arm are stale. 

Thus, AMD would need to increase its market share by around 2.6% annually through 2030 to achieve its 50% goal. At least in the x86 market, AMD has cleared a much higher bar over the past several years. In Q2 2023, the company’s server CPU revenue share was just 25.1%, meaning that AMD increased its share of the x86 market by more than 7% annually through Q1 2026.  

While this historical pace is encouraging, it shows AMD’s progress only against Intel—not including Arm’s traditional IP business, nor its move to become a CPU designer through its Arm AGI CPU. Additionally, Nvidia is pushing more aggressively into the CPU market, largely through its standalone Vera racks. However, the Diamond Rapids delay is one factor that could give AMD a leg up in continuing to take share from Intel. 

AMD has its hands full as some of the world’s strongest IP and chip-design companies are targeting the same TAM – including the incumbent Intel, mobile-IP superstar Arm, and the newest entrant, Nvidia. 

Nvidia’s CPU Strategy: Expanding Beyond GPUs 

Within AI-specific infrastructure, CPUs have been traditionally deployed as AI head nodes paired with GPUs in the same rack. A critical development to track is the emergence of standalone CPU racks as this marks a significant shift in architecture.  

Customers will increasingly be able to deploy full CPU racks independently without automatically having to increase GPU counts—one of the key arguments for why CPUs can increase their BOM share in AI clusters. 

Nvidia Vera Standalone CPU Rack Overview 

For example, Nvidia’s Vera rack marks the first time that it will market a standalone CPU rack; a clear signal of the current opportunity in this space. With 256 CPUs in the standalone Vera rack, customers can deploy nearly 7X more CPUs in one rack compared to the 36 CPUs in the Vera Rubin NVL72, which also contains 72 GPUs. Total CPU cores sit at 22,528 for the Vera rack versus 3,168 for the Vera Rubin NVL72. Note that Vera is based on Arm architecture, rather than x86 architecture. 

Arm specifically mentioned the standalone Vera rack as a reason why its 4X CPU core count growth estimate is likely conservative. Arm CEO Rene Haas said, “we probably have undercalled the CPU demand in terms of the transition here. We talked about a 4x increase. We could get our heads around a bigger number than that.”  

Haas went on to say that the number of CPU cores “probably will” exceed the number of GPU cores, even though the number of CPU chips may not exceed the number of GPU chips. 

Nvidia Quickly Eclipses AMD with $20 Billion in CPU Revenue in 2026 

Importantly, Nvidia said on its latest earnings call that the Vera rack opens up a $200 billion CPU TAM for the company—dramatically larger than AMD’s +$120 billion estimate. Within this, Nvidia says it has visibility into generating nearly $20 billion in CPU revenue this year, primarily for standalone Vera racks. Meanwhile, about 50% of AMD’s data center revenue comes from server CPUs, putting this at $2.9B or about a $12B run rate. I expect that to change but allows for a baseline comparison, which is that Nvidia is entering the market aggressively. 

When asked whether CPUs are cannibalistic to GPUs, Nvidia CEO Jensen Huang did not offer a direct yes or no, but framed CPUs as additive to GPUs. He argued that more AI agents require more orchestration—increasing CPU demand—but that more agents also require more inference—increasing GPU demand. This lines up with AMD CEO Lisa Su’s statements that CPUs are largely additive/incremental to their overall TAM.  

Additionally, the standalone Vera rack is just one of four ways that Nvidia targets the CPU market—and is the only one that could substantially change its current CPU-to-GPU ratio. Its other markets include selling head node CPUs paired with Rubin GPUs at a 1:2 ratio in the Vera Rubin NVL72. Nvidia also sells Vera alongside its ConnectX-9 SuperNICs for both storage and confidential computing use cases. 

Still, with the standalone Vera rack, Nvidia is indicating that it expects the CPU-to-GPU ratio to shift—creating a need for the product. Ultimately, while the mix of AI BOM should move toward CPUs, Nvidia is still positioned to capture growth from both chip types. It can benefit from the increasing size of the overall pie, rather than CPU spending going up at the expense of GPU spending. 

Arm Makes Historic Move into Merchant Standalone CPU Racks 

Arm is also forwarding the CPU rack approach through its AGI CPU. Leveraging Arm’s history of delivering high performance with low power requirements for mobile devices, the new AGI CPU is designed to offer a similar balance between high performance and low power consumption. 

The AGI CPU was co-developed with key partner Meta, the chip’s first customer, who revealed they turned to Arm almost two-and-a-half years ago to see if there was a CPU option that fit Meta’s needs: “put in a lot more cores per watt, but we do not want to compromise on the performance piece.” Meta had only been finding options satisfying one of the two criteria: meeting the performance but with too much power, or meeting the power but with too little performance. 

Arm CPUs Deliver Higher Performance Per Watt vs x86 

One of the main advantages that Arm touts is higher performance per watt. Based on internal estimates, the firm says the Arm AGI CPU can provide up to 2x greater performance per watt vs. Intel and AMD’s x86.  

Higher performance per watt is a key value proposition for hyperscalers, allowing more power to be dedicated to compute or networking equipment. 

Bar chart comparing Arm AGI CPU and x86 CPUs (with and without SMT) showing higher sustained performance per thread, threads per rack, and performance per watt for Arm in AI workloads

Bar chart comparing the performance of Arm’s AGI CPU against x86 CPUs (with SMT enabled and disabled) across three metrics: sustained performance per thread, sustained threads per rack, and performance per watt. Arm’s AGI CPU leads in all three categories, with roughly 1.2× higher performance per thread, up to 1.8–2.0× higher thread density per rack, and approximately 2× better performance per watt–highlighting Arm’s efficiency advantage in AI data center workloads, particularly for agentic AI applications where CPU orchestration, scalability, and energy efficiency are critical. Source: Arm Arm 

For more details on Arm, see my analysis from April: Arm Stock Could Win as Agentic AI Shifts the Bottleneck to CPUsArm Stock Could Win as Agentic AI Shifts the Bottleneck to CPUsArm Stock Could Win as Agentic AI Shifts the Bottleneck to CPUs 

In an air-cooled rack, Arm can pack 30 blades (or 60 CPUs) for a total of 8,160 cores in a 36kW power envelope, saying this configuration can deliver up to 2X the performance per rack versus x86 chips based on its internal estimates. Arm says this 30-blade design is “setting records for air cooled” racks that is not feasible with other systems, as power consumption is too high. 

Arm is taking this a step further with a fully-liquid cooled, 200kW open-standard rack in partnership with Super Micro, packing 168 blades, or 336 CPUs, delivering a total of up to 45,696 cores. Arm EVP of Cloud AI Mohamed Awad stated that while it is a “200-kilowatt rack. We actually will consume about half that much power. We ran out of space. That’s why we couldn’t put more cores in there.” 

This is one of the key advantages – it is not just about offering 2X the performance of x86 chips, but providing that performance boost while freeing up power for more compute or for more networking. 

Intel’s Fight to Maintain Server CPU Market Leadership 

Intel is positioning itself to be a stronger competitor in the server CPU market as it relates to agentic AI, announcing its intention to deploy rack-scale CPU systems at Computex. Intel’s new racks look to have quite an edge over Arm when it comes to core density, benefitting from its lead in cores at the individual chip level.  

Intel revealed two blueprints for its upcoming rack-scale CPU systems, with one design targeting maximum density and the other targeting latency-sensitive agentic AI workloads. The two designs can support 128 of Intel’s Granite Rapids Xeon 6 or Clearwater Forest Xeon 6+ chips, which will provide either 16,384 or 36,864 cores based on the chip of choice, alongside up to 384 TB of DDR5 memory per rack. 

Intel also leads on core counts at the individual chip level. Intel’s Xeon 6+ offers 288 cores per chip, slightly beating out AMD’s Venice at 256 cores and Arm’s AGI CPU at 136 cores; AMD has yet to release Verano’s core count. Despite the lead in core count, Intel’s Xeon 6+ only packs 288 threads whereas Venice offers up to 512 threads via multi-threading, allowing each core to handle two sets of instructions, reducing core idle time and increasing efficiency.

Bar chart comparing core and thread counts of major CPUs in 2026 including AMD EPYC Venice, Intel Xeon, Nvidia Vera, Arm AGI CPU, and hyperscaler chips like AWS Graviton and Google Axion

Bar chart comparing core and thread counts of major data center CPUs in 2026. AMD’s EPYC Venice leads with 256 cores and 512 threads, while Intel’s Xeon 6+ and Xeon 7 offer higher core counts at 288 but fewer threads due to no SMT. Nvidia Vera, AmpereOne, and Arm AGI CPUs have lower counts, while hyperscaler chips from AWS, Google, and Microsoft range from 64 to 192 cores. Source: TrendForce 

Intel vs Arm: Power Efficiency Battle 

Where Intel could find its edge in the rack-scale systems is power, with the blueprints fitting inside a 100kW power envelope. Compared to Arm, Intel is offering more than 368 cores per kW, while Arm’s AGI CPU is offering 228 cores per kW. This can also be viewed at the 200kW envelope of Arm’s AGI CPU, at which Intel could theoretically offer 73,728 cores across two 100kW racks, or more than 60% of Arm’s rack.  

This advantage stems from Intel’s 18A node, which in general offers up to 15% better performance per watt and up to 30% better density versus the Intel 3 node. Manufacturing on more advanced nodes is how x86 is fighting back against Arm – AMD’s Venice is the first CPU to ramp on TSMC’s 2nm (N2) process, which is designed to deliver 10%-15% higher performance at the same power level, or a 25%-30% reduction in power at the same  performance level.  

While Clearwater Forest just launched at the start of June, the most important part of Intel’s story is that its next-gen Xeon 7 ‘Diamond Rapids’ is rumored to be delayed. The new chip was originally expected to launch in the later part of 2026, yet is now expected to launch in 2027, giving AMD a bit more of a head start with Venice. 

CPUs Are the Next Major Bottleneck in AI Infrastructure 

It would be a mistake to think the AI trade begins and ends with Nvidia’s GPUs. Although GPUs are still the heart of AI compute, agentic AI is placing additional emphasis on making sure those accelerators do not sit idle while the rest of the system catches up. 

The addressable market is expanding overnight, as this no longer about adding a few more CPUs as head nodes next to GPU clusters. The bigger opportunity is the move to standalone CPU racks, which is a major architectural change that allows more orchestration capacity to be added without adding more GPUs at the same attach rate. Nvidia’s Vera rack is leading the way, with CEO Jensen Huang projecting roughly $20 billion in standalone CPU revenue this fiscal year, yet AMD, Intel, and Arm are not going to concede the market. 

This is the same framework the I/O Fund has used to identify massive AI winners across memory, networking and energy with CPUs now becoming the next bottleneck. Behind our paywall, we publish over 100 articles on the AI trade per year alongside portfolio allocations and real-time trade alerts.  

For example, we identified lesser-known AI winners, including Bloom Energy, up 1600% since our initial entry last year, a networking player that has delivered roughly 7X Nvidia’s returns YTD and an optical networking stock up more than 810% since November. 

Don’t miss out on the AI trade.
Join NowJoin Now

Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund own shares in AMD and NVDA at the time of writing and may own stocks pictured in the charts. 

Leo Miller, AI and Semiconductor Investment Writer at I/O Fund, contributed to this analysis. Leo Miller owns shares of NVDA.

👉🏻 Share with a Fellow Investor
Heare with a Fellow Investor
Help someone else benefit from this insight.

Recommended Reading:

Broadcom Offers Strong AI Growth at Scale; Yet Enters Circular Investing

Posted on by io-fund

By most measures, Broadcom offered a solid report with record revenue of $22.2 billion, up 48% YoY driven by AI semiconductor revenue of $10.8 billion, up 143% YoY. The Q3 outlook topped estimates on total revenue with management guiding for $29.4 billion, yet the Q3 AI guide came in below expectations for $17.2 billion. In addition, management did not offer a raise to previous commentary that they foresee $100 billion in FY27 AI revenue. The softer AI guide is due to a slight pivot in how they plan to supply Anthropic, which will be with XPU chips instead of AI systems, with the latter offering higher revenue yet weaker margins (more on this below).

Overall, Broadcom is well positioned, based on a combination of being the strongest XPU player and a formidable networking giant. In fact, networking comprised 40% of AI revenue this quarter, and even though it's expected to be lower in the near future, that is only because XPUs are expected to eclipse networking from 60/40 to 70/30.

Perhaps somewhat buried by the AI number miss is that Broadcom is entering the circular investing arena by standing up an external financing vehicle with Apollo, Blackstone and other investors to deploy 20GW of compute through 2028, with the first tranche valued at $35 billion. The announcement is a reminder that demand is being heavily funded for companies that are deep in the red and would otherwise see bad credit terms (such as Anthropic and OpenAI).

Below, we look at what caused the softer AI guide and why a softer AI guide is not a concern, whereas circular investing raises questions.

"Only Chips” is What Caused the Softer AI Guide 

The softer AI guide traces back to the fiscal Q3 and fiscal Q4 call awhile back when Broadcom management stated they would be delivering Ironwood racks: “Last quarter, one of these prospects released production orders to Broadcom, and we have accordingly characterized them as a qualified customer for XPUs and, in fact, have secured over $10 billion of orders of AI racks based on our XPUs.”

However, management was quite evasive when analysts had asked for clarification between chips and racks two quarters ago. There was more than one attempt for clarification, yet the one below stands out. Here is what was stated in the FYQ4 call:

“Vivek Arya: BofA Securities, Research Division:
And on the clarification, Hock, Anthropic racks versus chips.

Hock Tan: President, CEO & Executive Director:
I'd rather not answer that, but we're okay. As Kirsten said, we're good on our dollars and margin.”

However, on the earnings call this evening, the CEO had a change of heart and decided to stop dodging the question and rather inform investors they are only supplying the chips.

Ross Seymore 
Deutsche Bank AG, Research Division
And the rack versus chip side of things, is that all clarified now?

Hock Tan 
President, CEO & Executive Director
No racks it's all chip…

Kirsten Spears 
CFO & Chief Accounting Officer
We are on a chip business only.

Hock Tan 
President, CEO & Executive Director
We are only chips.

Kirsten Spears 
CFO & Chief Accounting Officer
Only chips.”

This shift in deal structure has puts and takes for how analysts model Broadcom’s AI revenue this year, as booking the full revenue for the full AI system inflates revenue (because the components would become a passthrough) yet would have diluted Broadcom’s margins. That’s why directly preceding the “only chips” exchange; the analyst was pressing Broadcom on how their margins would be affected by selling systems. Personally, I’m not a fan when there are repeated attempts by analysts to clarify guidance assumptions; those questions are not answered directly, and then a miss occurs after walking back the original framing. 

However, that opinion aside, the miss is inconsequential to the bigger picture. Broadcom emphasized they booked $30 billion in AI orders compared to the $10.8 billion shipped. Management also stated they have visibility into 2028, although that might not be a good thing if the visibility stems from sourcing chips well in advance while having to secure power and other supply constrained components.  

For additional color on the growing backlog and visibility, this was stated on the call: “See a lot of large — this few six customers now, they realize that lead time to get compute, you need lead time. You need to be thoughtful. And that's not just asking for wafers to get the chips or memory to ensure that HBMs are available or DRAM is available. They're also talking about, hey, I got to have the power, the power shell. So all this is planning ahead.

And what we are seeing the bookings that are coming is not for immediate delivery. Some are hope to have, but the reality, they all accept is they need to align quite a few other things in place before they can deliver. But they are placing their orders early and they're placing their orders now, and they are placing orders in fairly huge demand, which basically gives us a lot more visibility than we normally otherwise would have in semiconductors.” 

On the earnings call, an analyst pointed out that if you look at the 10 gigawatts that Broadcom expects to help deploy next year, priced at $15 billion to $20 billion, the $100B medium-term forecast seems low.  

Furthermore, there aresix customers driving the AI orders, whereas in the past Broadcom has been highly concentrated with Google as the primary customer. These customers include Anthropic, OpenAI, Meta, and two more unnamed customers.

Lastly, content per gigawatt was touched on, with the CEO stating Broadcom’s revenue will increase from one generation to the next: “Our revenue — our content per gigawatt will increase. Put it simply, our content from the fact that our compute chip will — XPU will go up in price very dramatically, particularly when you not only put SRAMs into it, as far as it cost, you start putting a lot — you start putting embedding CPU costs into the same XPUs and making those chips basically multi-die with lots of HBM.”

Broadcom to Backstop Anthropic; Deploy 20GW through 2028 with Blackstone, Apollo 

Broadcom is helping to arrange a $35-$36 billion private-credit facility arranged by Apollo and Blackstone to help fund and purchase the deployment of custom TPUs. According to Bloomberg, Broadcom has agreed to backstop part of the debt (about $25 billion) with its top-tier investment-grade credit rating around 5.75% compared to the portion without a backer at 8% to 9% yield. In other words, Broadcom’s strong balance sheet is the reason lenders are extending tens of billions to a customer that is not yet profitable. 

Oddly enough, when asked on the call if Anthropic’s deal was backstopped, the CEO pushed back and said the company is strictly supplying chips. Here is what was stated on the call: “As the deal we did with Anthropic is we use our TPU chips that we developed to provide the compute capacity to Anthropic. We want that it wasn't backstop in that sense. We were the ones providing the chips to Anthropic. We were the ones providing the compute capacity Anthropic.” 

Although Broadcom is not taking equity, and the capital is private credit, according to Bloomberg, the company is lending its credit rating to manufacture demand from a buyer that cannot yet fund the purchase on its own. Therefore, it does closely resemble a backstop. 

Noteworthy Discussions: Incremental Supply and 2027-2028 Commentary 

There was an exchange on the call that points toward 2027-2028 being strong years for Broadcom, with management stating: “Well, good question. Yes, for '27, we indicated about 10 gigawatts shipment in '27. That's still very much intact. That will be shipping — and we are planning to ship 10 gigawatts in '27. And that nothing has changed. Back half loaded, to that extend? Yes, and which really provides an interesting trajectory into '28 with this back half trajectory. So '28, we expect a lot more gigawatts.” 

Also, in another exchange, an analyst asked if Broadcom can secure incremental supply, which would point toward a ceiling to growth. There wasn’t much revealed in the exchange, yet an analyst having this concern in a very supply constrained market (CPUs, HBM, NAND, CoWoS capacity, etc) is noteworthy: 

Timothy Arcuri 
UBS Investment Bank, Research Division 

Right. But if a customer comes to you and wants incremental supply, are you able to go to your suppliers and get it the way that it seems like some of your competitors are? 

Hock Tan 
President, CEO & Executive Director 

Customers have been coming to us incrementally over the last few months. We expect that to continue. And by and large, yes. 

Q2 Revenue Beats by 0.3%; Q3 Guide Implies Acceleration to 84% YoY  

Broadcom reported Q2 revenue of $22.19 billion, beating consensus of $22.12 billion by a marginal 0.3%, growing 47.9% YoY and 14.9% QoQ. While the headline beat was modest — Broadcom’s smallest in five quarters — YoY growth accelerated for the fifth consecutive quarter, picking up another 18 points from 29.5% in Q1 and marking Broadcom’s fastest YoY growth since the immediate post-VMware-close quarters. 

For Q3 FY2026, Broadcom guided to revenue of approximately $29.4 billion, ahead of consensus for $28.47 billion. At the midpoint, the guide implies sharp acceleration to 84.3% YoY and 32.5% QoQ — a sequential dollar increase of more than $7.2 billion, which is itself larger than the company’s entire quarterly revenue base just three years ago. The QoQ dollar step-up of $7.2 billion exceeds the $2.9 billion QoQ step-up between Q1 and Q2, underscoring that Broadcom’s AI ramp is materially compounding. 

Fiscal 2026 consensus revenue estimates have inched slightly higher over the last three months, moving up 5.4% from $97.7 billion to $103.1 billion; the nearly $1 billion beat on Q3’s guide implies FY26 estimates have a bit more upside ahead. Fiscal 2027 consensus have jumped even more sharply to $161.0 billion (+56.2% YoY) from $135.9 billion in March, representing a +$25 billion revision driven by management’s commentary into >$100 billion in chip revenue alongside multiple multi-GW commitments from OpenAI, Anthropic and key customers Meta and Google. 

AI Revenue Up 143% YoY; Q3 Guide Implies >200% YoY but Short of $17.2B Estimate, Possible Q4 Decel 

AI semiconductor revenue was once again the centerpiece of the report. Q2 AI revenue grew 143% YoY and 28.6% QoQ to $10.8 billion driven by increasing demand for custom silicon and networking, beating management’s own guide of $10.7 billion (+140% YoY). YoY growth accelerated another 37 points from 106% in Q1, marking the fourth consecutive quarter of acceleration.  

For Q3, Broadcom guided AI semiconductor revenue to $16.0 billion, implying 200% YoY growth and a material acceleration to 48.1% QoQ. This sequential dollar step-up of $5.2 billion in AI revenue is more than double Q2’s $2.5 billion; however, it fell short of the $17.2 billion estimate.  

For FY26, Broadcom guided for $56 billion in AI revenue, up 180% YoY, while reiterating its >$100 billion guidance for FY27.  

This would plot out $20.8 billion in AI revenue in Q4, decelerating from the 48.1% QoQ guided in Q3 to 30% QoQ, and implying sequential dollar growth to moderate from $5.2 billion to $4.8 billion. JP Morgan’s Harlan Sur questioned about this deceleration dynamic, noting that 2X growth in 2H over 1H would put revenue closer to $60 billion, rather than the $56 billion guided. While the exchange with CEO Hock Tan suggests Broadcom may be taking quite a conservative stance in guiding through 2H while remaining positive on FY27’s prospects, the sequential deceleration on both a percent and dollar basis presents a risk to watch:  

“Hock, on this fiscal year, AI sort of 2x growth second half over first half, that would put AI revenues over $60 billion with sequential growth in fiscal Q4, but you gave us this $56 billion number, which is only like 1.5x half-over-half growth with 4Q AI actually being down sequentially. So if you could just help us kind of square the numbers there. 

Hock Tan
President, CEO & Executive Director 

To begin with, let's start with '26. Doing a math basically 2x to 2x, the first half, we ship about in total AI revenue, something in the range of $19 billion, you're going to be precise. So — and if you do what I indicate and 2x that in the second half, you get pretty much in the range of what we're talking about, which is around $56 billion, Harlan. 

So that number is still very, very — does tie up very well. Now your bigger question on the second half, which you're going to need a very detailed analysis of is, yes, we keep the momentum going as we expect to see in 2027. What we will see in 2027 is continued growth of the level we're talking about. And if you drive on that basis of what we're seeing here, almost 2x — in the range of 2x what 2026 will be. 

I think you will easily see that 2027 will exceed very easily $100 billion in 2027, which is pretty much what we indicated last quarter, and we are continuing to say that it will be over $100 billion in 2027. So in that sense, if anything else, it might be based on what we're doing, very much on track, if not stronger.” 

Semiconductors Up 79% YoY; Software In-Line at 9% YoY  

Semiconductor Solutions revenue was $15.01 billion in Q2, up 78.5% YoY and 20.1% QoQ, beating the company’s own guide of $14.8 billion (76% YoY). YoY growth accelerated 26 points from 52% in Q1, with AI now contributing approximately 72% of Semiconductor segment revenue, up from 67% in Q1.  

Infrastructure Software revenue was $7.18 billion, marginally below the company’s ~$7.2 billion guide and up 8.8% YoY and 5.4% QoQ. YoY growth rate decelerated from the elevated VMware-integration period a year ago. For Q3, Broadcom guided for Semiconductor revenue of $20.5 billion, up 124% YoY and 36.6% QoQ, and Infrastructure Software revenue of $8.9 billion, accelerating sharply to 31% YoY and 24% QoQ.  

Margins: Operating Leverage Drives GAAP Expansion  

Q2 margins highlighted the operating leverage thesis that has underpinned the Broadcom story since the VMware close, with GAAP profitability expanding as revenue scaled against a largely fixed cost base. 

Q2 GAAP gross margin was 69.5%, expanding 150bps YoY and 140bps QoQ. Adjusted gross margin was 77.1%, in line with management’s 77% guide and expanding 10bps QoQ from 77.0% in Q1 — notable because it dispelled the prior concern that the rising XPU mix would pressure gross margins. Adjusted gross margin remains down 230bps YoY (from 79.4% in Q2 FY25) due to a higher custom-silicon mix, but the QoQ stability suggests the mix headwind has largely played through.  

Q2 GAAP operating margin was 48.6%, expanding 980bps YoY and 430bps QoQ — a solid demonstration of operating leverage as semiconductor revenue scaled approximately $4.5 billion above Q2 FY25 levels while opex grew only ~6%. Adjusted operating margin was 67.3%, beating the 67% guide and expanding 200bps YoY and 90bps QoQ. For Q3, Broadcom guided adjusted operating margin to ~67% (flat sequentially).

Q2 GAAP net margin was 42.0%, expanding 890bps YoY and 390bps QoQ. Adjusted net margin was 54.4%, expanding 250bps YoY and 170bps QoQ.  

EPS and Adjusted EBITDA 

Adjusted EPS was $2.44 in Q2, beating estimates of $2.40 by 1.7%, marking Broadcom’s second consecutive quarter of sub-2% EPS beats. Adjusted EPS grew 54.4% YoY, accelerating from 28.1% in Q1. GAAP EPS was $1.91, growing 85.4% YoY. 

While Broadcom did not guide directly for Q3 earnings, the $1 billion beat on revenue and margin maintenance suggests potential upside to current estimates for $3.18 in adjusted EPS, up 88.1% YoY. This is also likely to put upwards pressure on FY26 estimates, which sit at $11.33, up 66.1% YoY. Similar to revenue, FY27 EPS estimates have seen a strong upwards revision over the last three months, up 27% from $14.56 in March to $18.50, driven by the expected surge in AI revenue next year.   

Adjusted EBITDA was $15.24 billion at a 68.7% margin, beating the 68% guide and growing 52.4% YoY and 16.1% QoQ. Adjusted EBITDA was guided to be ~68% of revenue in Q3, a marginal step-down versus Q2. 

Cash Flows and Balance Sheet  

Cash generation in Q2 was exceptional, with both OCF and FCF margins reaching post-VMware highs and dollar generation setting new records. 

Operating cash flow was $10.49 billion in Q2 for a 47.3% margin, up 60.1% YoY in dollar terms and 27.0% QoQ. OCF margin expanded 360bps YoY and 450bps QoQ as higher-margin AI revenue mix flowed through to cash conversion.  

Free cash flow was $10.26 billion for a 46.2% margin, up 60.1% YoY and 28.1% QoQ, with capex of just $231 million (1.0% of revenue, down slightly from $250 million in Q1). FCF margin expanded 350bps YoY and 470bps QoQ. 

Cash and equivalents climbed to $19.63 billion at quarter-end, up from $14.17 billion in Q1. Debt declined modestly to $64.91 billion. The combination of moderating buybacks and accelerating FCF means Broadcom’s net debt position has improved by approximately $6.6 billion over the past two quarters, providing meaningful flexibility for either an acceleration of buybacks, M&A, or — given the AI ramp — potential incremental capacity investments. 

Inventory rose sharply to $4.33 billion at quarter-end, up 46% QoQ from $2.96 billion in Q1, a meaningful supply-side signal that reinforces management’s confidence in the Q3 and Q4 AI ramp. Days sales outstanding extended to 44 days (from 40 days in Q1), reflecting the rising mix of larger hyperscaler customers with longer payment terms, though still well within historical norms. Both metrics — accelerating inventory build and modestly extending receivables — are consistent with a company gearing up for a sharp sequential ramp rather than one facing demand softness, mirroring similar supply-side signals seen at Nvidia and other AI infrastructure peers.

Conclusion: 

Broadcom’s AI revenue growth on a YoY and QoQ basis is stunning on all accounts, especially given its growth at scale. Most importantly, Broadcom is diversifying its customer base to six customers with bookings running 3X shipments with visibility into 2028.

Management did not provide an updated guide for FY27, which may be partly due to the chip-only content that led to next quarter’s miss, but it could also be they are not sure when their customers will secure the other supply constrained components.

The more immediate reason the report is likely selling off after hours due to bringing up an important modeling question, which is how much revenue should be assigned to each gigawatt of XPU compute? Rack-level assumptions imply a much larger revenue opportunity than chip-only content, thus we are seeing an adjustment after hours. As you can tell from my write-up, I think this was an avoidable communication issue on management’s part, especially given the repeated attempts by analysts to clarify the rack-versus-chip economics in previous earnings calls.

However, a minor miss on surging AI revenue will soon be water under the bridge. The larger concern is around circular AI investments, which are likely here to stay. Broadcom is now tethered to AI customers that need enormous compute capacity but are not yet profitable at the scale required to fund it internally. The creation of financing vehicles with Apollo, Blackstone and other investors is one strategic solution, yet it’s not exactly ideal to lend your own credit rating to manufacture customer demand, especially given Anthropic is likely years away from profitability.

That’s a wrap! I/O Fund just had one of our best quarters ever, helped by strong positioning ahead of the Nasdaq’s historic April rally. Let’s see if we can do it again. Keep an eye out for upcoming analysis on new stocks we may add to the portfolio as we rotate out of weaker names, plus my Q3 Top 15 AI Stocks report, due next month.

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

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

Recommended Reading:

Core Scientific: Multi-GW Pipeline, New Hyperscaler Interest but Still Tied to CoreWeave 

Posted on by io-fund

Core Scientific represents one of the stronger Miners as they are already earnings revenue on 243 megawatts with another 200 megawatts expected to be earning revenue in the coming months. This helps transition Core Scientific away from being pure speculation as the company is beginning to execute. With that said, Core Scientific must continue to execute to reach its full potential, and this is particularly important because of the company’s debt structure.  

As a reminder, Core Scientific is pivoting from being a Bitcoin miner to offering multisite infrastructure buildouts as a colocation data center provider. Therefore, at the moment, Core Scientific offers a challenging fundamental profile, with operating margins and cash flows still in the red. However, management stated the 243MW currently being billed will result in $350 million annualized colocation GAAP revenue, and raised their expected cash gross profit margin up 500 bps to 82.5% at the midpoint. The takeaway is that Core Scientific can offer enough profitability and visibility to support more growth.  

Notably, Core Scientific has expanded its gross power pipeline to 4.5 GW with 3GW of that pipeline leasable by customers, suggesting annual revenue opportunities of more than $5 billion under current deal economics at full scale. Although this is quite promising, it circles back to execution and debt structure, which is discussed more below.  

Lastly, Core Scientific still remains closely tied to CoreWeave (whose planned $9 billion acquisition of the miner fell through). The first attempt to diversify beyond CoreWeave with a hyperscaler fell through as a deal under exclusivity was allowed to expire. It’s a yellow flag that a hyperscaler deal fell through (raises important questions), although management seems optimistic as they are in discussions with three hyperscaler customers.  

Quick Recap on Miners’ Value Proposition 

Time-to-power is becoming a central bottleneck for AI data centers, as grid constraints rise, connection queues lengthen, putting the emphasis on quick, suitable on-site and behind-the-meter solutions such as Bloom’s fuel cells and GE Vernova’s gas turbines.  

Bitcoin miners offer a third solution for the time-to-power thesis, offering up to several GW of capacity in quick fashion, bypassing interconnection queues for greenfield builds, and offering cheaper electricity costs from long-term power contracts. For example, miners such as IREN and TeraWulf have touted electricity costs around $0.046-$0.047/kWh in the past, compared to PJM’s ~$0.08/kWh in 2025 and commercial sector rates averaging $0.08-$0.22/kWh. For a 400MW data center, electricity expenses at miner rates would be roughly $162 million, versus $280 million to $771 million under commercial sector average rates, or annual savings of ~42% to 79%. 

For neoclouds such as CoreWeave, these lease-based deals and cheaper electricity costs offer a compelling structure to quickly bring significant capacity online to scale revenue, without bearing the construction capex (around $12-14 million per MW) or dealing with power procurement.  

Overall, the value proposition is that miners are cheaper and faster than new, greenfield data center sites that are not energized, yet the downside is that they are capital constrained and may be unable to build-out capacity beyond what is currently in their pipelines. 

According to Core Scientific, they can offer five sites with the first of them ready-for-service (RFS) with a timeline of 18 months or less. Part of Core Scientific’s current strategy is to stop retrofitting the existing infrastructure and to pursue greenfield instead. Thus, the 18 months reflects a (very quick) newer build rather than the unpredictable nature of upgrading older sites. Here is what was stated on the call: “I think the thing that was much more difficult than we certainly gave a credit for was the — was actually executing on brownfield conversions, which is why everything you see that we're doing forward is actually a greenfield site with a very highly standard basis design that allows us to get kind of leverage over our supply chain and be super predictable in terms of our delivery dates.” 

Although this pushed back the timeline on when a Miner like Core Scientific becomes a more viable stock, it also could increase the predictability of deals getting signed and executed as we move further into the 18-month lead time (i.e., 2027). 

Gross Power Pipeline Expanded to 4.5GW, Lots of Execution to Get There 

When we first covered Core Scientific more than a year ago for Advanced members here, Core Scientific: Hypergrowth with 21X AI Segment Growth Potential, the miner’s contract power pipeline spanned 1.3GW, yet today, it is more than 3X higher at 4.5GW. Notably, this excludes the 590MW already contracted by CoreWeave, meaning Core Scientific’s gross power pipeline technically exceeds 5GW, more than double TeraWulf’s 2.3GW and in a similar boat as IREN and Applied Digital around 4.5GW each in North America.  

This 4.5GW power pipeline translates to 3GW of leasable capacity, with 1.5GW of that 3GW figure currently grid-connected. Two sites – Muskogee, Oklahoma and Pecos, Texas – account for the majority of Core Scientific’s pipeline, both with potential to scale to 1.5GW gross power each, or 1GW leasable. 

However, Core Scientific’s current focus remains on delivering its capacity for CoreWeave, ramping from the 243MW billable as of Q1 to 450MW by the end of Q2, with the target of reaching full capacity in early 2027:  

"Across our remaining contracted sites, we will continue delivering billable megawatts over the coming months while scaling execution on the CoreWeave contract, positioning us to deliver more than 450 billable by the end of the summer, while remaining on track to deliver the full 590 megawatts by the early 2027.” 

Executing on CoreWeave’s ramp will be the primary focal point for 2026 and early 2027, as management hinted that delivery of non-CoreWeave capacity within the 4.5GW pipeline will not occur until early 2027: “strategically positioning the business to sign attractive new customer contracts with capacity outside of CoreWeave available for delivery starting in early 2027.”  

What is Core Scientific’s Pipeline Worth? 

We can roughly translate what this power pipeline could suggest for Core Scientific’s revenue potential at 3GW of total leasable power. At roughly $1.4 million per MW per year, or the current run rate of its CoreWeave deal, its entire pipeline (should it materialize and at similar terms) could be worth $4.2 billion in annual revenue potential. Should it advance towards $1.7 million per MW, or in line with Applied Digital’s recent deal, that annual revenue potential moves to $5.1 billion.  

Overall, Core Scientific’s power strategy is all about ‘proactive positioning’, as management put it – securing land, labor and equipment to keep delivery timelines and ready-for-service dates on track for within 18 months, securing gas and behind-the-meter power to enable expansion, and showing prospective customers on-the-ground progress to entice deal-making discussions.  

As you can see, to get to the full pipeline, it comes down to strong execution. 

Shifting to Greenfield Development 

There was one interesting topic of discussion late in Q1’s call about lessons learned from developing multiple sites for CoreWeave, and how that translates into a competitive advantage. On this, management revealed that the advantage lies within their development approach – they are not retrofitting sites, but rather developing them from scratch: 

“I think the thing that was much more difficult than we certainly gave a credit for was the — was actually executing on brownfield conversions, which is why everything you see that we're doing forward is actually a greenfield site with a very highly standard basis design that allows us to get kind of leverage over our supply chain and be super predictable in terms of our delivery dates. 

Brownfield sites are highly unpredictable. They require a lot of customization. It's a lot of effort to try to retrofit an existing building. While sometimes that could be faster, it comes with a lot more complexity.” 

This is a key distinction we raised in our Hyperscaler Power analysis, Why Power is Critical for Data Centers and their Hyperscaler Customers:  

Brownfield sites (retrofitting) are the path other Bitcoin miners are taking as it is cheaper and faster than greenfield, allowing them to convert old Bitcoin mining halls into AI data center capacity at a relatively quick pace. However, greenfield builds – where the developer owns the land, power, and building – offer a higher degree of customization, though at a much higher cost and often with the longest timelines to completion due to permitting, site selection, and grid connection. 

It is that last point where Core Scientific gets its greenfield advantage – it does not have to deal with lengthy site selection or grid connection requestions, with a repeatable development playbook that can pencil in RFS dates within the next 12 to 18 months. It also represents (somewhat of) a faster path-to-market compared to typical greenfield hyperscale builds, which can take >16 months to reach first operations since starting construction.  This is evident in the Muskogee campus with the Polaris deal adding 440MW of contracted energy to its facility, bypassing the grid interconnection queue and opening the door for quick expansion (pending demand and capex). 

However, the main challenge is financing these greenfield builds faster than its miner peers, as these new sites are crucial in diversifying customer exposure outside of CoreWeave. For example, TeraWulf is targeting 2H 2027 delivery of up to 480MW at a single campus in Kentucky, while IREN is aiming to add more than 700MW in 2027. 

$3.3 Billion Financing Helping Accelerate Site Development 

We can roughly infer what Core Scientific’s pipeline would cost to build out, based on management’s estimates for development costs of million per MW. Core Scientific is estimating build costs to be roughly $11 million per leasable MW, meaning that its leasable pipeline of 3GW would cost in the ballpark of $33 billion.  

On this topic, Core Scientific recently closed a $3.3 billion secured senior note raise due in 2031, netting $2.9 billion in gross proceeds, to be allocated across its five sites currently under development.  

Of the proceeds, $2-2.2 billion is expected to help support development of roughly 1 GW in leasable capacity, or two-thirds of its current grid-connected 1.5GW of leasable capacity. It’s important to note that the $2-2.2 billion figure merely represents a 20% cash outlay necessary to land project financing, with this completing the remaining 80%, or ~$8.8 billion.  

All told, the 3GW pipeline would likely require $6-6.6 billion in cash and potentially more than $25 billion in related project financing to be fully developed under a similar structure. This is more than 6X Core Scientific’s current cash balance, meaning debt or project financing will be leaned on quite heavily. The problem with that is high interest – the $3.3 notes carry a 7.75% rate (essentially a junk bond), meaning Core Scientific will add $256 million in interest payments annually, placing further strain on its balance sheet. While it does help avoid diluting shareholders, it opens the door to execution risk, both by Core Scientific and its key customer CoreWeave.  

Importantly, management emphasized that they are not waiting for a deal to be signed before advancing site development, with current capital letting Core Scientific build and target 12 to 14 month ready-for-service timelines. Considering that debt is funding the non-CoreWeave-associated buildouts and that Core Scientific is not waiting for deals to be signed to continue building, analysts questioned what guardrails Core Scientific had on capex:  

“Is there any guardrail on how much CapEx you would start putting forward before getting a lease? 

Adam Sullivan, CEOAdam Sullivan, CEO 

“I mean the way we're thinking about it right now is we want to take the first data hall to full RFS. And as part of that, that means we're securing the labor, securing the trades, we're securing long lead equipment. And we're putting ourselves in a position where if a customer signs really within any time period leading up to the RFS, the first data hall, we can just continue to extend all of that labor that we have secured on site. So that's kind of our guardrail right now in terms of where we sit. But we feel very confident in the strategy and the ability to show the progress that we're making across each of these sites to customers is really what's forcing the engagement here because everyone is incredibly interested in capacity that's getting delivered in '27 right now.” 

The main readthrough here is that Core Scientific does not really have a financial guardrail in place, and instead is banking on a deal being signed, based on high interest in the market and the fact that they are making construction progress. This means that Core Scientific will likely be on the hook for a majority of the construction costs of the new facilities, a shift in strategy from its CoreWeave deal where the neocloud is fronting some of the capex bill for its capacity. Management also noted that they are looking to deploy behind-the-meter power solutions over the same 12 to 14 month timeframe, which, if not included in the above capex portrait, could add $500 to $600 million per GW to project costs. 

The longer that a deal takes to come to fruition, the more capex, and more debt, that Core Scientific will have to incur.  

Touching on Behind-the-Meter Power 

It’s necessary to briefly touch upon Core Scientific’s willingness to turn to behind-the-meter power solutions as a key method of increasing its power pipeline. Based on commentary for the planned Pecos and Muskogee expansions, Core Scientific expects behind-the-meter solutions to account for as much as ~1.86GW across both sites (to reach 1.5GW each), if additional grid power under load study does not pan out.  

One of the main benefits of the miners that we had originally highlighted was that miners already have power secured, yet the main risk here is that Core Scientific’s goal of having two GW-scale sites do not have power secured, and instead may rely on more expensive sources of power.  

Core Scientific noted that behind-the-meter offers a faster time to power than waiting for grid interconnection, which is reasonable considering the scale of these two sites;  however, as noted above, behind-the-meter solutions are not necessarily cheap, and could run as much as $500 million per GW. This could add more than $1 billion to development costs across the two sites, an expensive endeavor assuming power would have to be procured prior to a deal. The other risk is that there is no guarantee that Core Scientific would be able to secure the power necessary for both sites to expand, either via the grid or behind-the-meter. 

Core Scientific’s Crux – No Second Deal (Yet) 

The main drawback, to say, is that Core Scientific has yet to diversify beyond CoreWeave, partially because CoreWeave had attempted to acquire Core Scientific, likely limiting its deal-making ability. Additonally, Core Scientific had a hyperscaler in exclusive discussion across its Pecos and Muskogee campuses, yet the customer's exclusivity expired without a deal being signed.  

Despite that exclusivity agreement expiring, management explained that “three hyperscalers immediately engaged on those same sites, and we are now in active discussions.” CEO Adam Sullivan added that it was “hard to determine the exact reasons why” the original hyperscaler did not follow through with a deal, though Core Scientific believed it was “the best time for us to bring these back to market because hyperscalers were knocking at the door and asking questions about the sites. And we knew we could have an opportunity to bring another hyperscaler into the fray.”  

Given that Core Scientific essentially has taken a step back in the deal-making process, analysts questioned about the three hyperscaler engagements, and if this would be starting the process again from scratch or if there were previous discussions that could accelerate a potential deal. Management confirmed the latter, explaining that it was simply “bringing back both Pecos and Muskogee back to the table. And that's really why we are able to immediately reengage with those customers.” Core Scientifc added that it was also in conversations with AI labs, neoclouds and chipmakers.  

Though there was no indication around when a potential deal could be signed, management believes they are closer to a deal than in Q4 and uniquely positioned to close potential deals. This stems from their focus on having ready-for-service dates within the next 18 months with active construction progress, with five sites expected to have first data center halls ready in 2027:  

How does the negotiation get altered with some of these potential customers when you've secured the supply chain and you're kind of moving forward? Does that accelerate discussions? Does that keep them more engaged?  

Adam Sullivan, CEOAdam Sullivan, CEO 

Yes. I mean it definitely keeps them more engaged. I mean they rarely see sites that come across their desk where there's an RFS time line really within 18 months, but even more so less than that. And so for us, being able to show photos and videos of sites with active construction going on and the list of equipment that are on order that dramatically changes the dynamic of the discussions because this isn't just a photo of a piece of land. This is an active construction site actively progressing towards building a data center.  

Although management did not specifically discuss why the hyperscaler fell through, our readthrough is the change in tone from brownfield to greenfield may be where the delay came in. If you go back about 6 months ago, Miners were attempting to retrofit. According to this earnings call, that is a dead-end of sorts and greenfield is the way forward, which would naturally cause a delay in a deal.  

Financials 

Revenue Inflects in Q1 as Colocation Revenue Ramping  

Core Scientific’s revenue inflected in Q1 as Colocation revenue showed a strong ramp with billable capacity for CoreWeave reaching 243MW, up from 185MW in Q4. Q1 revenue was $115.2 million, up 44.9% YoY and 44.5% QoQ, accelerating from (16%) YoY and (1.7%) QoQ in Q4. 

For a segment breakdown: 

Colocation revenue was $77.5 million, up 804.5% YoY and 147.4% QoQ, driven by incremental capacity delivered to CoreWeave during the quarter. This marked a sharp acceleration from 267.8% YoY and 109.6% QoQ in Q4.  

Within Colocation, lease revenue was $59.2 million, up 892.4% YoY and 136.7% QoQ. Power fees passed through to CoreWeave were $21 million, while maintenance cost ($2.7 million).   

Management added that the 243MW of billable capacity represents roughly $350 million in annualized revenue, with 200MW of incremental billable capacity expected to come online by the end of the summer (Q2). This additional 200MW would represent roughly $288 million in annualized revenue, or $72 million quarterly; however, assuming half lands in Q2 due to the intra-quarter ramp timing, Colocation revenue would roughly estimate to $113 million next quarter, up 45.8% QoQ and 966% YoY. 

Digital Asset (Bitcoin) Mining revenue totaled $37.7 million across self-mining ($30.1 million) and hosted mining ($7.6 million), declining (46.9%) YoY and (22.1%) QoQ. Core Scientific is expecting a “meaningful step down” in miners in 2H as it transitions to Colocation. 

Currently, Q2 revenue is projected to be $134.5 million, accelerating to 71% YoY though QoQ growth would moderate to 16.8%. Q3 is projected to see a further acceleration to 112.8% YoY and 28.3% QoQ to $172.5 million in revenue driven by the capacity ramp. 

For the full year, revenue is currently projected to be $622 million, up 95% YoY, with FY27 estimated to reach $1.04 billion, up 66.5% YoY. Considering Core Scientific is aiming to deliver five sites in 2027 and satisfy the full 590MW for CoreWeave in the early part of the year (representing $850 million in annualized revenue), there is potential for upside to the current revenue estimate if it can contract out some of these sites to new customers. 

Operating Margin Impacted by Impairment Charge, Colocation Margin Dynamics 

Q1 saw gross margin improve double-digits YoY as Colocation takes a larger mix and as Bitcoin operations are wound down. However, impairment charges related to the Bitcoin operations had an outsized impact on operating margin.  

GAAP gross margin was 26.1%, up 15.8 points YoY and roughly flat QoQ.  

GAAP operating margin was (269.4%) due to recording a $266.5 million impairment charge in the quarter, widening from (59.1%) a year ago and (147.3%) in Q4. Excluding the impairment charge, operating margin would’ve been (38.1%).  

GAAP net margin was (301.3%), which was not comparable to 724.6% a year ago or 270.8% in Q4 as both quarters benefitted significantly from changes in fair value of warrants.

It’s also important to touch a bit upon Colocation margins, as power fees passed through to customers are recorded as revenue, yet because they are fully passed through, carry a 0% margin.  

Thus, Colocation reported an 57% gross margin overall in the quarter, up 52 points YoY and 11 points QoQ. However, when stripping out passed-through power costs, Colocation gross margin (lease revenue minus maintenance and other expenses) was 78%, up 21 points QoQ. Management added that they have “increased our target cash gross profit range for the CoreWeave contract to 80% to 85%, up from our original target of 75% to 80%” as they now have “much greater visibility into the associated cost structure given we are now billing for a meaningful portion of the contracted megawatts.” 

EPS 

Driven by the impairment charge, Core Scientific reported a large GAAP loss this quarter, though GAAP profitability is expected as early as Q3.  

GAAP EPS was ($1.06) in Q1, down from $1.25 a year ago and $0.42 in Q4. Adjusted EPS was ($0.11), improving from ($0.13) a year ago and ($0.18) in Q4.  

Looking ahead to Q2, GAAP EPS is projected to be ($0.02), likely accounting for no impairment charges, while adjusted EPS is projected to be ($0.06). Q3 is expected to see GAAP EPS turn thinly positive at $0.01, while adjusted EPS would remain negative at ($0.06); however, this profitability likely assumes no impairment charges, which are a real possibility given the expectation of a significant wind down in Bitcoin operations in 2H. 

Adjusted EBITDA in Q1 was $4.4 million for a 3.8% margin, up from (7.6%) a year ago and (53.5%) in Q4.  

Balance Sheet and Cash Flows 

One of Core Scientific’s advantages in the miner landscape is that CoreWeave is fronting a majority of the capex at up to $750 million ($1.5M/MW), whereas other miners are turning to debt and paying the entire construction/retrofitting costs themselves. However, Core Scientific’s current strategy of progressing greenfield builds pre-contract may require a higher degree of self-funding moving forward.  

Q1 operating cash flow was $249.9 million for a 216.8% margin, driven by the impairment charge and sale of ~$208 million in Bitcoin. This was up from a (56.6%) margin a year ago and 197.3% in Q4. 

Q1 free cash flow was ($136.7 million) for a (118.6%) margin, improving from (162.2%) a year ago and (152.7%) in Q4. Capex was elevated at $389.3 million, or ~3.4X of revenue. 

Deferred revenue was $654.2 million, up from $555.9 million in Q4. 

Cash was $1.0 billion, while debt was $2.1 billion; this does not include the $3.3 billion senior secured notes raised in Q1. Debt and project financing will need to be tracked closely given the current health of Core Scientific’s balance sheet with a ($1.3 billion) deficit. 

Conclusion 

Fundamentally, Core Scientific’s revenue is beginning to inflect as it delivers more capacity for CoreWeave, aiming to deliver an additional 200MW by the end of Q2 to take its total billable capacity to nearly 450MW, more than 75% of the way to its full 590MW obligation. Margins remain negative, though GAAP EPS is expected to potentially shift positive as early as Q3 as Colocation revenue ramps into year-end. 

Core Scientific is making solid progress in expanding its power pipeline, with up to 4.5GW of gross power potential with behind-the-meter and load expansions under study, offering up to 3GW of leasable capacity if fully developed. Pecos and Muskogee are expected to be the company’s primary campuses, both with potential to expand to 1GW of leasable capacity each, and likely the main cornerstones in diversifying exposure outside of CoreWeave to hyperscaler customers. 

While the company is working to progress rapidly with greenfield development of its non-CoreWeave-tied sites, aiming to have five sites ready for service in 2027, capex and debt needs must be watched closely as Core Scientific is funding this development itself. The weak fundamental profile of Core Scientific’s financials makes this stock an Advanced-only momentum play, and one we would only participate in for momentum purposes if we felt it was breaking out. Join Knox this week in his weekly webinar for more information.

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

Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund do not own shares in CORZ at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

Monolithic Power: Enterprise Data Growth Boosted by 35 Points, 800G Optical Growth Appearing

Posted on by io-fund

Monolithic Power Systems (MPS) is entering 2026 with solid AI-driven momentum in its Enterprise Data segment, recording growth of 97.7% YoY and 12.6% QoQ, a third consecutive quarter of double-digit sequential growth albeit decelerating each time.  

While management last quarter had laid out a ‘conservative’ floor for 50% YoY growth for Enterprise Data, this quarter saw management increase the floor to 85% YoY for the segment. This would represent a nearly $250 million raise in just one quarter, underpinned by strong ordering patterns continuing and increasing visibility in Q1. Additionally, the rollout of next-gen systems such as Nvidia’s Rubin and AMD’s MI450X serve as additional 2H catalyst.  

Monolithic also had an unexpected catalyst this quarter, seeing strong optical-driven growth emerge in its Communications segment, driving growth of 33.1% QoQ, the segment’s strongest sequential growth since Q3 2024. This growth was driven primarily by 800G optical modules, which are expected to see shipments rise as much as 2.6X this year. Additional levers for optics-driven growth include increasing optical attach rates with Rubin as well as the ramp of 1.6T.  

On the capacity front, Monolithic is increasing its near-term capacity goal to $6 billion, having reached its $4 billion target last quarter. This capacity expansion is imperative in preventing the company from becoming supply constrained given that FY26’s revenue estimates are already within 10% of that $4 billion mark.  

Brief Recap on Products, Vertical Power Delivery 

For a brief recap, Monolithic supplies a range of power management ICs, a broad suite of DC-DC converters, power modules, and multi-phase regulator modules. 

Power management ICs help regulate, distribute and optimize power within chips, and also help regulate heat dissipation to prevent overheating, an especially critical function as AI GPUs continue to push the thermal boundary higher.    

DC-DC converters help enable precise voltage regulation from 400V DC power entering the rack to lower voltage rails required by AI accelerators. DC-DC converters can also handle higher power densities, reducing power consumption and optimizing performance of increasingly-power hungry GPUs. 

Power modules combine DC-DC converters with built-in, integrated power MOSFETs, inductors, and other necessary components into a single module to lower BOM, reduce amount of external components required, and simplify AI chip design. 

Multi-phase voltage regulator modules (VRMs) are increasing in content as GPUs push into higher power requirements. GPUs operate at very low voltage at roughly 1V — but draw extremely high current, which means power must be converted from 12V at the board level down to the GPU core. Rather than relying on one oversized regulator, engineers distribute the load across multiple phases to improve efficiency and thermal management. 

However, as AI GPUs heat up to 2,000W and beyond (where Rubin and AMD’s MI450 generation are headed), the traditional placement of VRMs and lateral power delivery carried a major drawback that must be addressed.  

Lateral routing increases power loss and drives heat generation higher, causing overheating (when next-gen GPUs are already getting hotter) and reduced performance. This is essentially forcing a shift to vertical power delivery (VPD), which places voltage regulators directly under the PCB and shortening delivery lengths. VPD helps enable higher power density, lets modules more efficiently power GPU, CPU and memory rails, while also freeing up space on the PCB for additional HBM stacks or other components.   

With these key advantages and push to more powerful GPUs with each generation, MPS expects VPD to essentially become a non-negotiable in 2026: “This is just the direction of the market. It's the only energy-efficient solution you can put in place if you're going to operate in these high-voltage — high-current.”  

Enterprise Data FY26 Growth Raised from 50% to 85% 

While Q1 showed more evidence of Monolithic’s growing AI-driven revenue opportunities outside of its Enterprise Data segment, management’s updated growth guidance for the segment in FY26 arguably stole the show. This growth in Enterprise is likely being driven by VPD and higher content per socket, as VPD is becoming increasingly necessary with more powerful chips such as Rubin and the MI450. 

After guiding for a floor of 50% YoY growth last quarter, management raised this forecast to 85% YoY growth, a substantial 35 point raise in just one quarter (and with three more to go in the year):  

“We tend to be fairly conservative in how we look at these things, waiting for the backlog to be in place. So late last year, we talked about 30% to 40% growth year-over-year. In the last call, we kind of rose that to a 50% floor. And the strong ordering patterns that we saw start last year has kind of continued through Q1. So at this point in time, I think we're comfortable raising that floor up to around 85% year-over-year growth.” 

To put this in perspective, Enterprise Data revenue in FY25 was $701.9 million, so the updated forecast essentially represents a nearly $250 million raise just one quarter into the year – FY26 revenue would project to almost $1.3 billion at 85% versus $1.05 billion at 50%. 

Two pieces of additional commentary from management suggested that this growth guide is likely to move higher through the year – that the guide is underpinned by strong ordering patterns emerging through the quarter, and that Monolithic is not facing any supply chain constraints.  

As we had pointed out in our write-up covering Q4’s results, Monolithic Power: Strong AI Tailwinds to Drive 50% Enterprise Data Segment Growth in FY26, the 50% growth guide was supported by increased visibility into ordering patterns, yet visibility into 2H was limited. We explained that “the key takeaway is that 2H will likely be the determining factor for where Enterprise Data growth lands.”  

This quarter, the takeaway remains very similar. VP Tony Balow explained that Monolithic is comfortable increasing the growth floor to 85% because order visibility is extending with strong ordering patterns in Q1. Similar to last quarter, Monolithic did not offer any visibility into 2H, yet signals from Nvidia that Rubin will begin its initial ramp in Q3 and accelerating into Q4, alongside strong demand for server CPUs, all suggest order momentum can persist well into the year.  

The second piece of commentary relates to supply constraints, with management stating: “nothing about our outlook or anything we've said about Enterprise Data floor is because we see any constraints in the supply chain.” Simply put, supply will not be a constraint to growth, and will not prevent Monolithic from reaching or exceeding its 85% target. It also suggests that if demand strength and order momentum persists into Q2 and extends through 2H, that Monolithic likely has the supply in place to meet that upside.  

Combining these factors of persisting order momentum, increased visibility, and a lack of supply constraints with product-driven tailwinds in server CPUs and across AI accelerators and servers, all suggest this 85% growth guide could have more upside. If Monolithic raises growth by ~15 points in each of the next two quarters for ~115% YoY, Enterprise Data revenue would project out to $1.51 billion, or a $200 million raise from Q1’s guide (also smaller than the nearly $250 million raise this quarter). 

Looking ahead, Monolithic hinted at a key advantage they have as XPUs heat up to 2,000W and beyond – monolithic solutions built on a single piece of silicon: “We are the best in the market segment because we provide a total monolithic power solutions. And we can use a single piece of silicon versus our competitor use multiple piece of silicon. And that clearly shows our advantage.” Fully integrated power modules provide higher power density, and can offer more efficient power delivery even with more compact designs. Monolithic is also planning to further increase power density on smaller modules, as it plans to move from 60nm silicon to 40nm. 

This also leads into a second advantage, an ability to be flexible with integrations. Monolithic can offer fully integrated, cost effective modules to help reduce board space, accelerate chip design and increase system efficiency in increasingly complex chips, or offer a range of discrete components if its customer needs require those:  

“And we do what is the most cost effective – and how we do the integrations. And we have the capabilities to integrate or disintegrate, okay? And the integrations, we can put it in one module. And that's a huge advantage. And with the multiple other chips, and if you use particularly discrete power components, discrete power FETs, and it's very difficult to do for manufacturing the modules.”  

These monolithic and integration-based advantages may become increasingly important in driving growth alongside VPD as next-gen GPU platforms ramp and as chips progress past 2,000W. 

Communications Unexpectedly Strong, Benefitting from AI Networking 

Monolithic previously mentioned optical modules as a driver of growth for Communications (nearly a year ago at this point), yet Q1’s call was filled with discussion over 800G optical modules driving this unexpected QoQ inflection for the segment.  

Communications revenue rose 33.1% QoQ to $111.5 million in Q1, a rapid acceleration from 4.8% QoQ in Q4 and marking the segment’s fastest sequential growth since Q3 2024. On a dollar basis, Communications recorded almost the same QoQ growth as Enterprise Data at $27.8 million versus $29.3 million. Management chalked up the sequential increase to strength in both optical modules and networking switches.  

The reason why Monolithic is seeing strong growth ties in to what was discussed above, in its ability to offer a full module within optical modules with higher power density, better efficiency in more compact sizes: “why are we winning all these segment is because the power density, as I said earlier. And nobody want to waste the power and efficiency is — power density is directly related to power efficiency. And so they want a smaller size, and they want to have a higher efficiency.” 

As usual, Monolithic would not offer segment-level guidance for Q2, but hinted that Communcations is also seeing strong order momentum and will grow faster than corporate average this year: 

Quinn Bolton, NeedhamQuinn Bolton, Needham 

“I wanted to ask on the Comms segment. It was up 33% sequentially in March, it sounds like it's going to be one of the faster-growing segments in the June quarter. When I look at optical modules, I think 800-gig modules are more than doubling in '26. So — my question is, do you think the comms segment could actually grow as fast, if not faster, than Enterprise Data this year given those trends? 

Tony Balow, VP FinanceTony Balow, VP Finance 

“I think as ordering patterns have continued to be strong and extend, we still don't have them all the way through the year. So I think it's pretty tough for us to call all the way through the back half right now. But certainly we put that end market above the corporate average.” 

Monolithic does not have visibility into 2H, yet considering the strength of demand across the optics industry for >800G speeds, it’s unlikely that ordering patterns will materially slow. For example, as we highlighted in our free newsletter on Lumentum, TrendForce has predicted that “optical transceivers shipments of 800G and higher will hit 24 million units in 2025, then jump by 2.6 times to nearly 63 million units in 2026.” This 2.6X growth, or almost a 40 million increase in unit volumes, offers a strong backdrop for Monolithic’s optics-driven growth through the rest of 2026.  

Running the math on management’s commentary for Communications to grow above corporate average, coupled with calendar-Q2 guidance from other optical beneficiaries paints quite a positive picture for the segment this year.  

To start, assuming Communications grows roughly 40% YoY – below Q1’s 55.5% YoY growth but roughly 8 points faster than estimated FY26 corporate growth of 32.5% — projects the segment’s FY26 revenue out to $432.7 million.  

This is below Q1’s annualized run rate of ~$445 million, suggesting the segment sees no chance of sequential growth throughout the remainder of the year, an unlikely scenario considering the estimated >800G shipment growth and combined tailwinds from networking switches, where Monolithic has opportunities to provide power across switches, NIC cards, and other processors within the trays.  

Assuming ~15% QoQ for the segment in Q2 (as it is not entirely optics driven but also to account for potential strength in switching) and mid-single digit QoQ in 2H, revenue would project out to $520 million, up more than 68% YoY. Perhaps a bit speculative, moving the needle higher to 20% QoQ in Q2 and 10% QoQ in the back half would project Communications revenue at $555 million, up nearly 80% YoY.  

It should be noted that Monolithic’s presence in 1.6T optics is a bit unclear, as commentary this quarter primarily surrounded 800G modules; however, this may simply be due to timing as we are still quite early in the 1.6T ramp cycle and 800G could account for the bulk of growth and revenue so far. The reason optics and 1.6T (and switches) could emerge as a strong driver through 2026 is because higher power consumption at faster data rates is a critical factor to solve, especially in scale-out as optical transceiver and switch content is projected to increase sharply with Rubin. This is where Monolithic’s expertise lies in offering highly efficient power-dense modules, with CEO Michael Hsing hinting that fast execution is helping them capture the market.  

Goldman Sachs estimates that current optical transceiver (800G/1.6T) to GPU attach ratio for the GB300 racks range between 1:2 to 1:3, depending on cluster architectures in either two or three-layer configurations. With Rubin, GS projects this ratio to increase to 1:4 to 1:6, while spine, leaf and top-of-rack switch counts would increase 1.8-2.2X. This sheer content growth within transceivers supports strong optics-driven growth for Monolithic extending through 2027 as Rubin ramps.  

Near-Term Capacity Plan Increased to $6B 

Monolithic announced last quarter that it had reached its capacity target of $4 billion, and this quarter it unveiled a new near-term capacity target of $6 billion, a 50% increase. Management emphasized that this upcoming capacity will look to be geographically diverse inside and outside of China to preserve supply chain diversity. 

While comments on capacity were limited otherwise, it’s rather imperative for Monolithic to quickly expand beyond the $4 billion mark. Current revenue estimates for FY26 sit at $3.7 billion, meaning immediate capacity expansion will likely be critical in supporting future revenue upside throughout the remainder of the year, and to prevent Monolithic from capping its revenue upside.  

Historically, it can be roughly inferred that it took Monolithic around two years to expand from $2 billion of capacity (around FQ4 23) to the $4 billion mark last quarter, yet this updated plan may need to be accelerated. This could put more emphasis on higher capex, which already reached a record $70.9 million in Q1. 

Quick Note on SiC and 800V 

As we noted in our prior analysis, Monolithic was named as a key silicon provider and industry partner for Nvidia’s planned 800V DC architecture shift, which it believes will be needed to address rising power needs with next-gen rack architectures, from Rubin Ultra and beyond.   

For a quick refresher, Monolithic has begun sampling its 800V solutions and was the first to do so, per the CEO, yet it expects revenue ramps from these solutions to land in 2027 to 2028.  

800V would represent a major shift in where Monolithic Power sits as multi-phase controllers, ICs and PMICs are located on the accelerator board (or motherboard). 800V DC is about rack-level power distribution and this also shifts MPS from specializing in low voltage MOSFET-based devices to offering high voltage Sic/GaN devices in the future.  

While discussions in the past have suggested Nvidia may be seeking GaN solutions, whereas MPS is offering SiC-based ones, management this quarter emphasized that their solutions will remain SiC-based.  

Financials 

Revenue Inflecting in Q1, Accelerating in Q2 

Monolithic’s revenue began to inflect on both a YoY and QoQ basis in Q1, with Q2’s guide implying this acceleration strengthens next quarter. Q1 revenue was $804.2 million, up 26.1% YoY and 7.1% QoQ, accelerating from 20.8% YoY and 1.9% QoQ in Q4. Growth was mixed on a segment basis (discussed in more detail below), with Enterprise Data and Communications leading the sequential growth while Consumer and Industrial showed double-digit QoQ declines.  

For Q2, Monolithic guided for $890 to $910 million in revenue, accelerating to 35.4% YoY and 11.9% QoQ at the midpoint. This would mark Monolithic’s fastest sequential growth since Q4 2024. Management also added that sales channels have been very lean, implying they are shipping at demand levels and suggesting that they are not facing any supply constraints or headwinds to growth.  

For the full year, Monolithic has not provided guidance, though current consensus estimates point to 32.6% growth to $3.7 billion, a roughly six point acceleration from 26.4% growth in FY25.  

Key Segments 

Monolithic’s growth was mixed across its key segments, with Enterprise Data and Communications recording the strongest growth in Q1: 

Enterprise Data revenue was $262.8 million, up 97.7% YoY and 12.6% QoQ and accounting for 32.7% of revenue. Monolithic said the QoQ increase was driven by increased sales of power management solutions for AI and server applications, though QoQ growth decelerated from 21.9% in Q4. YoY growth accelerated 78 points in the quarter. 

Communications revenue was $111.5 million, up 55.5% YoY and 33.1% QoQ, accounting for 13.9% of revenue. This marked a sharp acceleration from 4.8% QoQ and 31.2% YoY in Q4, with growth driven by growth in 800G optical modules and networking switches. 

Storage & Computing revenue was $174.5 million, down (7.5%) YoY but up 7.6% QoQ, accounting for 21.7% of revenue. Storage was the primary driver this quarter with HDD and SSD remaining strong, while notebook revenue remained soft. Monolithic also began sampling its first high-speed interface product for DDR5, but noted not to expect this to be a contributor in 2026. 

Automotive revenue was $152.4 million, up 5.1% YoY and 0.9% QoQ, accounting for 18.9% of revenue. This decelerated from 17.6% YoY while QoQ was roughly steady after being essentially flat in Q4. Auto revenue is expected to be roughly flat in the first half before ramping later in the year.  

Consumer revenue was $54.5 million, down (4.2%) YoY and (17.5%) QoQ, accounting for 6.8% of revenue.  

Industrial revenue was $48.6 million, up 14.2% YoY but down (11.2%) QoQ, accounting for 6% of revenue. 

Margins Improving Down the Line 

Gross margins continue to remain flat with minimal expansion, with management noting headwinds arising in 2H. Operating margins are showing signs of improvement, likely tied to increasing server and optical module growth.  

GAAP gross margin was 55.3%, roughly flat YoY and QoQ, while adjusted gross margin was 55.5%, down marginally YoY and flat QoQ. Management provided some color as to the lack of expansion and flagged headwinds arising in the second half: “For the last 4 quarters, we've been flat at 55.5%, which is at the low end of our gross margin model for growth, which ranges mid-50s to upper 50s. For Q2, as you noticed, we did have the confidence to increase incrementally our gross margins, mainly because we've gotten better visibility to our backlog. We saw this happening in the fourth quarter of last year and it's continued into the first quarter of this year. So that has, again, given us some confidence. We do, however, do see some strong headwinds potentially in the second half.” Management also hinted that yield improvements in modules are still improving, but not creating much of a headwind. 

GAAP operating margin was 30%, up 3.5 points YoY and 3.4 points QoQ and coming in fairly ahead of guidance for 28.3%, suggesting more operating leverage is now appearing. Adjusted operating margin was 35.8%, up 1.1 points YoY and flat QoQ. Management noted that input component costs are rising, but they will look to offset that with price raises to maintain margins.  

GAAP net margin was 24%, up 3 points YoY and 1.4 points QoQ; adjusted net margin was 31.2%, up less than a point YoY and roughly flat QoQ. 

For Q2, Monolithic guided for a tiny step up in gross margins, projecting GAAP gross margin to be 55.1% to 55.7% and adjusted gross margin of 55.3% to 55.9%, both up marginally QoQ at midpoint. GAAP operating margin was guided to be 30.7%, up nearly 6 points YoY and less than a point QoQ; adjusted operating margin was guided to be 36.8%, up 2 points YoY and 1 point QoQ. 

EPS  

Monolithic has a strong bottom line, with adjusted EPS forecast to top $24 this year. However, considering the minimal margin expansion, EPS growth is forecast to largely track revenue growth through the year. 

Q1 GAAP EPS was $3.92, up 40.5% YoY and slightly ahead of estimates for $3.86. Adjusted EPS was $5.10, up 26.2% YoY and beating estimates by 4%, marking its largest beat since Q1 2024. 

For Q2, GAAP EPS is projected to be $4.72, accelerating to 69.8% YoY, while adjusted EPS is projected to be $5.86, accelerating to 39.2% YoY.  

For FY26, GAAP EPS is projected to be $19.40, up 51.4% YoY, while adjusted EPS is forecast to be $24.02, up 35.2% YoY. 

Cash Flows and Balance Sheet 

Cash flows were solid in Q1, with operating cash flow rebounding to north of 30% after a soft Q4.  

Operating cash flow was $250.3 million for a 31.1% margin, down from a 40.2% margin in the year ago quarter but up sharply from 14% in Q4.  

Free cash flow was $179.4 million for a 22.3% margin, down from 33.9% in the year ago quarter but up from 8.5% in Q4.  

Cash and equivalents totaled $1.37 billion, while debt remained zero. 

Inventories jumped nearly 10% QoQ to $619.2 million, while accounts receivable surged more than 18% QoQ to $302.1 million.   

Conclusion 

Monolithic is seeing strong, multi-faceted growth emerge across its Enterprise Data and (unexpectedly) Communications segment, underpinned by a shift to VPD with increasingly powerful GPUs, and its power density and cost advantages stemming from its monolithic approach and flexibility with integrations.  

Enterprise Data remained a core growth driver with revenue up 97.7% YoY and 12.6% QoQ in Q1, marking a third consecutive quarter of double-digit sequential growth. Strong order momentum and increasing visibility led management to raise the segment’s FY26 growth floor from 50% to 85%, a nearly $250 million increase, with upcoming growth levers from next-gen GPU platforms arising in 2H.  

Communications emerged as an unexpected catalyst in Q1 with 800G optical modules driving 33.1% QoQ growth for the segment. Optics could emerge as a strong secondary growth story to Enterprise Data given that >800G optics are expected to increase 2.6X this year, while optical attach rates are expected to surge with Nvidia’s Rubin platform.  

Working to expand capacity to $6 billion, a 50% increase from $4 billion in Q4, is necessary given Monolithic is quickly approaching that $4 billion level and has remained constraint-free on the supply side (so far). Quickly expanding capacity should allow MPS to capitalize on the multiple tailwinds above without becoming constrained.

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: