Earlier this month, Arm unveiled an AGI CPU to address one of AI’s biggest bottlenecks, which is orchestration. During the chatbot craze of 2023-2025, GPUs did most of the heavy lifting while CPUs had become an afterthought. Yet with agentic workloads, which is perhaps the single largest catalyst on the horizon for the AI trade in 2026 and beyond, the importance of CPUs is set to increase.

In agentic workflows, the GPU still handles inference, but between each inference call, the CPU is doing the orchestration – which are best described as handling tool calls, API requests and memory tasks. AI agents are surfacing this new constraint, which is how to prevent latency and underutilized GPUs following the exponential growth of orchestration needs.

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

Regarding how Arm fits in, the company’s expertise in lowering power requirements could matter more than the market expects. After years of supplying the architecture IP behind other companies’ CPUs, Arm is preparing to directly compete with its customers and x86 CPU competitors by transitioning to a chip designer themselves. This comes during a time when CPU cores are expected to go up 4X from 30 million CPU cores per gigawatt to 120 million CPU cores per GW.

Brief History of Arm

We covered Arm two years ago in our free newsletter, Arm Stock: AI Chip Favorite Is Overpriced, and how its mobile background and power-efficient focus positioned it well for increased AI adoption.

Arm offers the most popular CPU architecture in the world with more than 350 billion chips shipped since its inception, with the company essentially building a monopoly in the mobile CPU market with 99% market share stemming from its ‘heterogenous compute’ design and RISC architecture.

This design helped facilitate lower power requirements as the architecture allows different CPU parts to work together for improved efficiency, with workloads to be allocated across both high-performance and low-performance CPU cores to lower energy by balancing performance.

Arm is translating this expertise in power efficient chips to the data center, powering both Nvidia’s Grace and Vera CPUs, as well as custom CPUs from Amazon, Google and Microsoft. For example, Google says its Axion CPUs can offer up to 65% better price-performance and 60% better energy efficiency versus x86 alternatives, while Microsoft’s Cobalt 100 and Amazon’s Graviton4 CPU both have shown significant performance and price-performance advantages over competing x86 products.

The Role of CPUs in Agentic AI and the Coming 4X Increase in CPU Cores

Agentic AI represents a natural evolution from the query-and-response based nature of chatbots, to a more complex system capable of running dozens of different tasks and tools autonomously to reason through a problem and provide a response.

As a result, CPUs will play a more integral role in agentic and multi-agent systems to help solve how the system will schedule dozens of concurrent API requests and tool calls across independent agents, as quickly as possible with minimal delay. This is the orchestration constraint: how dozens of agents can make hundreds of concurrent requests needed to complete their independent workflows without causing significant latency or GPU underutilization.

Multi-agent systems are also expected to drive an exponential increase in token generation, which Arm estimated at up to a 15X increase in tokens per user, due to the increase in tool calls and API requests associated with each agent. This is expected to drive CPU core demand much higher, at a time where key x86 suppliers AMD and Intel battle growing supply constraints.

Arm CEO Rene Haas detailed at the company’s latest ‘Arm Everywhere’ event that a typical AI data center of today will feature around 30 million CPU cores per GW. Solving the flow bottlenecks of agentic inference will require “CPUs near the head node, CPUs next to the accelerator rack, more CPU racks inside the data center,” driving CPU core demand as much as 4X higher to 120 million cores per GW, per Haas.

CPU-Centric AI Systems

We are seeing evidence of how agentic AI and the importance of CPUs are beginning to affect incoming architecture designs. If you did not catch last week’s free stock newsletter, Nvidia Stock Prediction: The Path to a $20 Trillion Market Cap is Strengthening, I want to relay the importance of one sentence: “The goal is no longer to simply sell faster and more powerful chips, but to deliver superior economic value at the system level relative to custom silicon (in other words, let the battle begin).” 

Think of the Groq LPX racks that cost Nvidia a record-breaking $20 billion that it is soon deploying, addressing the memory-intensive decode phase of inference to significantly accelerate token throughput. This system-level focus is now moving to CPUs, with Nvidia pivoting to deploy its Grace and Vera CPUs as standalone racks. Nvidia says that when “paired with Rubin GPUs as a host CPU, or deployed as a standalone platform for agentic processing, Vera enables higher sustained utilization by removing CPU-side bottlenecks that emerge in training and inferencing environments."

Arm’s big announcement was not just that it was launching its first in-house silicon and CPU rack platform, but that it will now be directly competing on the standalone CPU system side. Arm’s foray into the standalone rack market offers a new choice for hyperscalers and data center operators to seamlessly deploy CPU-centric racks alongside GPUs, customize the CPU-to-GPU ratio to optimize for agentic orchestration and enable power-efficient AI inference at scale from start to finish. It will also provide another outlet to avoid vendor-lock in to Nvidia’s ecosystem with both air-cooled and liquid-cooled CPU racks.

Key Advantages of Arm’s ‘AGI CPU’ for Agentic AI Workloads

Arm also marked its long-awaited foray into physical chip development with its ‘AGI CPU’, launched at its Arm Everywhere event last week. The company’s pivot into physical CPU and rack development is one the AI industry will watch with great anticipation given Arm’s history of owning significant IP in the mobile space combined with the company setting out to solve agentic AI’s orchestration challenges.

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.

2X Performance of x86 and Record CPU Rack Density

The new CPU features up to 136 of Arm’s highest-performing Neoverse V3 cores, drawing just 300W of power in a single-unit (1U) dual-node server (blade) featuring two chips. This compares to x86 chips, such as AMD’s fifth-gen EPYC CPUs, which deliver 128 to 192 cores per chip but at 390W to 500W in a two-unit (2U) rack.

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.

Comparison of sustained AI workload performance shows Arm’s AGI CPU outperforming x86 CPUs (with and without SMT) in performance per thread, total threads per rack, and performance per watt — highlighting Arm’s efficiency advantage for agentic AI orchestration. Source: ArmArm

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:

“So if you have a CPU that can draw less power, it could be just as performant, but use less power, it means you have more leftover for everything else that you want to do. That means more inference and more compute. That means more intelligence.”

mid

One Petabyte of Memory and Low-Latency Chip Design

However, Arm architected the new chip with other key optimizations in mind, notably on the memory side. The AGI CPU features 96 PCI Gen6 lanes with CXL 3.0 connectivity, which Arm says allows the new CPU to be attached to any accelerator of choice, allowing flexibility of deployments with Nvidia or AMD GPUs or custom chips.

The chip also features up to 6TB of memory, providing more than 1 petabyte (PB) of low-latency memory in the liquid-cooled 200kW rack. Arm is achieving this extreme low latency of <100 nanoseconds from memory via a dual chiplet design, with each chiplet having both the memory and I/O directly onboard to avoid multiple links across the silicon.

Arm says the new chip’s performance advantage over x86 could enable “up to $10B in capex savings per GW of AI data center capacity,” making it a compelling option for current and future agentic AI-optimized deployments to save money, save power and avoid Nvidia-lock in from its accelerator-agnostic nature.

Looking beyond AMD and Intel, Arm’s in-house design may also outperform Nvidia’s design based on Arm’s IP as the AGI CPU offers a step up in core count and memory at the rack level versus Nvidia’s Vera CPU, despite Vera running on custom Arm ‘Olympus’ cores.

Nvidia says the Vera CPU rack features 256 CPUs, each packing 88 cores for a total of 22,528 cores across the rack; this offers 45,056 threads with multi-threading. Vera racks will also offer up to 400 TB of memory capacity with 315 TB/s of memory bandwidth with PCIe Gen 6 lanes and CXL 3.1 connectivity, whereas Arm offers 2.5X the capacity though just ~274 TB/s of bandwidth at 6GB/s per core.

Quantifying the Impact of CPU-Optimized Agentic AI

Arm’s executives provided some initial commentary about long-term growth projections, seeing the new CPU as a rather lucrative revenue opportunity by the turn of the decade.

Arm expects the AGI CPU to generate $1 billion in revenue in fiscal 2027 and 2028 (June 2026 to March 2028), with the first generation of the AGI CPU now available and ramping further in 2H.

Arm outlined a second generation of the chip tentatively on deck for 2027, likely contributing to some of the revenue generation in 2028 and beyond. On a broader long-term view, Arm estimates the AGI CPU line could generate $15 billion in revenue in 2031, or ~60% of its 2031 revenue target of $25 billion, with a third generation potentially launching before then, though Arm provided no timeline for that chip.

This represents around a 15% share of what Arm believes will be a $100 billion TAM for data center CPU. This is a higher TAM than what Nvidia CEO Jensen Huang discussed at GTC, saying: “I'm not expecting CPUs to be that much and call it because CPUs just don't add up to much, okay. And so you could say CPU is another 5%” – or around a $50 billion TAM on $1 trillion.

The challenge in scaling from a fresh launch to $15 billion in revenue in a matter years is definitely doable, considering the current pace of capex growth and tens of billions in AI accelerator revenue flowing to Nvidia, AMD and Broadcom each quarter. However, the path ahead is not free of challenges, as Arm must still compete against Nvidia’s new standalone deployments, as well as the custom Arm-based CPUs hyperscalers have already built – Amazon’s Graviton, Google’s Axion, and Microsoft’s Cobalt.

Financials

Arm’s Financials Offer Superior Gross Margins

Arm has a profitable business model that constitutes licensing revenue and royalty revenue. The company reported a strong gross margin of 97.6% in Q3 FY2026 ending December.

Arm’s gross margin has remained consistently near 98% over the past five fiscal quarters, underscoring the strength and durability of its high‑margin licensing and royalty business model. Source: Company IR

Arm reported a GAAP operating margin of 14.9% and an adjusted operating margin of 40.7% in the recent quarter. The difference between adjusted operating margin and GAAP operating margin is that the company is a recent IPO and has high stock-based compensation of $285 million or 23% of revenue.

ACV Grew by 28%

The company’s licensing and other revenue grew by 25% YoY and down (2%) QoQ to $505 million in FQ3. The revenue growth decelerated from 56% YoY and 10% QoQ growth in the previous quarter. Licensing revenue has been lumpy, and management mentioned on the earnings call that it varies quarter to quarter due to the timing and size of high-value deals. So, annualized contract value or ACV is considered a key indicator of the underlying licensing trend. ACV grew by 28% YoY and 1% QoQ to $1.62 billion in FQ3. ACV grew by 28% in the past three quarters.

Arm’s annual contract value growth accelerated from single‑digit levels in FY2025 to a sustained 28% year‑over‑year pace through FY2026, signaling strengthening demand and improved visibility in its core licensing business. Source: Company IR

AGI CPU Could Become a $15 Billion Revenue Line by FY2031

Arm is launching its own chip which will help the company grow revenues and profits. The company expects more than $1 billion in revenue from this business over the next two years with the majority of revenue to be recognized in FY2028 ending March.

Arm expects an exponential ramp to $15 billion in revenue in FY2031. The strong growth is primarily driven by solid demand and the increasing complexity of chips will lead to a significant rise in average selling prices. Gross margins are expected to be at least 50% and the adjusted operating margin of over 30% for the chip business.

The company’s licensing and royalty revenue is expected to be $10 billion in FY2031 ending March, with a gross profit margin of 99% and over 65% adjusted operating margin. Management increased this long-term adjusted operating margin guidance by 500 basis points.

The company’s total revenue in FY2031 is expected to be $25 billion and adjusted EPS is expected to be over 9. The current consensus revenue estimates for FY2031 of $21.18 billion and adjusted EPS of $8.39 are lower by 18% and 7.3%, respectively.

Consensus revenue estimates project Arm’s revenue to grow from $4.9 billion in FY2026 to $21.2 billion by FY2031, with year‑over‑year growth accelerating sharply in the later years, reflecting expanding licensing, royalty, and AGI CPU contributions. Source: Seeking AlphaSeeking Alpha

Royalty Revenue Is Entering a Sustained 20% CAGR Phase

The company’s royalty revenue has grown at about a 14% CAGR in the past five years. It has accelerated to over 20% in the past two years as Armv9 and Compute Subsystems (CSS) have started to ramp. Looking forward, management expects that the royalty revenue Compound Annual Growth Rate (CAGR) will be 20% over the next five years. The company’s royalty rates have doubled from Armv8 to Armv9 architecture, and again to CSS.

Valuation

Arm is currently trading at a P/S ratio of 36.3 and a forward P/S ratio of 28.6. The company is trading significantly higher than its other semiconductor peers like Broadcom’s forward P/S ratio of 14.4 and Nvidia’s forward P/S ratio of 11.6.

The company’s revenue growth is expected to accelerate in the next five years compared to the previous period. The company’s revenue CAGR has been 19.3% from FY2021 to FY2026E. Analysts expect revenue to grow at a CAGR of 34% from FY2026E to FY2031E and will be even higher at 38.5% if we use the $25 billion management guidance. However, when looking at the AI segment of many semiconductor peers, the growth rate does not stand out, per se, to justify the high valuation. Rather, the consistency of licensing and royalties' revenue does stand out, and this recurring revenue will create a nice baseline when you combine higher growth from their merchant CPUs.

Arm’s forward price‑to‑sales multiple has remained elevated over the past year, with the forward P/S above 30x, reflecting strong expected future revenue growth. Source: YChartsYCharts

Cash Flows

The company’s cash flows have been lumpy due to high working capital and high capex to support the long-term growth. However, with the expected strong future profit growth, the cash flows should improve. The company also has a strong balance sheet with cash & short-term investments of $3.54 billion and no debt.

Conclusion:

Arm may be a right place-right time stock, to where the years where Arm sat uneventful compared to peers could payoff as they are more prepared to aggressively compete in the AI market. The valuation is an unknown, and an important aspect given merchant CPU revenue will take time to scale.

To offset this, Arm offers consistent (and rare) recurring revenue, strong margins and far less speculation than, say, an AI software stock that must invest heavily to remain competitive and prove they can build user traction. Instead, Arm offers decades of IP excellence, a 99% penetration in mobile, all while its x86 competitors like AMD and Intel struggle to keep up with their growing backlogs.

The AI trade has been in question lately, yet when the market regains confidence in AI again, Arm will undoubtedly be part of a select, core AI semiconductor pack positioned to deliver for the long haul. 

Since our inception in May 2020, I/O Fund has delivered a cumulative return of 326%— if we were a hedge fund, we’d rank #1 and if we were a tech ETF or Mutual Fund, we’d rank #3 in the United States. To get our Top 15 AI stocks, real-time trade alerts, and weekly 1-hour webinars from a proven team in AI and tech stocks, sign up now.sign up 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.

Recommended Reading:

• Nvidia Stock Prediction: The Path to a $20 Trillion Market Cap is Strengthening
• Nvidia Stock to See New Growth Catalyst; 35X Faster AI with Groq 3 LPX
• Palantir Stock is Out of Favor, but is the Growth Engine Still Intact?
• “Tech Bubble” Warnings Cost Investors a 550% Nasdaq-100 Run

Last week at GTC, Jensen Huang stated that Nvidia has a path to $1 trillion in cumulative sales across the Blackwell and Rubin generations from 2025 through 2027. If you follow Nvidia’s stock closely, this isn’t new information; rather it’s roughly aligned with what analyst forecasts already had baked in. 

The distinction is crucial for investors as separating what’s already priced in from what can make a meaningful difference in stock returns. The latter typically offers alpha, while the other potentially sets up an investor for losses (hence the saying: “buy the rumor, sell the news” “buy the rumor, sell the news”). 

The math for Nvidia to see $1 Trillion in Revenue was already there. 

We must go back to October to more fully understand why the statement that Nvidia has visibility to $1 trillion in revenue through 2027 is anti-climactic.  

Last October, Huang stated that the combined revenue from Blackwell and Rubin was an estimated $500 billion through the end of 2026. Our firm modeled something similar nearly two years earlier, when my original Nvidia $10 trillion market cap thesis was published, stating we would see a $320 billion data center segment in 2026 (FY2027).  

Beth Kindig of the I/O Fund first laid out the case for Nvidia reaching a $10 trillion market cap in June 2024 — a view Jensen Huang later expressed publicly in March 2026 nearly two years later. 

Blackwell revenue was $184 billion in 2025 when you combine compute and networking, along with the $320 billion expected in 2026, comes out to the $500 billion quoted at GTC in October. This means our model proved correct roughly 16 months before the CEO confirmed it. When I first made the $320 billion data center prediction in June 2024, it resulted in 56% upside, versus a 15% decline by the time the CEO effectively confirmed the thesis in October 2025.  

Being early can pay off at many points along a stock’s trajectory. Of course, this is modest compared to the I/O Fund getting ahead of the Street on Nvidia in 2018–2019, which led to returns of more than 4,000%. But each milestone still matters when you’re talking about one of the world’s most valuable companies, as generating outsized returns becomes far more difficult at this stage. 

Now consider the trajectory to $1 trillion laid out by Huang. Bridging from $500 billion (approximately $184 billion in 2025 and $320 billion in 2026) to $1 trillion implies about $500 billion in CY2027, or about 54% year-over-year growth for about $125B a quarter.  

The Street had largely modeled this in, with FY28 (CY27) quarters at $114B, $119B, $125.5B, and $132B for a total of $490.5B for the fiscal year ending in January.  

In other words, there was not much alpha in the comment, which helps explain why the stock didn’t move much from the “blockbuster” $1 trillion comment. 

Chart showing Nvidia’s long‑term revenue forecast for FY2027–FY2031, including annual estimates, year‑over‑year growth, and multi‑period trend revisions that highlight consistently rising analyst expectations.

On the heels of the $1 trillion cumulative revenue comment, Huang publicly stated this week on a podcast that he sees a path to a $10 trillion market cap for Nvidia.  

This is the exact thesis I first published in June of 2024 in the article “Here’s Why Nvidia Stock Will Reach $10 Trillion Market Cap by 2030," 

“Nvidia has a market cap of $3 trillion today. We believe Nvidia will reach a $10 trillion market cap by 2030 or sooner through a rapid product road map, it’s impenetrable moat from the CUDA software platform, and due to being an AI systems company that provides components well beyond GPUs, including networking and software platforms.” 

Admittedly, 16 months later, that is no longer a contrarian call. Once the CEO goes on record with a number like that, it’s a sign the $10 trillion market cap narrative is transitioning from offering alpha to becoming increasingly priced in.

mid

Which is exactly why I pushed the thesis further out, publishing a new article last November that Nvidia has a path to end the decade with a $20 trillion market cap. 

How Nvidia Gets to $20 Trillion. 

The $20 trillion market cap will not come from GPU unit growth alone, though unit growth remains very important. Rather, the value proposition will increasingly focus on economic output. This marks a tremendous shift for how Nvidia is evaluated. 

As the AI market shifts toward inference, Nvidia’s product cycles will be optimized around token economics such as throughput, latency, power efficiency and cost per token. The goal is no longer to simply sell faster and more powerful chips, but to deliver superior economic value at the system level relative to custom silicon (in other words, let the battle begin).  

Leading up to this, Nvidia was competing on performance metrics, and MLPerf benchmarks still matter of course. But going forward, workload economics and system-level efficiency will play a much larger role in how their systems are evaluated. 

Historical Mispricing — How Analysts Missed Nvidia by 4X 

Nvidia’s stock price is highly dependent on upward revisions – more so than any company that I can recall. To put it plainly, Wall Street’s estimates have consistently underestimated this company, and this gap is critical for where investors can continue to find an edge. 

Three quarters after Nvidia’s breakout earnings report in May of 2023, analysts had eight months to price in the AI trade. Consensus for the fiscal year ending in 2028 was set at $138.3 billion. Today, that same estimate stands at $480 billion.  

Further out, the error widens with Jan 2031 estimates of $208 billion versus where they are today at $757.6 billion or nearly 4X higher. 

Chart showing Nvidia’s consensus revenue revision trend from 2018 to 2026, highlighting steady upward revisions for FY2027–2032. Key estimates—such as $369.42B for FY2027, $479.97B for FY2028, and $757.63B for FY2031—have climbed sharply, reflecting rising expectations for Nvidia’s AI, data center, and accelerator demand. The sustained increase in long‑term forecasts reinforces bullish sentiment around Nvidia stock and its multi‑trillion‑dollar market cap outlook. and its multi‑trillion‑dollar market cap outlook.

Only One Year Ago, Nvidia Revenue Estimates Were Far Too Low 

If we repeat this exercise and look back exactly a year ago, we will see that analysts continue to miss the target. Just one year ago, the fiscal year ending January 2028 was expected to see revenue of $294 billion. Today, consensus is at $480 billion. All else equal; those revisions could potentially represent alpha of 63%.  

If you look at FY2031, estimates were $343.5 billion compared to $757.6 billion – meaning analyst estimates were off by 2X from where they are today. 

Chart showing Nvidia stock revenue revision trends from 2018–2026, highlighting steadily rising analyst estimates for FY2027 through FY2032 as expectations for Nvidia’s AI and data center growth continue to increase.

How This Ties into Nvidia Reaching a $20 Trillion Market Cap by 2030 

My $20 trillion market cap thesis for Nvidia is grounded in certain assumptions, primarily that Nvidia reaches $930 billion in data center revenue in a single year prior to the close of the decade combined with a price-to-sales ratio of 22X – a few points below the stock’s 3-year median P/S of 28X.  

Right now, analyst estimates sit at $757 billion for the fiscal year ending January 2031. However, given the estimates have doubled in the past year alone, the 23% difference in estimates compared to my firm’s base case of $1 trillion seems achievable. 

As I recently emphasized in an interview with Bloomberg Asia, analysts revising estimates intra-quarter is one of the most important catalysts for this stock. 

What’s Pressuring Nvidia’s Valuation in 2026 

Despite odds favoring Nvidia ending the decade at $930 billion or more in annual revenue, the more pressing issue is valuation. The stock has been trading at a significant discount to its historical valuation, yet buyers are not stepping in. Meanwhile, Broadcom is trading right well above its 3-year median and AMD is two points above its 3-year median. 

Chart comparing Nvidia, AMD, and Broadcom P/S ratios and 3‑year median P/S ratios from 2023–2026, showing Nvidia trading below its historical valuation while Broadcom and AMD remain closer to their 3‑year medians.

Source: YCharts 

I’ve heard some outlandish theories about this disconnect, but I believe the reason Nvidia is seeing a weaker valuation is fairly straightforward: the inference market offers immense opportunity for Nvidia yet is expected to lower Nvidia’s overall percentage of the AI accelerator market. Yes, this means Nvidia’s near-monopoly is set to end. 

I’ve covered this dynamic in detail in my Broadcom stock analysis here and and AMD stock analysis here.Broadcom stock analysis here and and AMD stock analysis here. 

According to TrendForce, custom silicon represents 20.9% of the market in 2025 yet is expected to expand to 27.8% of the market in 2026. Adding to the competitive pressure, AMD is expected to release Helios MI400s in the second half of 2026, which could further eat into Nvidia’s GPU market share, adding to the pressure of custom chips gaining 7 points with architectures like Google’s TPUs. 

Chart showing global AI server shipment share by accelerator type from 2023 to 2026, based on TrendForce data. GPUs remain dominant, while custom silicon grows from 20.9% of the market in 2025 to a projected 27.8% in 2026, highlighting accelerating adoption of ASIC‑based AI infrastructure.

Nvidia Versus Custom Silicon is Overly Simplistic 

The assumption that losing AI accelerator market share should result in a lower valuation is overly simplistic for four reasons. 

Capex Growth Expands the Entire AI Accelerator Market: The first reason is that capex is the primary multiplier. Because capex continues to grow the overall pie, the market will expand faster than any single architecture can absorb. Compute demand is compounding; a shrinking slice of a rapidly growing pie can still mean explosive revenue growth. 

GPUs Remain the Most Flexible Architecture for New Workloads: The second reason is that GPUs remain the default when workloads change. The versatility of GPUs is a competitive advantage as Big Tech does not always know what next quarter or next year will bring. Consider too that custom silicon is not only inflexible yet takes years to design. When a new model architecture emerges or workloads shift, GPUs step up in ways that custom silicon can’t. For example, during the reasoning model era, architectural breakthroughs such as OpenAI releasing o1 and DeepSeek releasing R1 required significantly more compute at inference. Custom silicon is a better choice when workloads are stable versus rapidly evolving like we saw with reasoning models and inference-time scaling.

Nvidia Monetizes at the System‑Level: Nvidia is monetizing the system rather than just the chips, evidenced by the explosive networking growth of 263% this past quarter. On that note, speaking of Broadcom, Nvidia’s management team stated that Nvidia is now the world’s largest Ethernet company, overtaking former Ethernet giant Broadcom, and this was accomplished in just a few years’ time.  

Fourth – and most importantly, the value of what Nvidia offers is rapidly shifting from raw compute to token economics. If Nvidia continues to lead in performance per watt and performance per rack, its premium valuation can persist. Big Tech will prioritize unit economics, which means if Nvidia’s systems cost 2-3X more, the goal will be to produce more tokens per watt than the alternative to offset the premium. 

This point is critical for how Nvidia plans to defend its positioning over the next few years.  When token volume scales 10X, 100X or even 1000X, Nvidia’s ability to sell more units increases, along with the platform of more networking, more software and more tooling. 

In that framework, it will be less about which systems cost $40,000 versus $15,000 and more about which platform delivers better economics at scale. 

The Market Is Not Pricing in Nvidia’s Inference Opportunity 

Last week, in the article “Nvidia Stock to See New Growth Catalyst; 35X Faster AI with Groq 3 LPX, I argued that the $1 trillion revenue comment through 2027 wasn’t the headline. The real development was the new Groq 3 LPX racks delivering up to 35X higher throughput per megawatt. 

Why the Groq 3 LPX Integration Is a Major Catalyst 

The 256-chip LPX rack introduces Groq’s unique SRAM‑based architecture that allows Nvidia to offload decode‑phase workloads and massively increase token throughput. This primarily targets trillion‑parameter LLMs, million-token context, and multi‑agent systems, which are bottlenecked less by compute and more by how efficiently a system can move data and generate tokens. Paired with the new Vera Rubin GPUs, Nvidia claims this architecture can deliver up to 35X higher throughput per megawatt, with seamless integration into Vera Rubin deployments.   

The Groq acquisition is aimed to solve the limiter of inference throughput per watt, where memory bandwidth can become the gating factor to token output and cost. Nvidia is preparing to position its GPUs to be among the best inference options available, utilizing Groq’s unique SRAM-based architecture to significantly turbocharge token throughput and accelerate inference performance.   

Nvidia expects Groq will help drive up to a 15X increase in tokens per second, directly translating into higher tokens per megawatt, which is already scaling by a factor of 10X between Blackwell and Rubin. If these claims hold true, then cheaper inference will unlock more usage, and more usage should lead to higher revenue and higher profits as the AI monetization wave plays out. 

Nvidia is positioning its new Groq 3 LPX racks as a ‘token accelerator’ functioning in tandem with Vera Rubin GPUs to significantly boost token throughput and address the upcoming multi-agent future. The Groq LPUs are not meant to replace GPUs in inference workloads, but rather compliment them by optimizing for memory-intensive decode.   

Off the bat, Nvidia expects that combining Rubin GPUs and Groq racks will drive a substantial increase in token throughput, with Nvidia VP Ian Buck claiming the combination “moves us from a world where 100 tokens per second is a reasonable throughput to one of 1500 TPS or more for AI agent intercommunication.”   

To visualize this, anything over 100 TPS feels near-instantaneous, such as for chatbot users; in other terms, this would represent 1,500 words per second, or ~275X the average human reading speed. This distinction and shift from 100 TPS to 1,500+ TPS is more important than it might appear, as 100 TPS is optimized for human consumption, such as chatbot outputs, while 1,500 TPS is optimal for machine consumption, such as multi-agent communication, autonomous long-form reasoning and real-time AI systems that all require continuous, low-latency token generation.  

Rubin + Groq Racks and the $300B Annual Revenue Opportunity 

The introduction of the Groq LPUs as the seventh chip in Rubin’s co-design also represents a natural shift in Nvidia’s rack scale strategy that may help deepen its moat, where it disaggregates compute and bandwidth via different specialized architectures to optimize inference at the rack and system rather than chip level. Nvidia is moving quickly with the new combined infrastructure, with Groq chips in volume production at Samsung and CEO Jensen Huang saying they would be shipping around the Q3 timeframe.  

 Nvidia foresees a rather large opportunity from this new integration, with CEO Jensen Huang explaining at GTC that he believes the Groq racks could account for up to 25% of a data center footprint to extend the performance and value of Vera Rubin, as well as future chips. Overall, Huang added that combining Vera Rubin with the Groq LPX racks could unlock a $300 billion annual revenue opportunity for customers.   

While some analysts had cautioned that reaching the upper end of this would depend on buyer appetite and ‘ultra-premium’ tiers such as up to $150 per million tokens (nearly ~10X of GPT 5.4’s cost), the scale of the opportunity reflects Nvidia's belief that inference-optimized rack-level systems will become a key part of future AI infrastructure buildouts. 

Read more about the importance of the Groq 3 LPX racks and why the acquisition represents an important catalyst for the Nvidia’s stock. an important catalyst for the Nvidia’s stock.  

Nvidia Stock Broke Minor Support at $176 

Quick note on pricing: 

Nvidia topped in late October 2025, which was one of several occurrences that signaled growing weakness in the broader market. Since then, we have seen numerous large block trades hit Nvidia’s price while in this consolidation zone. This implies larger institutions are either accumulating a big move higher or distributing before another bout of volatility hits. 

Chart showing Nvidia’s daily price action with key support near $170, Fibonacci extension targets, and Elliott Wave counts highlighting potential breakout and downside levels.

Nvidia is barely holding the lower end of this consolidation zone, and even broke minor support at $176. This is not ideal; however, final support is $170. Below this zone, and we would likely see Nvidia move toward the $155 – $135 region. If Nvidia can instead break over $200, then it would strongly suggest that it is setting up for the next leg higher.  

Conclusion: 

As discussed, analysts had difficulty pricing in the training market while it was in motion despite a clear product roadmap of the incoming GPU generations. Based on the history of analyst estimates being up to 4X too low 5-years out and 2X too low 1-year out for the training market opportunity, the chances that Nvidia’s inference opportunity is correctly priced in is fairly low in my opinion. This is especially true at this moment in time because inference has not made a dent yet on the return on capital for Big Tech companies. Furthermore, Groq is a recent acquisition and its impact on Nvidia’s revenue trajectory is not fully modeled yet.  

That is not to say that current estimates are 4X too low, but rather that the 23% gap between current estimates of $757 billion and the $930 billion gets closed is a reasonable assumption. It’s also reasonable that if Nvidia can prove strong execution in the inference market, the company will make a case for its premium valuation again. 

The I/O Fund’s portfolio is on fire this year with a lesser-known AI semiconductor stock up 170%+ YTD and another lesser-known AI semiconductor stock up 90%+ YTD. We are the team that predicted Nvidia would become the world’s most valuable company in 2019 – years before Street consensus, and we have dominated the AI trade in recent years.170%+ YTD and another lesser-known AI semiconductor stock up 90%+ YTD. We are the team that predicted Nvidia would become the world’s most valuable company in 2019 – years before Street consensus, and we have dominated the AI trade in recent years. 

In fact, our high-performing tech portfolio with cumulative returns of 326%, which would place us as #1 if we were a hedge fund and #3 if we were a tech ETF or mutual fund. To get a 60-page analysis on our Top 15 AI Stocks, sign up now.326%, which would place us as #1 if we were a hedge fund and #3 if we were a tech ETF or mutual fund. To get a 60-page analysis on our Top 15 AI Stocks, sign up 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 NVDA at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

• Nvidia Stock to See New Growth Catalyst; 35X Faster AI with Groq 3 LPX
• Palantir Stock is Out of Favor, but is the Growth Engine Still Intact?
• “Tech Bubble” Warnings Cost Investors a 550% Nasdaq-100 Run
• My Top 2026 Stock Pick for the AI Boom

At GTC this week, Jensen Huang stated the revenue opportunity for Nvidia’s artificial intelligence chips may reach at least $1 trillion through 2027, up from a previous target of $500 billion. While that grabbed most of the headlines, there was another jaw-dropping statistic that will set the stage in the coming years – which was the ability to drive up to 35X higher throughput per megawatt with its new Groq 3 LPX racks.  

The 256-chip LPX rack introduces Groq’s unique SRAM‑based architecture that allows Nvidia to offload decode‑phase workloads and massively increase token throughput. This primarily targets trillion‑parameter LLMs, million-token context, and multi‑agent systems, which are bottlenecked less by compute and more by how efficiently a system can move data and generate tokens. Paired with the new Vera Rubin GPUs, Nvidia claims this architecture can deliver up to 35X higher throughput per megawatt, with seamless integration into Vera Rubin deployments.  

In some ways, this acquisition draws parallels to Nvidia’s $6.9 billion acquisition of Mellanox, which my firm covered for premium research members in 2020. Mellanox was a strategic purchase to clear the bottleneck at the time on GPU performance, which was scale-out networking. By combining Nvidia’s GPUs with the strength of Mellanox’s InfiniBand, smart NICs and switching, Nvidia was able to turn accelerators into clusters by removing the limiter at that time (scale-out networking). 

The Groq acquisition is aimed to solve a different limiter, which is inference throughput per watt, where memory bandwidth can become the gating factor to token output and cost. Nvidia is preparing to position its GPUs to be among the best inference options available, utilizing Groq’s unique SRAM-based architecture to significantly turbocharge token throughput and accelerate inference performance.  

Nvidia expects Groq will help drive up to a 15X increase in tokens per second, directly translating into higher tokens per megawatt, which is already scaling by a factor of 10X between Blackwell and Rubin. If these claims hold true, then cheaper inference will unlock more usage, and more usage should lead to higher revenue and higher profits as the AI monetization wave plays out. 

Below, we cover how Nvidia, the de facto leader in training, is now shifting its focus to inference architecture as the next catalyst. 

Why Nvidia is Rethinking AI Inference Architecture 

Year after year (and generation after generation), Nvidia has proven that it can consistently deliver massive efficiency gains on inference throughput and token processing speed. For example, Nvidia’s GB200 NVL72 boosts per-GPU throughput by up to 30X versus the HGX H100 platform, while the GB300 NVL72 boasts up to a 50x increase in AI factory output via a 10X increase in tokens per second per user and a 5X increase in throughput per MW.  

Versus the Blackwell NVL72 systems, Nvidia says Vera Rubin can deliver up to 10X more throughput per megawatt, rapidly compounding performance gains from its Hopper generation in just three years.  

Source: Nvidia 

However, the more important piece of the puzzle is not just the rapid pace of these throughput gains, but how Nvidia can continue to deliver exponentially more throughput gains and how Nvidia will accelerate inference workloads further. The key answer to this is Groq, and ‘inference disaggregation’ at the rack level.  

Inference disaggregation refers to splitting up the two-step process of token generation, prefill and decode, instead of running both steps together. The prefill phase processes the entire input token sequence in parallel and stores information in the KV cache, relying heavily on GPU compute and not as much on memory (yet). The decode phase generates the output tokens one by one in a sequential manner, relying on the KV cache and previous tokens, making it extremely reliant on memory bandwidth and capacity to rapidly access cached tokens. When discussing how AI workloads are memory constrained, it comes from the decode phase.  

When both prefill and decode shared the same hardware (the GPUs), the two would interfere with each other and lead to delays, as a new prefill request would either force the system to pause decodes and prioritize the prefill, or run both again at the same time, elongating response times.

mid

With inference disaggregation, prefill and decode can be scaled and scheduled on different optimized hardware via Nvidia’s Dynamo; in this case the Rubin GPUs handle prefill and Groq LPUs handle decode. With disaggregation and the LPU’s massive memory bandwidth, Nvidia CEO Jensen Huang says the two combined can deliver up to 35X higher throughput per MW on trillion-parameter LLMs: 

“What if we disaggregated inference altogether with a piece of software called Dynamo? What if we rearchitected the way that inference is done in the pipeline, so that we could put the work that makes perfect sense on Vera Rubin and then offload the decode generation, the low latency, the bandwidth limited challenged part of the workload for Groq. And so we united, unified processors of extreme differences, one for high throughput, one for low latency. 

It still doesn't change the fact that we need a lot of memory. And so Groq, we're just going to add a whole bunch of Groq chips, which expands the amount of memory it has. And so if you could just imagine, out of 1 trillion parameter model, we have to store all of that in Groq chips. However, it sits next to NVIDIA Vera Rubin, where we could hold the massive amounts of KV cache that's necessary in processing all of these agentic AI systems. It's based upon this idea of disaggregated inference. We do the prefill, that's the easy part, but we also tightly integrate the decode. 

So the attention part of decode is done on NVIDIA's Vera Rubin, which needs a lot of math and the feed forward network part of it, the decode part is done — the token generation part is done — on the Groq chip. The 2 of them working tightly coupled together over today, Ethernet with a special mode to reduce its latency by about half. 

And so that capability allows us to integrate these 2 systems. We run Dynamo, this incredible operating system for AI factories on top of it, and you get 35x increase, not to mention additional new tiers of inference performance for token generation the world has never seen.” 

Inference disaggregation is not an entirely new concept, but rather it is the way Nvidia is approaching disaggregation that makes this move noteworthy. Instead of seeing disaggregation as a separate, service-layer optimization, such as what AWS is eyeing with its recent partnership with Cerebras, Nvidia is pushing to directly embed disaggregation into the rack to maximize throughput. 

Inside Groq’s SRAM Architecture and Its Massive Bandwidth Advantage 

Groq’s chips feature a completely different memory-based architecture than Nvidia’s GPUs, utilizing SRAM instead of HBM. This unique architecture gives Groq’s language-processing units (LPUs) a significant advantage in the decode phase and in low-latency, high-query inference workloads from extremely higher bandwidth.  

SRAM offers a major trade-off versus DRAM and HBM when it comes to memory storage capabilities within AI accelerators. Unlike typical DRAM, SRAM does not require capacitors and stores data without the need for periodic refreshing, as long as power is available. Because of this design, SRAM can offer the fastest memory access speeds with minimal latency, though at the cost of having a mere fraction of the capacity of HBM chips – the LPUs have just 500MB of capacity versus 288GB of HBM in its Rubin GPUs. 

Despite having just 500MB of capacity, each LPU delivers 150 TB/s of SRAM bandwidth — this is nearly 7X the 22 TB/s HBM bandwidth per Rubin GPU. In the rack-scale configuration, the Groq 3 LPX delivers an astounding ~2.5X increase in total scale-up bandwidth and a 25X increase in SRAM bandwidth versus HBM bandwidth. 

The Groq 3 LPX combines 256 individual LPUs for a total of 128GB of SRAM capacity, yet it offers 40 PB/s of SRAM bandwidth versus 1.6 PB/s of HBM bandwidth in the Vera Rubin NVL72. Total scale-up bandwidth reaches 640 TB/s versus 260 TB/s in the NVL72. This also dwarfs the upcoming NVL576 rack which offers just 4.6 PB/s of HBM bandwidth. 

This 25X increase in bandwidth is precisely the reason why Nvidia is aiming to offload decode and low-latency workloads to the LPX racks, as more bandwidth means more weight data can be processed per second, which, at its core, means more tokens can be generated per second.   

Nvidia Positioning Groq 3 LPX as a ‘Token Accelerator’ 

Nvidia is positioning its new Groq 3 LPX racks as a ‘token accelerator’ functioning in tandem with Vera Rubin GPUs to significantly boost token throughput and address the upcoming multi-agent future. The Groq LPUs are not meant to replace GPUs in inference workloads, but rather compliment them by optimizing for memory-intensive decode.  

Off the bat, Nvidia expects that combining Rubin GPUs and Groq racks will drive substantial increase in token throughput, with Nvidia VP Ian Buck claiming the combination “moves us from a world where 100 tokens per second is a reasonable throughput to one of 1500 TPS or more for AI agent intercommunication.”  

To visualize this, anything over 100 TPS feels near-instantaneous, such as for chatbot users; in other terms, this would represent 1,500 words per second, or ~275X the average human reading speed. This distinction and shift from 100 TPS to 1,500+ TPS is more important than it might appear, as 100 TPS is optimized for human consumption, such as chatbot outputs, while 1,500 TPS is optimal for machine consumption, such as multi-agent communication, autonomous long-form reasoning and real-time AI systems that all require continuous, low-latency token. 

The introduction of the Groq LPUs as the seventh chip in Rubin’s co-design also represents a natural shift in Nvidia’s rack scale strategy that may help deepen its moat, where it disaggregates compute and bandwidth via different specialized architectures to optimize inference at the rack and system rather than chip level. Nvidia is moving quickly with the new combined infrastructure, with Groq chips in volume production at Samsung and CEO Jensen Huang saying they would be shipping around the Q3 timeframe. 

 Nvidia foresees a rather large opportunity from this new integration, with CEO Jensen Huang explaining at GTC that he believes the Groq racks could account for up to 25% of a data center footprint to extend the performance and value of Vera Rubin, as well as future chips. Overall, Huang added that combining Vera Rubin with the Groq LPX racks could unlock a $300 billion annual revenue opportunity for customers.  

While some analysts had cautioned that reaching the upper end of this would depend on buyer appetite and ‘ultra-premium’ tiers such as up to $150 per million tokens (nearly ~10X of GPT 5.4’s cost), the scale of the opportunity reflects Nvidia's belief that inference-optimized rack-level systems will become a key part of future AI infrastructure buildouts. 

AI Monetization is Arriving, and Tokens are the Currency 

As we had covered in our Bloom Energy analysis, My Top 2026 Stock Pick for the AI Boom, the real risk to the AI economy lies in the physical constraints of scaling these AI ambitions — not in compute availability from companies like Nvidia or Broadcom, and certainly not in Big Tech’s software capabilities, but in power availability, thermal management, and infrastructure that were never designed for this magnitude of demand. 

This is the core challenge the AI industry now faces, and this means the most important equation for the upcoming inference-driven monetization wave is how many tokens can be generated, served, and monetized within a fixed power envelope. With Vera Rubin and the new Groq racks, Nvidia is increasingly orienting its GPU roadmap around that point, aiming to exponentially increase tokens per second per watt. It is about creating a platform that is not just faster, but able to deliver more of that monetizable output (tokens) per watt.  

Nvidia CEO Jensen Huang made this point extremely clear at GTC, explaining that “everybody is looking for land, power and shell. Once you build it, you are power limited. Within that power limited infrastructure, you better make for darn sure that your inference — because you know inference is your workload, and tokens is your new commodity, that compute is your revenues — that you want to make sure that the architecture is as optimized as you can.” With Vera Rubin, Huang emphasized that Nvidia is “going to take our token generation speed, token generation rate from 2 million to 700 million, a 350x increase” for GW-scale AI data centers. To roughly estimate what this could look like using a cost of $1 per million tokens, this would represent a step function from $2 in revenue to $700, before applying that to scale.  

While achieving the 350X increase in token generation may be reserved for hyperscalers operating at maximal scale and efficiency, this can be translated across the industry to emerging neoclouds and data center operators alike. Think of it this way — for a data center with a fixed 100MW power envelope, the amount of users and tokens that can be served with Vera Rubin and Groq racks are multiples higher than Blackwell and other generations.  

This means that driving TPS per MW higher is essentially a multiplier on revenue and margins, as more tokens under the same power footprint translates directly to higher revenue and lower costs per token processed. As Nvidia puts it, up to 10X more tokens per MW and up to 10X lower cost per million tokens with Rubin versus Blackwell – put differently, if it cost a cloud provider $10 to serve 1 million tokens that generated $15 in revenue, it would net $5 in profit. With Rubin, if it can generate 10 million at that same $10 cost, profit could reach as much as $140. 

The above scenario assumes high revenue for AI inference, which may compress as the AI inference market is built out. Yet, even with a ~67% compression in token costs (revenue) from $15 to $5, there will still be $40 in profit at the 10 million tokens, or an 8X increase. 

This does not mean that Nvidia is not immune to rising competition from custom silicon, as hyperscalers continue to turn towards custom chips to optimize for specific inference workloads and dramatically lower serving costs. For example, Alphabet lowered Gemini’s inference serving costs by 78% through 2025 via model optimizations, utilization and efficiency improvements, and its newest TPU generation is likely to drive further cost reductions through 2026. Meta also recently expanded its custom silicon roadmap with four new chips, focusing on ranking and recommendation model performance, genAI models, and inference via increasing HBM bandwidth and capacity each generation.  

Among the hyperscalers and startups with the deepest pockets, custom silicon will likely remain a key choice in AI deployments for its ability to offer much lower costs and high performance for optimized workloads. However, for neoclouds and companies without capital to build and deploy ASICs at scale, Nvidia is creating an extremely compelling value proposition by offering a platform optimized for token throughput at scale.   

Conclusion 

Nvidia is leveraging Groq’s SRAM-based LPUs and extreme memory bandwidth to significantly accelerate inference and token throughput by offloading the decode phase to the new chips. When paired with Vera Rubin, Nvidia claims this architecture can deliver up to 35X higher throughput per megawatt for trillion parameter LLMs. As the AI industry now faces power and infrastructure constraints rather than compute, the key differentiator in the upcoming AI inference monetization wave will be how to extract the highest number of tokens per megawatt to maximize revenue.  

For years, Nvidia’s dominance has been synonymous with training. Now, the company is making it clear it wants to own inference, which is the part of AI that actually scales into everyday usage and recurring revenue. The market latched onto Jensen Huang’s $1 trillion AI chip visibility through 2027, but the bigger tell may be what Nvidia is optimizing for next: tokens per megawatt. If Groq 3 LPX helps Rubin deliver anything close to the claimed throughput gains, Nvidia’s next growth leg won’t be about building bigger models—it will be about making inference cheap enough that demand explodes. 

The I/O Fund predicted Nvidia would become the world’s most valuable company in 2019 – years before Street consensus. Today, our team runs a high-performing tech portfolio with cumulative returns of 326%, which would place us as #1 if we were a hedge fund and #3 if we were a tech ETF or mutual fund. To get a 60-page analysis on our Top 15 AI Stocks, sign up now.Nvidia would become the world’s most valuable company in 2019 – years before Street consensus. Today, our team runs a high-performing tech portfolio with cumulative returns of 326%, which would place us as #1 if we were a hedge fund and #3 if we were a tech ETF or mutual fund. To get a 60-page analysis on our Top 15 AI Stocks, sign up now.

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

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

Recommended Reading:

• Palantir Stock is Out of Favor, but is the Growth Engine Still Intact?
• “Tech Bubble” Warnings Cost Investors a 550% Nasdaq-100 Run
• My Top 2026 Stock Pick for the AI Boom
• I/O Fund Jumps to 326% Cumulative Return, Ranking Among Wall Street’s Best

Palantir stock has softened alongside a broader software selloff, raising questions about whether its premium valuation is still justified. Beneath the muted price action, the company delivered sharp revenue re‑acceleration, triple‑digit U.S. commercial growth driven by AIP, expanding profitability, and record-leading metrics that suggest its long‑term growth engine could still be intact. 

Palantir is Underperforming AI Hardware Stocks 

Software stocks have been out of favor, with some down as much as 45%. Palantir’s stock is down about 10%, better than most software peers, yet it has still lagged AI hardware with stocks like Micron up as much as 50% year-to-date. Despite the softer price action, Palantir’s valuation still trades meaningfully above its 5-year and 3-year median. 

Chart comparing Palantir’s forward price‑to‑sales (PS) ratio against its 3‑year and 5‑year median PS ratios. The forward PS line rises significantly above both long‑term medians, reaching 49.85 by March 2026, illustrating how Palantir’s valuation has expanded beyond historical averages. Source: YChartsYCharts

Palantir Stock Earnings: Revenue Re‑Acceleration and Margin Expansion     

Palantir’s underlying fundamentals continue to challenge investors who believe real value is found in cheap stocks. This past quarter, the company offered a rare re-acceleration in growth with revenue accelerating to 70%, an impressive 57-point acceleration over the last ten quarters, while guiding revenue to accelerate further to 73.6% in Q1. US commercial momentum remained unphased, with revenue accelerating 16 points sequentially to 137% YoY, surpassing the $500 million mark in the quarter.

Chart illustrating Palantir’s year‑over‑year revenue growth from Q2 2022 through Q1 2026G, highlighting a powerful acceleration. This underscores Palantir’s strong multi-year revenue momentum and reinforces the narrative of accelerating demand by its AI platform. 

I’m a growth investor through-and-through, yet it would be a mistake to think Palantir’s strength is found in the top line as I could point toward a dozen or more stocks with higher growth rates. Rather, it is Palantir’s ability to balance hypergrowth with profitability that sets it apart from SaaS peers. While revenue accelerated, profitability expanded alongside it: adjusted operating margin reached 57.4%, and adjusted EBITDA margin came in at 57%.  

Palantir’s Rule of 40 (revenue growth + adjusted operating margin) rose to 127%, up 46 points YoY and 13 points QoQ, putting the company in a class of its own. 

Palantir’s Rule of 40 came at 127% in Q4 2025, up 13 points QoQ and 46 points YoY. 

Cash generation remained equally strong with adjusted free cash flow of $791M for a 56% margin in the quarter—while management guided for FCF margin to expand further next year. 

US Commercial: The Key Driver of the Palantir Stock Bull Case 

Palantir’s AIP-driven US commercial segment remains the company’s core revenue driver. In a series of previous analyses, my firm has covered key elements as to why the artificial intelligence platform (AIP) is able to drive such strong growth. To briefly review, Palantir’s Artificial Intelligence Platform (AIP) integrates generative AI with operational data and workflows, and, when combined with Palantir’s other platforms, Foundry and Apollo, it provides an AI service mesh that can run hundreds of microservices, scale compute via its Rubix engine, and orchestrate updates through Apollo. 

The differences matter as unlike traditional AI-enabled database or business intelligence competitors, Palantir can operate effectively even when data sets are incomplete or fragmented—situations where most models struggle. In that regard, traditional business intelligence companies require a complete data set, whereas Palantir can handle situations where one isn't available. You can think of the competitive advantage as actionable depth, as Palantir has described it: “the reasoning that goes into decision-making, not just data.” 

mid

Additionally, Palantir’s knowledge graph, referred to as Ontology, is a distinct advantage. The graph offers better context than a large language model would on its own – or as Palantir states, it’s “the reasoning that goes into decision-making.” Palantir made key upgrades to AIP with the introduction of AI-forward-deployed engineers (FDEs) and the AI Hivemind, and brought Ontology to the edge, enabling deployment on mobile devices. 

Read more about Palantir’s products in our analysis “Palantir Stock 2026 Forecast: Is Its High Valuation Sustainable?” and Palantir Stock Forecast 2025Palantir Stock 2026 Forecast: Is Its High Valuation Sustainable?” and Palantir Stock Forecast 2025 

How AIP Is Powering Triple‑Digit Commercial Growth 

In the prior earnings report, US Commercial revenue grew 29% QoQ and 121% YoY to $397 million in Q3, accelerating sharply and supporting the view that AIP is driving existing expansions and new customer conversions.   

The acceleration continued this past quarter with US commercial revenue rising 137% YoY and 29% QoQ to $507 million in Q4. This led to a $2 billion annualized run rate in the quarter, up from a $1 billion run rate at the start of 2025. 

With that said, Palantir’s International commercial revenue is not nearly as strong as US commercial with revenue growth of just 8% YoY and 12% QoQ to $170 million in Q4.

Chart displaying Palantir’s US Commercial year‑over‑year revenue growth from Q1 2022 through Q4 2025, highlighting the dramatic expansion of demand driven by the company’s AIP platform. 

Palantir Stock Forecast: What Key Metrics Signal for 2026 and Beyond 

Unlike 99% of other companies reporting this season, Palantir’s call offered little substance to sift through, with more clues on its growth opportunities hidden in the key metrics. 

Palantir’s NRR expanded 5 points to 139% in Q4; on a YoY basis, NRR has risen 19 points. In the earlier thesis, Palantir’s NRR had already been expanding (to 134% in Q3), with management emphasizing that the metric does not capture revenue from new customers acquired over the last twelve months—meaning upside can still be building beneath the surface.   

RPO then surged in Q4—up 62% QoQ and 143% YoY to $4.21 billion—while billings rose 91.1% YoY and 21.5% QoQ to $1.49 billion. Both of these key metrics witnessing this sharp step-up in tandem provides further confidence in Palantir’s 2026 accelerations panning out with the potential for upside to its initial guidance as each quarter progresses. 

Palantir also booked record TCV of $4.26 billion, up 138% YoY, with commercial TCV of $2.6 billion, up 161% YoY and 83% QoQ.

Chart showing Palantir’s remaining performance obligations (RPO) from Q1 2023 through Q4 2025, highlighting a substantial increase from $0.94 billion to $4.21 billion over the period. This sharp expansion signals a rapidly strengthening multi‑year demand pipeline, reinforcing Palantir’s visibility into future revenue and supporting expectations for continued acceleration in 2026. 

The Other Side of Palantir's Story: Government Likely to Strengthen

Palantir recorded its best week since August as conflict in Iran broke out, as analysts were quick to point out that the conflict would help Palantir prove its value to the US military and continue to see strong government-driven deal momentum.  

Palantir’s exact role in the conflict is a bit unclear, as some reports suggest that Palantir’s Maven Smart Systems, powered by Anthropic’s Claude AI, was utilized to integrate fragmented data from multiple US agencies, from drone footage to satellite imagery and radar signals, to help map military movements, what other LLMs are not necessarily capable of doing. However, the Pentagon’s growing disputes with Anthropic, which saw the government deem Anthropic as a ‘supply chain risk’, could hinder its role, as defense contractors such as Lockheed Martin are expected to work to remove Anthropic from their supply chain per the Pentagon’s orders. 

Despite that uncertainty, Palantir’s platform does remain model agnostic and analysts believe there are ‘adequate alternatives’ to Claude that could be utilized, while the broader theme of conflict and the Defense Department saying they would award “bigger, longer contracts” for proven weapons systems support growth in Palantir’s government pipeline. As a reminder, Palantir had already won a $10 billion contract with the Army last summer, and any indication that increased government deal volumes for AI solutions could flow through to Palantir.  

Palantir’s government remained strong with 60% YoY and 15% QoQ growth in the fourth quarter to $730 million, and government accounted for nearly 52% of revenue in Q4. Palantir highlighted mission impact across the DoD and momentum in civil agencies, including an up to $448 million contract with the US Navy to modernize the shipbuilding supply chain and accelerate delivery of naval vessels. 

Margins and Cash Flow Are Reinforcing the Operating Model 

As stated earlier in the analysis, Palantir’s Rule of 40 score expanded 46 points YoY to 127%, which the company defines as revenue growth plus adjusted operating margin. Adjusted operating margin in Q4 was a record 57.4%, while Palantir guided for adjusted operating margin to remain strong next quarter at 56.8% at midpoint. 

Cash flows remain best-in-class with management guiding for adjusted free cash flow margin to expand in 2026 from an already strong 51% in 2025, projecting adjusted FCF up more than 77% YoY to $3.925–$4.125 billion. 

Why EV/EBITDA Shows Palantir Is Cheaper Than It Appears

When we revisit valuation, we see that Palantir’s strong bottom line has resulted in an EV/EBITDA that is as low as the April 2025 selloff – reinforcing the fact that Palantir is more of a bottom-line story than a top line story in terms of what truly sets the stock apart: 

Chart showing Palantir’s EV/EBITDA ratio that is as low as the April 2025 selloff. Source: YChartsYCharts

Will Palantir’s 2026 Setup for Beat/Raise Potential Mirror 2025? 

For 2026, Palantir initially guided for fiscal 2026 revenue to accelerate from 56.1% to nearly 61% YoY, driven by US commercial revenue accelerating six points to >115% YoY. Driving such an acceleration at these growth rates is undeniably difficult, yet there are hints that Palantir could go above and beyond these figures by this time next year. 

Looking back to 2025, Palantir’s beat/raise pattern offers a framework for how 2026 could play out if key metrics continue to strengthen. In 2025, management’s initial guide proved conservative as the year progressed; the open question now is whether Palantir can deliver similar upside from a higher revenue base.  

For example, in Q4 2024, Palantir had initially guided for roughly 30.8% YoY growth in FY25. This was then raised to 36% in Q1, then again to 44.7% in Q2. By Q3, Palantir’s FY25 guide was raised to 53.5%, before ending the year with growth of 56.2%, more than 25 points faster than originally anticipated four quarters prior. FY26 is already starting off at 60.6% projected growth, with all four quarters to go.  

Technical Analysis: Where Palantir Stock May be Headed Next 

Palantir has been tracing a large 5-wave pattern off the 2022 low. Regardless of how the internal waves are organized, the pattern remains incomplete to the upside — meaning the larger uptrend is still intact, and the current weakness is most likely just a correction. That said, given the magnitude of this 5-wave structure, investors should expect corrections of a larger degree than what they have experienced during the earlier stages of this advance. 

Price action currently supports two scenarios: 

Primary Count — A sustained break above $195 would invalidate the primary count and instead project a 5th wave advance toward the $240 region. This would complete the larger 3rd wave patern, as the final push higher will be met with lower volume and momentum as we push higher. Once completed, it will give way to a multi-month 4h wave decline. 

Alternative Count — We are in a significant 4th wave decline. The current bounce is expected to fail somewhere in the $186 — $195 range before rolling over into a final leg lower. Target support for the completion of this 4th wave sits between $119 and $87. The line in the sand for the entire bullish thesis is $66. A close below that level would cast serious doubt on the continuation of the larger uptrend.

Chart displaying a long‑term Elliott Wave analysis of Palantir (PLTR) from late 2023 through 2026, mapping out impulsive waves (1–5) followed by an ABC corrective structure. The price action shows a strong uptrend through waves (1), (2), (3), (4), and (5), with Fibonacci extension targets marked above current levels, including 236%, 250%, and 261.8% projections ranging roughly from the mid‑$200s to nearly $400. The chart also outlines potential corrective zones labeled (A), (B), and (C), with support areas between approximately $87 and $120 highlighted as key downside levels. Trendlines and moving averages show broader upward momentum, while future price projections extend into 2027.

Conclusion: 

Palantir’s Q4 report showed that the company’s AIP-driven momentum remains robust with no signs of slowing, further supported by most key metrics accelerating in unison. Palantir’s NRR expanded 5 points to 139%, its Rule of 40 score expanded 46 points YoY to 127%, and record TCV and RPO were the cherry on top of a strong quarter.    

Palantir also guided for revenue to accelerate to nearly 61% YoY in 2026, driven by US commercial revenue accelerating to >115% YoY. Driving an accelerate at multi-billion dollar scale is difficult, yet the company’s key metrics suggest growth rates may continue to move higher. 

Lastly, on valuation, the bottom-line ratio is more favorable than the sales multiple implies, as earnings power has improved due to the company’s margin expansion. Nobody has a crystal ball, but the company’s fundamentals suggest Palantir is positioned to trade higher than where it sits today if execution remains intact.  

Since our inception in May 2020, I/O Fund has delivered a cumulative return of 326%— if we were a hedge fund, we’d rank #1 and if we were a tech ETF or Mutual Fund, we’d rank #3 in the United States. 326%— if we were a hedge fund, we’d rank #1 and if we were a tech ETF or Mutual Fund, we’d rank #3 in the United States.    

Being early to many lesser-known AI winners helped us to achieve these results. To get our Top 15 AI stocks, real-time trade alerts, weekly webinars and deep-dive research from a proven team in AI and tech stocks, Sign up now.Top 15 AI stocks, real-time trade alerts, weekly webinars and deep-dive research from a proven team in AI and tech stocks, Sign up now.

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:

• “Tech Bubble” Warnings Cost Investors a 550% Nasdaq-100 Run
• My Top 2026 Stock Pick for the AI Boom
• I/O Fund Jumps to 326% Cumulative Return, Ranking Among Wall Street’s Best
• Bitcoin After the Cycle Peak: What Comes Next and How We’re Positioning

The market is fixated on when Big Tech will generate economic value from the $650 billion+ being poured into AI data center expansion annually. The market is missing the point. Monetization has never been Big Tech’s weakness as explosive revenue growth and high margins have defined their businesses for decades. While execution risk always exists, these companies remain the world’s most reliable operators at scale.  

Instead, the real risk to the AI economy lies in the physical constraints of scaling these AI ambitions — not in compute availability from companies like Nvidia or Broadcom, and certainly not in Big Tech’s software capabilities, but in power availability, thermal management, and infrastructure that were never designed for this magnitude of demand.  

Nvidia’s GPU road map is bringing about an immediate need to overhaul data centers as most data centers today are incapable of powering the kilowatts required for rack-scale systems. Blackwell power requirements of 120 kW for the GB200s and 140 kW for the GB300s represents a 2X increase from the H200s 70kWs. As we look out over the next 1-2 years, it’s expected Nvidia will ship rack-scale systems requiring 300-600 kW – or a 5X increase from what was needed per system in the first half of 2025. 

Therefore, it is not enough to say the AI economy needs more power, but rather it needs power urgently. These are two entirely different matters; for example, the first could be supported by the expansion of nuclear power and the electrical grid, but the latter cannot. In fact, combining these two is something very few companies can do. 

This leads me to my Top Pick for 2026 – Bloom Energy.  

Bloom Energy offers onsite power generation through solid oxide fuel cells that are behind the meter to reduce dependency on the grid.  By providing behind-the-meter generation, Bloom reduces reliance on utility infrastructure and accelerates time-to-power for customers. An added benefit is the United States is the largest producer of natural gas, therefore, Bloom does not struggle to secure supply given the United States has large, well-developed gas supplies and pipeline infrastructure. 

The I/O Fund’s History on Bloom Energy 

We first covered the surging power demand from AI data centers in our June 2024 newsletter, AI Power Consumption: Rapidly Becoming Mission-Critical, with Bloom Energy quickly rising to the top of our list for its ability to solve the critical time-to-power constraint. From there, we drilled deep into this stock in early 2025 yet used technical analysis to hold off and wait for the April lows for our first entries. 

We made our initial buys at $16.64 and $17.04 in April 2025, helping position the stock as our biggest winner of 2025. During 2025, we held the position at allocations as high as 15%, with real-time trade alerts sent to our Members throughout the year. Today, the stock trades at $160.90, while many of Wall Street’s most renowned firms followed later in 2025 and entered at significantly higher prices. 

With that said, it requires strong conviction in not only Bloom Energy’s positioning but also the sheer pressure from AI’s primary bottleneck to believe the stock could see a repeat year of strong performance. Below, I lay out why I believe Bloom Energy is setting up to do exactly that. 

Chart showing Bloom Energy (BE) the I/O Fund's 2025 buy and trim actions, with green arrows labeled ‘Started Our Position’ and ‘Bought,’ and red arrows labeled ‘Trimmed’ during upward price movements.

Power is the #1 Constraint for AI Data Centers 

Before we drill deeper into Bloom Energy’s unique positioning, it’s well worth the time to revisit the mounting pressure in the AI energy bottleneck. Consider that companies like Microsoft and Meta are spending hundreds of billions annually on AI, with tens of billions allocated to Nvidia’s Blackwell GPUs.  

Any delay in powering these systems deepens both risk and market perception, as it not only pushes out revenue and profits but also extends the period in which Big Tech remains underwater on capex returns. A long timeline for power availability increases both timing risk and financial leverage.  

Competitively speaking, power availability is also an advantage as providers that can energize and deploy GPUs faster will have a meaningful head start over competitors stalled by power constraints. While the concept is straightforward, the stakes are immense, as it is not only the scale of these AI investments to consider but also the fierce competition to secure power can amplify the consequences of a delay.

mid

AI leaders are in unison this is the predominant challenge the industry faces. Commentary from executives at hyperscalers, neoclouds, Bitcoin miners, colocation providers and commercial real estate firms all point to power as a key constraint (and consideration) facing the market this year and next.  

CBRE said in its H1 2025 North America Data Center Trends Report that “power availability and infrastructure delivery timelines remained the most decisive factors shaping site selection, leasing activity and pricing across all major U.S. markets.” 

Equinix executives stated that “the amount of power we need isn't sitting around on the grid. And so we are planning, and I think most people in the room that are doing data center development are ensuring you have clear line of sight to that power before you take down any land or plan any data center capacity.” 

A survey by Bloom Energy of 44 hyperscaler and colocation developers found that availability of power was the number one consideration for new site selection, with 84% of respondents placing that in the top 3 with an average rating of 7.8 out of 10. 

Amazon CEO Andy Jassy said that “you see some of the constraints and they kind of exist in multiple places, [but] the single biggest constraint is power.” Microsoft CEO Satya Nadella said Microsoft needs “power in specific places so that we can either lease or build at the pace at which we want.” 

Google Cloud’s Thomas Kurian explained that “more powerful chips… take a lot more power. And power is, in many cases, a short resource.” Arm’s CEO Rene Haas has said that without improvements in efficiency, "by the end of the decade, AI data centers could consume… 20% to 25% of U.S. power requirements. Today that’s probably 4% or less." 

AI Data Center Power Demand Forecast for 2030: Projected to Surge 8,050% 

Data center power demand is expected to grow at an accelerated clip through the end of the decade and beyond, driven by the two main drivers of more powerful GPUs and surging growth in inference. 

In 2024, we had revealed that “Wells Fargo is projecting AI power demand to surge 550% by 2026, from 8 TWh in 2024 to 52 TWh, before rising another 1,150% to 652 TWh by 2030. This is a remarkable 8,050% growth from their 2024 projected level. AI training is expected to drive the bulk of this demand, at 40 TWh in 2026 and 402 TWh by 2030, with inference’s power demand accelerating at the end of the decade.”AI power demand to surge 550% by 2026, from 8 TWh in 2024 to 52 TWh, before rising another 1,150% to 652 TWh by 2030. This is a remarkable 8,050% growth from their 2024 projected level. AI training is expected to drive the bulk of this demand, at 40 TWh in 2026 and 402 TWh by 2030, with inference’s power demand accelerating at the end of the decade.” 

However, we have more data from the IEA that projects global data center power demand to more than double from ~415 TWh in 2024 to ~945 TWh by 2030 under its base-case scenario, or growth of roughly 530 TWh. The agency’s AI ‘lift-off’ scenario projects demand reaching 1,250 TWh, or growth of ~835 TWh, more closely aligning with Wells Fargo’s projection.  

Regardless of where AI demand falls relative to these projections, the trend and takeaway is rather clear – AI is set to drive data center power demand much higher by 2030. We can also look at this from a GW perspective, with numerous projections all pointing to substantial growth in data center capacity. 

Boston Consulting Group forecasts 45 GW of growth in global data center power demand in just three years from 82 GW in 2025 to 127 GW by 2028, with this more than doubling from 2023’s 60 GW.  

Overall, BCG expects generative AI power demand to rise at a 65% CAGR from 2023 through 2028, with AI training increasing at a 30% CAGR and inference rising at a rapid 122% CAGR. Under BCG’s scenario, gen AI will account for more than one-third of global data center power demand by 2028.   

Stacked bar chart showing global data center power required to meet projected computing demand from 2020 to 2028, rising from 43 GW to 127 GW. Bars display segments for GenAI, other AI + HPC, and traditional workloads, with CAGR rates shown for 2020–2023 and 2023–2028. 

On the other hand, McKinsey projects data center capacity will rise ~2.5x to 219 GW by 2030, up from a similar ~82 GW baseline in 2025. McKinsey projects AI training and inference demand to rise at a nearly 29% CAGR by 2030, driven by inference, rising at a 35% CAGR from ~21GW  to ~91GW. In total, AI would be contributing ~112 GW of the projected total 137 GW demand growth.  

This is quite a substantial amount of projected capacity growth over the next three to five years. But, more importantly, what level of capex does this require?  

Given our prior calculations for each GW to cost between $30 to $38 billion from the ground up (and now towards >$40 billion with Nvidia’s Blackwell Ultra), building out 112 GW of AI training and inference capacity by 2030 could necessitate as much as $4.3 trillion in capex.  

Looking more directly at the power side, and more specifically what Bloom’s TAM could be in the realm of on-site generators, Bernstein analysts estimate that generators and turbines could account for ~6% of capex per GW. This would equate to roughly $1.8 to $2.4 billion per GW, or in the long-term scenarios noted above with 112 GW of growth tied to AI, as much as $258 billion. BofA takes a more conservative approach at roughly ~2% of capex per GW, or ~$800 million, placing this 112GW forecast opportunity at nearly $90 billion. 

Why Bloom Energy Stands Out in a Crowded Energy Industry 

Time-To-Power Solutions for AI Infrastructure 

Our primary message has been “time to power” for Bloom, and the company continues to stand out for this very reason as it is finding strong product market fit in AI data center power needs. This is a key advantage as on-site power is becoming more of a necessity as grid constraints and connection timelines rise.  

As we had noted above, the industry is expecting to see significant demand growth over the next few years, yet the primary hurdle is that the grid is not able to keep up with such rapid demand in a short timeframe. For example, PJM (home to Data Center Alley in Northern Virginia as well as fast-growing data center markets in Pennsylvania and Ohio) fell short of its reliability requirements in the last two capacity auctions, with the most recent 2027/28 planning year, falling ~6.6GW short.  

A similar dynamic is unfolding in Texas, where ERCOT’s interconnection queue has reached roughly 226 GW as of mid-November, nearly quadruple the 63 GW recorded at the end of 2024. Of that total, approximately 165 GW comes from data center projects targeting approval by 2030, whereas ERCOT added only 23 GW of new capacity in 2024–25 — about 10% of the queued demand. 

This further validates Bloom’s positioning by enabling new data center projects to come online sooner with on-site, behind the meter power without sitting in interconnection queues for years at a time. Bloom has already proven that it can quickly establish data center power solutions in a rapid manner, completing shipments to Oracle Cloud Infrastructure in just 55 days of its 90-day delivery request.  

Its fuel cells are also fuel-flexible and can run on natural gas, biogas, or hydrogen, and provide continuous power with 99.9-99.999% reliability metrics. They are also modular in nature and can scale from 20 MW to 500 MW+, allowing flexibility in deployments and ease of scaling. Bloom is also continuously improving on price-performance, stating that its fuel cells have seen double digit YoY cost reductions each year for the past ten years, and a 10X increase in power production in the same footprint versus ten years ago.  

Bloom Energy vs. Gas Turbines for Data Centers 

Bloom also has an advantage over gas turbines when it comes to on-site power demand, as GE Vernova had stated in December that its gas turbines are sold out through 2028 with less than 10% remaining in 2029, meaning any new orders would not be delivered for another 3+ years. Nuclear has been floated as a solution to meet GWs of demand, though restarting facilities take years and SMRs are not expected to be commercially viable at scale until the 2030s.  

From Oracle to Quanta: Bloom’s Rapid AI Power Deployment  

Doubling capacity this year to 2GW gives Bloom an outlet to meet immediate-term demand from data centers throughout this year into 2027. 

A subsidiary of American Electric Power (AEP) had entered into a deal with Bloom in November 2024 for the purchase of 100MW of solid oxide fuel cells with the option to purchase 900MW more, for a total of 1GW.  

On January 4, AEP disclosed that its subsidiary had exercised a “substantial” portion of this option for $2.65 billion as part of its plan to develop and build a fuel cell power generation facility in Wyoming. This is rumored to be for Crusoe and Tallgrass Energy’s 2.7GW ‘Project Jade’ campus currently under development, said to be scalable to 10GW in the future.  

Bloom also quietly secured a major purchase order from AI server manufacturer Quanta Computer’s subsidiary QMN at the very end of 2025, with it buying three fuel cell microgrid systems for ~$502 million to provide reliable back-up power to its B16, B18 and B19 plants in California to ensure that high-value AI server manufacturing is not interrupted in times of inclement weather or wildfires.  

As a quick recap of some of Bloom’s prior deals, it had announced a partnership with Brookfield in October, which will see the asset management firm invest up to $5 billion in Bloom’s fuel cell tech and potentially finance deals for Bloom to be the preferred on-site power provider for Brookfield’s AI data center portfolio. This spans Brookfield’s $100 billion global AI Infrastructure program announced in November, in partnership with Nvidia and the Kuwait Investment Authority, which is aiming to build ‘AI factories’ on Nvidia’s Vera Rubin stack under Brookfield’s new cloud company Radiant.    

Bloom also struck a deal with Oracle back in July to deploy its fuel cells for onsite power at select Oracle Cloud Infrastructure (OCI) data centers, with Bloom serving as the first and second source of power for a single data center, with management hinting that this partnership is likely to expand over time.  

Financials 

2025 Revenue up 37.3%, 2026 Guided to Increase 58%  

Bloom once again delivered revenue more than 20% above analysts' expectations, with Q4 revenue of $777.7 million beating estimates by 20.5%. This represented 35.9% YoY growth, decelerating from 57.1% YoY growth in Q3; however, sequential growth was very strong at 49.8% QoQ, accelerating from 29.4% QoQ in Q3 – this is because Q4 is typically Bloom’s seasonally strongest quarter. The company announced its total current backlog at $20 billion, including $6 billion in product backlog, up 2.5X, and $14 billion in service backlog, up 1.5X. 

Bar chart titled ‘Revenue YoY’ showing Bloom Energy’s year‑over‑year quarterly revenue performance from Q4 2023 to Q4 2025, with Q4 2025 highlighted at +35.9% YoY, reflecting revenue growth to $777.7 million driven primarily by strong AI demand. 

Source: Company IR 

For the full year, Bloom reported record revenue of $2.02 billion, driven by significant AI data center growth and demand from commercial and industrial sectors. This represented 37.3% YoY growth. 

For 2026, Bloom guided for a sharp acceleration to 58% YoY at the midpoint of its guide for $3.1 to $3.3 billion, supported by its capacity expansion towards 2GW. This is a notable 24% beat over the consensus estimates and also would represent just 16% of its total $20 billion backlog.   

Product Revenue grew by 35% YoY and 66% QoQ in Q4 2025 

Products, installation, and service revenue growth remained solid in Q4, though electricity revenue continued to decline. 

Product revenue was $638.5 million in Q4, up 35.4% YoY and 66.1% QoQ, though YoY growth did decelerate from 64.4% as Q4 faced a much tougher, seasonally strong comp. FY25 product revenue increased 41.1% YoY to $1.53 billion.  

Installation revenue was $67.3 million in Q4, up 86.4% YoY, though this did decelerate from 105.2% growth in Q3. FY25 installation revenue increased 66.9% YoY to $204.1 million. 

Service revenue was $61.7 million, up 14.7% YoY, decelerating slightly from 15.5% in Q3. FY25 service revenue increased 6.9% YoY to $228.3 million. 

Electricity revenue did reaccelerate in Q4 but growth continued to decline. Q4 revenue declined (5.3%) YoY to $10.2 million, improving from Q3’s (25.1%) decline. FY25 electricity revenue was $60.3 million, up 14.2%.

Bar chart titled ‘Growth by Segment’ showing Bloom Energy’s year‑over‑year growth across Products, Installation, Service, and Electricity from Q4 2023 through Q4 2025. 

Source: Company IR 

Bloom Energy Q4 2025: Margins Rebound Sharply QoQ 

Bloom’s margins showed a sharp sequential rebound in Q4 but remained lower on a YoY basis. Full year margins showed expansion across the board with the exception of GAAP net margin, while GAAP operating margin moved a bit further into positive territory albeit remaining razor thin. 

• Bloom Energy’s adjusted gross profits grew by 10.2% YoY to $248 million with an adjusted gross margin of 31.9%, an improvement of 1.5 percentage points sequentially.  
• GAAP operating margin was 11.3% in Q4, up 9.8 points QoQ. Adjusted operating margin was 17.1%, an improvement of 8.2 points QoQ. Bloom noted that it continues to focus on reducing product cost and driving operating leverage, which will likely be much more visible in 2026 based on its current guide. 
• The company’s adjusted net profits grew by 13.1% YoY to $134.1 million with an adjusted net margin of 17.2%, compared to 20.7% last year and a significant improvement over 6.8% in the previous quarter. Bloom’s margins are improving, primarily driven by operational efficiency, product cost improvements, and operating leverage. 

Bar chart titled ‘Margins’ showing Bloom Energy’s adjusted gross margin and adjusted operating margin by quarter from Q4 2023 to Q4 2025, with Q4 2025 displaying 31.9% adjusted gross margin and 17.1% adjusted operating margin. 

Source: Company IR 

• The company’s 2025 gross profits grew by 45.2% YoY to $587.4 million. While adjusted gross profits grew by 44.9% YoY to $612.44 million with an adjusted gross margin of 30.3%, an improvement of 1.6 percentage points YoY. The strong gross margins were primarily due to product cost improvements.  
• FY25 GAAP operating margin expanded 2 points to 3.6%, remaining quite thin, while adjusted operating margin expanded 3.6 points to 10.9%, ahead of guidance for 8.6%.  
• Looking ahead, the company’s margins are expected to improve in 2026. Management guided adjusted gross margin to be 32%, an improvement of 1.7 points YoY. While the adjusted operating margin is expected to improve 3.2 points to 14.1% primarily due to operating leverage. 

2026 Adjusted EPS Guided to Increase 85% 

Bloom reported GAAP EPS of $0.00 in the quarter, though adjusted EPS saw a large 50% beat, coming in at $0.45 versus the $0.30 estimate. Analysts expect strong adjusted EPS growth in the coming quarters, with Q1 growth of 314.2% YoY to $0.12 and 153% YoY to $0.25 in Q2. 

Bar chart titled ‘Non‑GAAP EPS’ showing Bloom Energy’s quarterly adjusted EPS from Q3 2024 to Q4 2025, with values ranging from –$0.17 to $0.45. Q4 2025 adjusted EPS came at $0.45, beating estimates by 50% 

Source: Company IR 

For FY25, GAAP EPS was ($0.37), widening from ($0.13), while adjusted EPS was $0.76, increasing 171.4% YoY. For FY26, Bloom guided for adjusted EPS to be $1.33-$1.48, up 84.9% YoY at the midpoint and beating estimates by 25.5%. 

Turning to adjusted EBITDA, Bloom reported $146.1 million in Q4 for an 18.8% margin, down 6.9 points YoY but up 7.4 points QoQ. FY25 adjusted EBITDA was $271.6 million for a 10.9% margin, up 3.6 points YoY. 

Cash Flows and Balance Sheet 

Q4 is seasonally Bloom’s largest quarter for cash flows, with operating and free cash flow margins in excess of 50% this quarter, though this was much lower than the >80% margins it reported in Q4 2024. However, these large margins simply offset weak cash flows in the rest of the year, with full-year margins in the single-digit range.  

Operating cash flow was $418.1 million in Q4 for a 53.8% margin, down from an 84.6% margin in the year ago quarter. FY25 operating cash flow was $113.9 million for a 5.6% margin, down 0.6 points YoY. Bloom is guiding for operating cash flows to be ~$200 million in FY26, representing a ~6.3% margin at midpoint.  

Free cash flow was $395.1 million in Q4 for a 50.8% margin, down from an 82.7% margin in the year ago quarter. FY25 free cash flow was $57.2 million for a 2.8% margin, expanding 0.5 points YoY. 

Bloom reported $2.45 billion in cash and debt of $2.61 billion compared to $595.1 million and $1.13 billion in the previous quarter as Bloom raised $2.5 billion in convertible notes while also paying $975 million in existing debt in the quarter. 

Valuation 

Bloom Energy is trading at a P/S ratio of 17.1 and a forward P/S ratio of 12.7. The strong AI-demand led the company’s stock outperformance of 291.2% in 2025, and YTD return of 61%. As a result, Bloom Energy is trading at a premium to its average forward P/S ratio of 4.6. The company’s forward P/S ratio peaked at 17.7 on November 03, 2025, and is now 28% below that level. On the bottom line, the company is trading at a forward P/E ratio of 103.8.  

Chart showing Bloom Energy Corp. (BE) forward price‑to‑sales ratio from March 2025 to February 2026, rising sharply through late 2025 and currently trading at 12.70. 

Source: YChartsYCharts

Conclusion 

We are in the era of “what you see is what you get” – meaning, those offering strong earnings reports right now are setting up for a strong runway as future generations of AI accelerators will only be more power hungry.  

Bloom Energy has a compelling story, and after two decades, it appears like the stars are aligning for this alternative energy company. There is far less speculation today than at the start of 2025, when we first made our three initial buys, with two on the April 4th low at $17.04 and $16.64 for a 5% position. We then added another 3% on July 24th at $32.93, taking tactical gains in September as Bloom’s strong relative performance had made it one of the I/O Fund’s largest allocations at up to 15%. Bloom ended the year as one of the I/O Fund’s best performing stocks with an average return of 305%, with one entry returning 422%. The company’s customer base has expanded to six major accounts; margins are improving, and utilization of Bloom’s SOFCs has increased due to ongoing product enhancements all leading to strong price appreciation. 

Whether with Bloom Energy or the many other lesser-known AI stocks that my company has successfully identified early in their cycle, the test for investors will be figuring out how to hold-on while this market unfolds in the coming quarters (and years). 

We aim to offer support at every stage — from identifying products and solutions early in their cycle, to examining financials for confirmation that companies are executing, to breaking down technicals in our weekly webinars for a disciplined approach during market highs and lows. 

Since our inception in May 2020, I/O Fund has delivered a cumulative return of 326%— if we were a hedge fund, we’d rank #1 and if we were a tech ETF or Mutual Fund, we’d rank #3 in the United States. 326%— if we were a hedge fund, we’d rank #1 and if we were a tech ETF or Mutual Fund, we’d rank #3 in the United States.  

Being early to Bloom Energy and many lesser-known AI winners helped us to achieve these results. To get our Top 15 AI stocks, real-time trade alerts, weekly webinars and deep-dive research from a proven team in AI and tech stocks, Sign up now.Top 15 AI stocks, real-time trade alerts, weekly webinars and deep-dive research from a proven team in AI and tech stocks, Sign up now.

Royston Roche, Equity Analyst at I/O Fund contributed to this analysis.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

• The Future of AI Stocks? TSMC Commentary Suggests AI Megatrend
• S&P 500 Outlook 2026: Rising Volatility Risk and Key Support Levels
• Bitcoin After the Cycle Peak: What Comes Next and How We’re Positioning
• I/O Fund Jumps to 326% Cumulative Return, Ranking Among Wall Street’s Best

In 2025, the I/O Fund posted a 37% return, bringing its cumulative returns since inception to 326%, outperforming the broader markets and leading institutional tech portfolios by as much as 294%. On an annualized basis, the I/O Fund has averaged returns of 29.2%, outpacing many of Wall Street’s most renowned firms.  

• The I/O Fund’s 2025 return of 37% outperformed the Nasdaq-100 by 17% and the S&P 500 by 21%.  
• Since its May 2020 inception, the fund has achieved a 326% cumulative return, surpassing the Nasdaq-100 by 152% and the S&P 500 by 192%.  
• Since inception, the I/O Fund has maintained a notable 294% lead over other institutional technology portfolios.

An investment of $10,000 with the I/O Fund's picks at inception versus other all-tech portfolios would see a portfolio value of $42,552 with IOF versus $13,192 with institutional tech-focused portfolios. The difference in value is 223%.

I/O Fund Ranks Among Top Tech ETFs and Mutual Funds

On a cumulative basis, the I/O Fund’s stellar performance ranks among the top performing Tech ETFs and Mutual Funds, with the 326% cumulative return hypothetically placing the I/O Fund at #3.

While the I/O Fund’s portfolio was not exclusively invested in semiconductors, similar to many of the top ETFs on the chart below, we held high allocations in a select group of semiconductor winners, which played a significant role in driving our cumulative performance. This extended well beyond Nvidia to include several lesser-known AI semiconductor names.

I/O Fund Reports 56% Equity-Only Returns in 2025 

The I/O Fund’s 37% portfolio return combines both equities and cryptocurrencies, with the latter weighing on results with Bitcoin down (5%) and altcoins down as much as (40%). When excluding cryptocurrencies, the team’s equity strategy delivered a staggering 56% return, a result that ranks the I/O Fund team among the highest-performing investment teams in the United States on both an annual and cumulative basis. 

• The I/O Fund’s equity-only portfolio return of 56% would rank as the third highest recorded for any U.S. equity-only portfolio in 2025.
• The I/O Fund had 11 positions beat the Nasdaq-100 in 2025, up from 10 positions in 2024 and up from 7 positions in 2023.

The returns shown above reflect the performance of our equity strategy in 2025 and provide a more appropriate comparison to ETF thematic approaches. I/O Fund equities are concentrated entirely in AI stocks. 

I/O Fund’s 29.2% Annualized Return Ranks Among Leading Hedge Funds

The I/O Fund actively manages risk through hedging and raising cash. Therefore, the closest comparison in terms of style would be hedge funds. The I/O Fund’s annualized returns of 29.2% since inception compare favorably with some of Wall Street’s most established investment teams. 

The I/O Fund launched mid-year in 2020 on May 9th. The comparison below illustrates the strength of our annualized return over the past five years, outpacing several well-known hedge funds, including Pershing Square and Coatue.

What Drove the I/O Fund’s Outperformance in 2025

The last few years have demanded a high level of precision as the markets quickly and sharply rotated from cloud towards AI, with valuations resetting, and major indices advancing to consecutive strong years – conditions that have made sustained outperformance exceptionally difficult for active managers.

The I/O Fund strives for breadth and consistency by measuring how many positions outpaced the Nasdaq-100. Our results in 2025 exceeded previous results, notably this was on a high base from 2024.

• The I/O Fund had 11 positions beat the Nasdaq-100 in 2025, up from 10 positions in 2024 and up from 7 positions in 2023.

Additionally, the I/O Fund’s biggest winners were built through strategic entries that occurred below the January 1 opening price as tactical purchases through March and April allowed us to realize gains that exceeded the stock’s annual performance. Although we reserve the complete list for our paid members along with real-time alerts for every trade we make, below are a few highlights we can share with you: 

• Nvidia remained a key anchor in the I/O Fund’s portfolio, though we actively managed the position with purchases in March and April in alignment with our publicly-stated buy target zones.  
• We tactically reduced Bitcoin from a 10% allocation to 1% through April to December 2025, locking in gains from numerous purchases made through 2023 and 2024. 
• Bloom Energy became the I/O Fund’s top performing stock of 2025 with a 305% average return, with one entry as high as 422%. 
• An AI semiconductor stock became a leading allocation in the I/O Fund’s portfolio, with a 140% average return across all entries, driven by five tactical buys through February to April 2025. 
• We entered and exited a Bitcoin miner for a 93% average gain. 
• We captured a 79% return on a lesser-known AI data and software platform, with one entry as high as 94%. 
• We realized a 51% return on a leading AI advertising and social media firm.

Nvidia Stock Positioning  

Leading up to the release of the Hopper GPUs, we were net buyers of Nvidia in 2021 through early 2023. On average, it was held as a 15% position throughout 2022 and 2023. As we moved into 2024, Nvidia was allowed to exceed this allocation to become our first ever 20% position.

However, in mid-2024, we shifted our focus by looking more deeply at suppliers for Nvidia. Although the stock remained a core holding, we used our oversized allocation to raise cash at periods of perceived risk. Most notably, we talked about Nvidia hitting the $90 – $80 region for months, including in two free newsletters in January 2025, Where I Plan to Buy Nvidia Stock Next, and DeepSeek Creates Buying Opportunity for Nvidia Stock.

In February of 2025, we cut half of our position with an average cost basis of $130.88. We were then able to grab shares of Nvidia roughly 30% lower at $94.48 on April 4th and again at $87.99 on April 7th. The low for the year was $86.60.

Bitcoin Allocation Strategy  

While we were exceedingly bullish in Bitcoin in 2023 and early 2024, buying most dips, we lean heavily into technical analysis to manage our position. Using our technical analysis and internal indicators and other liquidity signals, such as those outlined in the free newsletter from August, Is Bitcoin’s Bull Run Nearing a Top? What the Herd Missed at $16,000 and is Missing Now and subsequent free webinar, we began trimming our position.

From April through December 2025, we reduced our Bitcoin position from 10% down to a 1% allocation, locking in gains between $85,000 and $113,000, making the multi-year bull run in Bitcoin one of the top actively managed positions in our firm’s history.

Bloom Energy as Top 2025 Winner

We first covered surging power demand from AI data centers in our June 2024 newsletter, AI Power Consumption: Rapidly Becoming Mission-Critical, with Bloom Energy quickly rising to the top of our list for its ability to solve the critical time-to-power constraint.

We made three initial buys in Bloom that defined our core position, with two on the April 4th low at $17.04 and $16.64 for a 5% position. We then added another 3% on July 24th at $32.93, taking tactical gains in September as Bloom’s strong relative performance had made it one of the I/O Fund’s largest allocations at up to 15%. Bloom ended the year as one of the I/O Fund’s best performing stocks with an average return of 305%, with one entry returning 422%.

AI Networking Stock Allocation

We built a leading 11% allocation to an AI networking stock from February to April after identifying its key positioning within the networking stack and its product positioning enabling larger and faster AI clusters to be architected.  

We trimmed 1/3 of this position in September to lock in an average 295% return from our initial four purchases and free up cash, shortly before the stock sold off 50% in just over two months.

Early to a Bitcoin Miner 

The I/O Fund first pitched this Bitcoin miner to its Discovery members, identifying it as one of the first to lead the transition to lucrative, high-revenue AI data center hosting with multi-billion dollar deals already secured. Shortly after, the I/O Fund brought this idea to its Advanced members as Portfolio Manager Knox Ridley believed he saw a potential positive setup emerging within his technical analysis on the chart.

mid

The I/O Fund then entered this Bitcoin miner with two purchases, both on April 4th, before exiting the stock in late July; we attempted to re-enter this position in October and November but closed the position for a small loss.

Overall, the I/O Fund realized an average return of 93% on this miner. 

Unlock our portfolio, real-time trade alerts on every position, live weekly webinars and more by signing up for Advanced Market Signals today. Learn more here.Learn more here.

What Sets the I/O Fund Apart 

At the heart of I/O Fund, we believe that transparency is key to our success over the last few years. We keep our members in the loop with real-time trade alerts and audited performance reviews, and we take pride in holding ourselves to the utmost highest standard when it comes to accountability as very few research platforms (if any) offer this. 

Why Real-Time Trade Alerts Matter 

Real-time trade alerts are sent to our premium members at the moment we decide to buy, sell, trim, or add to a position. That may sound straightforward, but in practice, it’s one of the most demanding ways to manage a portfolio, as each and every decision is recorded publicly the moment it’s made.

That level of transparency places meaningful pressure on the portfolio team, which is exactly the point. It’s the same standard registered fund managers are held to when they file trades, yet it’s rarely offered across investment research.  There’s a reason most research platforms avoid real-time alerts, active position management, and detailed transparency: the more granular the reporting, the higher the stakes. When decisions are logged instantly, there’s nowhere to hide. 

Every portfolio team makes mistakes. The difference is that by making mistakes publicly, it has sharpened our decision-making and strengthened our discipline over time. 

Verified Returns, Accountability: $210,000 Invested in Transparency

One of the biggest gaps across research platforms is the lack of verified performance. Institutional investors don’t take claims at face value — they require proof. Hedge funds are required to report returns precisely because it reduces posturing and selective storytelling.

We apply that same mindset here. To date, the I/O Fund has invested over $210,000 into accountability and transparency for members since inception.  

Early on, we used a forum-based system for trade alerts. By 2021, we transitioned to dedicated SMS and email infrastructure using Twilio and Mailchimp — tools designed to minimize outages and delays. This alone costs $30,000–$40,000 per year, depending on trading activity. 

In addition, we engage an independent accounting firm in San Francisco to mathematically review and verify performance across both our equity and crypto accounts. Each audit takes several months and costs $4,500 to $5,500. To date, we’ve completed seven audits, totaling $32,500.  

Accountability isn’t free, but it’s a standard we hold ourselves to. 

Conclusion: Consistency, Accountability, and Strong Portfolio Returns

Over the past year, we delivered 11 positions that outperformed the Nasdaq-100, continuing a multi-year trend of identifying winners early. Many of our biggest winners were built at prices below the January 1 opening levels, allowing us to realize returns that exceeded the stock’s annual performance. In addition, we drastically cut back our crypto positions to stave off losses starting in August, despite many crypto influencers calling for aggressive price targets. 

This consistency helped extend the portfolio’s cumulative return to 326% since inception in May 2020, with annualized returns averaging 29.2% over that period. Across thousands of portfolio options, these results place us firmly among the top-performing investment strategies in the United States.  

We take the responsibility of providing our Members early, actionable research tools just as seriously as the pursuit of the upside. Thank you for your continued support and confidence.

You can read our full press release here: I/O Fund Proves Leadership in AI Stocks with 326% Cumulative Return and 29.2% Annualized Return.You can read our full press release here: I/O Fund Proves Leadership in AI Stocks with 326% Cumulative Return and 29.2% Annualized Return.

These results have been independently audited by an accounting firm in San Francisco. Additional performance details and disclosures are available on the I/O Fund website for Premium members. 

Just as important as performance is access to timely, actionable tools. The I/O Fund’s Advanced Tier provides:

• Real-time trade alerts on entries, trims and exits
• Deep-dive research on lesser-known AI Stocks
• Weekly, one-hour webinars held Thursdays at 4:30 pm Eastern

Join today with a limited-time offer for $100 off our flagship tier. Sign up now.Join today with a limited-time offer for $100 off our flagship tier. Sign up now.

Please note, past results are not a guarantee of future outcomes. Reference our terms and conditions here.terms and conditions here.

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:

• Bitcoin After the Cycle Peak: What Comes Next and How We’re Positioning
• S&P 500 Outlook 2026: Rising Volatility Risk and Key Support Levels
• The Future of AI Stocks? TSMC Commentary Suggests AI Megatrend
• The $530 Billion AI Question: Which Big Tech Stock is Winning?

TSMC is one of the least sensational management teams in the AI stocks space, yet management explicitly called AI a multi-year “megatrend” in their most recent earnings call, with demand now being pulled not just by chip designers, but directly by hyperscale cloud providers seeking to lock in capacity.  

Management stated: 

“Our customers’ customers, who are mainly the cloud service providers, are also providing strong signals and reaching out directly to request the capacity to support their business. Thus, our conviction in the multiyear AI megatrend remains strong, and we believe the demand for semiconductor will continue to be very fundamental.”Thus, our conviction in the multiyear AI megatrend remains strong, and we believe the demand for semiconductor will continue to be very fundamental.” 

When the world’s most advanced foundry says hyperscalers are coming to them directly for capacity, it signals that AI demand remains foundational. Perhaps most importantly, TSM is not a “flip the switch" business model to where demand can be turned on and turned off quickly. Wafer capacity must be planned years in advance, which makes these signals particularly meaningful. 

TSMC already sits beneath tens of trillions of dollars in market capitalization, with customers including Apple, Nvidia, Broadcom, Amazon, AMD, and Google. While each is pursuing AI through a different mix of merchant GPUs, custom silicon, and software, they all converge at the same point: TSMC’s advanced manufacturing. As the roadmap progresses to N2 and A16, customer dependence on TSMC’s leading-edge capacity increases. 

The company reported record Q4 revenue of $33.73 billion, up 25.5% year over year and 1.9% sequentially, exceeding the midpoint of guidance by 2.8%. But the more important takeaway was not the quarter, rather it is the visibility that TSMC provides. 

Below, we break down how this multi-year AI megatrend translates into durable visibility for TSMC, expands pricing power for advanced nodes, and a longer runway for earnings growth on arguably one of the strongest AI stocks on the bottom-line in the space. 

TSMC Raises AI Accelerator Forecast to mid to high 50% CAGR: Evidence of a Multiyear Megatrend  

TSMC stands out as the most reliable barometer for tracking AI demand trends. During the Q4 earnings call, management raised the forecast for revenue growth for AI accelerators to a mid- to high 50% CAGR over the 5-year period from 2024 to 2029, up from the mid-40% CAGR provided during the Q4 2024 earnings call. This conveys strongly that AI demand is expanding.  

Similarly, the company’s long-term revenue growth forecast was raised to 25% CAGR in U.S. dollar terms for the 5-year period starting from 2024 from the earlier 20% CAGR provided during the Q4 2024 earnings call. Management expects AI accelerators to drive the largest share of incremental revenue growth, while overall growth will continue to be supported by smartphones, high-performance computing (HPC), IoT, and automotive over the next several years.

mid

AI accelerator revenue accounted for a high-teens percentage of total revenue in 2025, up from the mid-teens in 2024. As manufacturing complexity continues to rise, engagement lead times for advanced chips have extended to at least two to three years and thereby providing better long-term visibility.  

The company’s capex in 2025 grew by 37.4% YoY to $40.9 billion. Management expects 2026 capex in the range of $52 billion to $56 billion, implying a YoY growth of 32% at the midpoint. Notably, 70–80% of 2026 spending will be allocated to advanced process technologies (7nm and below), signaling management’s confidence in sustained, long-term demand driven by AI. TSMC takes two to three years to build a new fab with the increase in 2026 capex spending suggesting strong structural growth for the years 2028 and 2029. 

Responding to an analyst’s question on whether the industry is in an AI bubble, Chairman and CEO C.C. Wei gave a grounded answer. “Okay. Gokul, you essentially try to ask us, say, whether the AI demand is real or not. I'm also very nervous about it. You bet because we have to invest about USD 52 billion to USD 56 billion for the CapEx, right? If we didn't do it carefully, and that would be big disaster to TSMC for sure.” Management mentioned that they spent $101 billion in capex in the last three years and they expect capex to be significantly higher in the next three years.

TSMC’s capex is expected to grow 32% YoY to $54 billion in 2026, suggesting strong AI-demand in the next few years.  

Source: Company IR 

Advanced Nodes: Why TSMC’s Semiconductor Moat is Widening 

While 2nm defines the next phase of the roadmap, 3nm remains the node supporting most AI deployments today. 

The company’s advanced 3nm node offers roughly 15% better performance than 5nm at equal power and transistor density, with die sizes estimated to be ~42% smaller. TSMC also states the 3nm process can reduce power consumption by up to 30%, underscoring power efficiency as a key competitive advantage. 

This efficiency helps deepen TSMC’s moat. While Samsung introduced 3nm chips in 2022, it has lagged TSMC on yield and power efficiency by an estimated 10%–20%. This advantage is reflected in pricing power, with TSMC charging roughly 25% more for 3nm versus 5nm, as customers are willing to pay a premium to avoid Samsung. 

TSMC offers multiple 3nm variants, including N3E, N3P, and N3X, allowing customers to tailor designs across consumer and AI workloads. N3E serves as the baseline IP platform, delivering approximately 18% higher performance and 34% lower power versus N5. N3P provides incremental performance and efficiency gains, while N3X is optimized for high-performance computing at the expense of higher leakage. 

The company entered volume production of its most advanced node, N2, in 4Q 2025, marking a transition from FinFET to gate-all-around (GAA) transistor architecture. By wrapping the gate around all sides of the channel, GAA improves electrostatic control and reduces leakage versus FinFET designs. 

N2 introduces NanoFlex technology, enabling designers to mix cell types and optimize for performance or power by adjusting nanosheet dimensions. According to management on the Q2 2025 earnings call, N2 delivers 10%–15% speed improvement at the same power or 20%–30% power reduction at the same speed, along with more than 15% chip density gains versus N3E. 

TSMC expects strong demand from both smartphones and AI HPC applications, with a rapid ramp in 2026. The company also introduced N2P, an enhanced version of N2, with volume production scheduled for the second half of 2026. Additionally, A16, featuring super-power rail (SPR) technology for complex HPC designs, is also on track for volume production in the second half of 2026. 

As chips migrate to advanced nodes—such as Nvidia’s Rubin moving to 3nm and AMD building GPUs and CPUs on 2nm—TSMC stands to benefit from rising pricing power, as these nodes command significant wafer premiums in exchange for material performance and power efficiency gains. 

In Q4 2025, advanced nodes below 7nm accounted for 77% of wafer revenue, up from 74% in Q3. 3nm rose to 28% of wafer revenue, while 5nm accounted for 35%, compared with 23% and 37%, respectively, in the prior quarter. 

TSMC 3nm revenue has ramped up from 22% of wafer revenue in Q1 2025 to 28% in Q4 2025, signaling transition toward high-performance AI. 

Source: Company IR 

TSMC Stock: $122B Record Revenue on Surging AI Chip Demand 

TSMC reported record Q4 revenue of $33.73 billion, up 25.5% year over year and 1.9% quarter over quarter, beating the midpoint of guidance by 2.8%. As expected, the upside was driven primarily by strong AI-related demand. Revenue growth is expected to accelerate next quarter, with management guiding Q1 revenue of $34.6 billion to $35.8 billion, implying 37.9% year-over-year growth and 4.4% sequential growth. 

For the full year, 2025 revenue grew 35.9% year over year to a record $122.42 billion. Looking ahead, management guided 2026 revenue growth to be close to 30% year over year in U.S. dollars. 

HPC revenue continued to expand, rising 4% quarter over quarter in NT dollars and accounting for 55% of total revenue in Q4. For FY25, HPC revenue in NT dollars increased 48% year over year and represented 58% of total revenue.  The sequential deceleration in HPC revenue in Q3 appears consistent with a platform transition, as Nvidia prepares to shift from Blackwell—built on TSMC’s customized 5nm N4P process—to the upcoming Rubin architecture, which relies on TSMC’s customized 3nm N3P node. 

With Rubin chips expected to enter mass production in the second half of 2026, we will look for HPC revenue growth to reaccelerate in the coming quarters. Management also highlighted a strong ramp for N2 in 2026, reinforcing sustained AI-driven demand. 

As shown in the chart below, TSMC experienced a similar slowdown in HPC growth during the second half of 2022 and the first half of 2023 amid the transition to Nvidia’s Hopper architecture. Nvidia began mass production of Grace Hopper chips in May 2023, after which HPC revenue growth resumed.

TSMC High-performance Computing (HPC) revenue grew by 4% QoQ in Q4 2025, up from flat sequentially in Q3 2025 

Source: Company IR 

TSMC Q4 Results: GAAP EPS grew by 40% 

TSMC’s ability to generate exceptionally strong profits showcases that the company is one of the best-managed companies in the world. Despite the rising inflation, tariff concerns, technological advancement, trade wars, overseas fab expansion, and geopolitical tensions, TSMC has overcome these challenges by continuing to generate superior profits. Margins continue to expand due to cost controls, higher capacity utilization rates, economies of scale, and better price negotiation with customers and suppliers.   

TSMC’s Q4 GAAP EPS grew by 40.2% YoY, beating estimates by 5.2%. That strength is expected to continue into 2026. Earnings are projected to grow roughly 55% year over year in Q1, followed by another strong step-up in Q2 with 40.1% growth. After delivering more than 50% EPS growth in 2025, consensus estimates call for earnings to rise by more than 20% annually in both 2026 and 2027, pointing to a sustained growth profile rather than a one-year spike.

TSM’s Q4 2025 EPS grew by 40.2% YoY to $3.14, beating estimates by 5.2%. These results highlight the company’s dominant position in the AI chip supply chain and growing demand for high-performance computing (HPC). 

Source: Company IR/YCharts 

In Q4, TSMC generated $23.4 billion in operating cash flow, up 21.9% year over year, with an operating cash flow margin of 69.4%, down modestly from 71.4% in the prior-year period. 

Free cash flow increased 48.8% year over year to $11.9 billion, with margins expanding nicely to 35.2% up from 29.8% last year. Notably, capital expenditures grew just 2.5% year over year to $11.5 billion in the quarter, supporting strong free cash flow conversion. Management has indicated this moderation is temporary, as noted above, capex is expected to accelerate meaningfully in the coming years as advanced-node demand ramps. 

Conclusion 

When the world’s most advanced foundry raises its long-term AI growth forecasts, commits more than $50 billion annually to capacity, and openly states that hyperscalers are coming directly to secure wafers years in advance, it provides a level of visibility that few companies in the AI ecosystem can match. TSMC’s willingness to materially expand capex through 2026, knowing that fabs take years to come online, signals confidence that extends well beyond the next product cycle and into the latter half of the decade. 

We’ll expand on how this visibility feeds into the I/O Fund's broader AI positioning—and where we see the next phase of opportunity—in our Q1 2026 Top 15 AI Stocks report. This 50+ page report totalling over 20,000 words identifies many lesser-known AI stocks leading the charge in the three critical pillars of the AI Megatrend: AI networking, AI energy and AI inference. Sign up now to receive this report.

Royston Roche and Damien Robbins, Equity Analysts 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:

• The $530 Billion AI Question: Which Big Tech Stock is Winning?
• Palantir Stock 2026 Forecast: Is Its High Valuation Sustainable?
• Top 10 Tech Stocks of 2025: How the AI Trade Defied the Skeptics
• Nvidia & Beyond: I/O Fund’s Best Free AI Stock Research in 2025

Big Tech is expected to invest $530 billion for building AI infrastructure in 2026, while the path to near-term monetization remains a question mark. As investor scrutiny around capital expenditure intensifies, the key question is no longer who is spending the most on AI, but who is translating that spend into measurable revenue and sustainable margins. 

The market is right to question the shift in role for Big Tech, as these companies have swiftly moved from global demand engines for consumer and enterprise software to now being large consumers themselves of a rather expensive endeavour.  

When asked about Big Tech for the last few years, my answer has been consistent – my firm is heavily positioned in the suppliers. For example, while many are celebrating Google’s stock performance last year, stocks like AMD and Micron effortlessly outperformed the 2025 Big Tech winner – as did dozens of lesser-known suppliers. This is easily visible in the QQQs returning 20.2% last year compared to SMH at 48.7%.  

But I digress, as although Big Tech has paled in stock returns compared to select AI semiconductor stocks, my firm is anticipating an incoming monetization wave driven by inference. The moment that Big Tech and others can effectively monetize their large investments will unfold in the coming years. Big Tech is not muted; it’s just a bit later in the cycle given they’re predominately software companies.  

My firm strives to be early to trends, positioning in cloud in 2019 ahead of Covid, then began rotating out of cloud to position for AI in late 2022 around extreme lows and ahead of Nvidia’s Hopper release. We then added AI energy names to our portfolio ahead of the Street in February – April of 2025 (again) before the broad market and around extreme lows.

On a similar note, I fully expect we will be well-positioned for the moment AI can finally monetize – which brings us back to Big Tech.  

The report below compares Big Tech companies across revenue, AI segment revenue, margins, capex intensity, and balance sheet health, to help assess which companies are best positioned to convert AI investment into durable growth and profits. With infrastructure spending at historic highs, the stakes have never been higher. 

Microsoft ($MSFT): Azure Acceleration Powered by AI  

Microsoft has consistently delivered double-digit revenue growth across all three quarters of 2025, driven by strong demand for cloud and AI.  

• The company’s Q3 revenue grew by 18.4% YoY to $77.67 billion.  
• Analysts expect revenue to grow 15.3% YoY to $80.3 billion in Q4, representing more than 300 bps deceleration due to seasonality and the revenue decelerated 370 bps in the same period last year. 

Yet under the hood, Microsoft Azure is witnessing accelerating demand, rising 40% in Q3, driven by better-than-expected growth in the company’s core infrastructure business, primarily from its largest customers and AI-demand. The company’s productivity and business processes segment has also demonstrated steady, balanced growth driven by continued adoption of Microsoft 365.  

Microsoft’s revenue growth has accelerated from 13.3% in Q1 to 18.1% in Q2, and a further acceleration of 30 basis points in Q3 to 18.4% YoY. 

Source: YChartsYCharts

The last direct update on AI revenue we got from Microsoft is from fiscal Q2 ending December 2024, where AI revenue was stated to be $13 billion, growing at a pace of 175% year-over-year. Microsoft did denote in fiscal Q3 that AI services contributed 16 points to Azure’s 35% YoY growth, up from 13 points in the prior quarter.  

Assuming AI drove Azure’s reacceleration from 35% to 39% in the past two quarters, the base case for Microsoft assumes AI contributing approximately 22 points to growth as of the prior quarter, fiscal Q1, or around 56% of its YoY growth in dollars. This could imply AI revenue at 26% of Azure’s total revenue, or around $25 to $26 billion on a nearly $100 billion annual run rate for Azure. Microsoft has also struck a new deal with OpenAI under which it has committed to $250 billion of compute capacity through 2032, and also signed a deal with Anthropic for $30 billion of compute capacity.

The company has consistently delivered strong profits, driven by higher software and cloud services revenue, and operational efficiencies. The company’s Q3 operating margin improved by 400 basis points sequentially and 230 basis points YoY to 48.9%, driven by operating leverage and higher-margin revenue. 

Microsoft’s Q3 operating margin has improved YoY and sequentially. 

Source: YChartsYCharts

Microsoft’s capex in Q3 was $34.9 billion, an increase of 75% YoY from $20 billion in the year-ago quarter. The company’s strong capex growth was primarily driven by increasing demand for its Cloud and AI offerings. With strong accelerating demand, Microsoft is increasing its spending on GPUs and CPUs. Therefore, total spending is expected to increase sequentially in the next quarter. The company maintains a strong balance sheet with a net cash position of $58.8 billion. 

Takeaway: Given Microsoft’s large capex spending, the market is likely to be unconvinced in the near-term the AI contribution to Azure is happening quickly enough. If our math is correct, AI’s contribution is less than 1/5 of this year’s capex and we are years deep now – meaning its much lower in terms of contribution to cumulative AI capex. 

Keep in mind, Microsoft has publicly stated they will absorb grid costs for their AI infrastructure to avoid passing it off to consumers – which is ethically the right thing to do – yet this increases total capex cost per GW. Recently, Microsoft President Brad Smith stated: “We will pay utility rates that are high enough to cover our electricity costs in part by collaborating with utilities on plans to add the electricity supply that we will need.” It’s likely other Big Tech companies will follow suit. 

Google ($GOOGL): Record 83% Capex Surge to Scale Search and Google Cloud Momentum 

Google’s Q3 revenue grew by 16.2% YoY to $102.35 billion. Analysts expect Q4 revenue to grow 15.4% YoY to $111.3 billion. The company is witnessing momentum across its business segments with Google Cloud being a standout, with accelerating revenue growth from 28% in Q1 to 32% in Q2, and further accelerating to 34% YoY growth to $15.2 billion. This has been primarily driven by Google Cloud Platform, Generative AI Solutions, and AI Infrastructure.  

The Cloud business continues its trajectory of accelerated growth, with AI-driven revenue emerging as a key catalyst. Cloud backlog increased 46% QoQ to $155 billion, underscoring strong demand visibility and sustained momentum across enterprise deployments. Most importantly, over 70% of existing Google Cloud customers use the company’s AI products.

Google’s revenue growth has accelerated from 11.8% in Q1 to 14.1% in Q2, and a further 210 basis points acceleration in Q3 to 16.2% YoY.

Source: YChartsYCharts

Google said that its enterprise AI products in GCP are generating ‘billions’ in quarterly revenue but provided no specific breakout beyond that, and also did not provide an update on Search. However, assuming a similar split as Microsoft for AI contribution in the latest quarter, such as ~17 points or half of growth, this would project quarterly AI revenue to be ~$1.9 billion, or nearly $8 billion annualized, less than one-third of Microsoft’s current run rate. Google signed a deal with Anthropic in October said to be worth tens of billions, giving the Claude parent access to up to one million TPUs, bringing more than 1GW of capacity online in 2026.

mid

AI’s contribution to Search revenue could be larger, assuming that the acceleration from 10% YoY in Q1 to 15% YoY in Q3 is entirely driven by AI and AI Overviews, or 5 points. Search growth was ~12-13% in the second half of 2024 after the launch of AI Overviews and assuming ~2 point contribution from AI by Q4, this would roughly estimate AI’s contribution to Search at $800 million. As of Q3, under that 5 point assumption, AI’s contribution would be estimated at roughly $2.3-2.4 billion, or closing in on a $10 billion annual run rate.

Google’s Q3 operating margin improved 180 basis points YoY and 160 basis points QoQ to 34.1%, driven by operational efficiencies. Google’s high-margin advertising business remains the core profit engine, delivering durable and resilient earnings growth.

Google's Q3 operating margin has improved YoY and sequentially. 

Source: YChartsYCharts

Google reported the fastest capex growth among Big Tech in Q3. The company’s capex grew by 83% YoY to $23.95 billion. Sequentially, it grew by 7% from $22.4 billion in the previous quarter. Management stated in the Q3 earnings call that they are witnessing positive returns on AI investments. “I would say it's not just early signs because we're seeing returns, obviously, in the Cloud business. You've heard us talk about the fact that we already are generating billions of dollars from AI in the quarter.”  

Looking ahead, the company expects to invest aggressively due to strong demand from cloud customers and growth opportunities across the company. Management expects 2025 capex to be in the range of $91 billion to $93 billion, up from the previous estimate of $85 billion. It represents YoY growth of 75% at the midpoint. Capex is expected to increase further in 2026, which supports our view that AI stocks will benefit. Google’s balance sheet remains robust, with a net cash balance of $76.9 billion, the highest among Big Tech companies.

Takeaway: 

Google reported the strongest returns last year across Big Tech, likely due to a combination of Google Cloud accelerating alongside an easier monetization with Search given advertising can see an immediate impact from automation. The company repeatedly emphasizes that integrating AI into Search improves advertiser outcomes for more effective campaigns, therefore, investors should keep an eye on the Search inflection as much (if not more so) than Google Cloud. 

In April 2025, Google introduced Ironwood v7, its first TPU designed specifically for inference. Another impetus for 2025’s strong performance is using its own silicon to drive inference at a lower cost curve, which will result in improved unit economics for Search and GCP. We discussed in detail the Google TPU Ironwood v7 rollout in our recent report to our free newsletter subscribers.

Amazon ($AMZN): AWS Reported the Fastest Growth since Q4 2022 

Amazon has the second slowest revenue growth among the Big Tech companies. The company’s Q3 revenue grew by 13.4% YoY to $180.17 billion. Revenue is expected to grow by 12.4% YoY to $211.14 billion in Q4. Q3 AWS revenue grew by 20.2% YoY to $33 billion, accelerating by 270 basis points from 17.5% growth in Q2. The company reported the fastest AWS growth since Q4 2022, driven by strong demand in AI and core infrastructure.   

Amazon’s revenue growth has accelerated from 8.6% in Q1 to 13.3% in Q2, then in Q3 to 13.4% YoY.  

Source: YCharts

Amazon has provided a few hints on its AI revenue, saying in Q2 that AI was a “fast-growing triple-digit year-over-year percentage multibillion-dollar business” and in Q3 that its Trainium chips had grown 150% QoQ and were a multi-billion dollar business. Again, assuming AI contributed roughly half or slightly above half of AWS’ 20% YoY growth, or 10 to 12 points, this would project AI revenue to be ~$2.8-3.3 billion, or more than $10-12 billion annualized. This would represent 8-10% of AWS’ TTM revenue.  Amazon also signed a $38 billion, multi-year deal with OpenAI in November to give the AI firm immediate access to Nvidia GPUs on AWS.

Amazon has the weakest margins among Big Tech companies, as the majority of its revenue is generated by its lower-margin e-commerce business. In Q3, operating margin improved modestly by 20 basis points YoY but declined 30 basis points sequentially to 11.3%.

Amazon’s Q3 operating margin has improved marginally YoY and down sequentially. 

Amazon is the biggest spender among the Big Tech companies. Amazon’s capex in Q3 rose 55% YoY to $35.1 billion, with the company raising the 2025 capex guidance to $125 billion, up 51% YoY. The company’s CEO, Andy Jassy, said in the Q3 earnings call, “You're going to see us continue to be very aggressive investing in capacity because we see the demand. As fast as we're adding capacity right now, we're monetizing it.” The company has a net cash position of $43.5 billion.  

Takeaway:  

AWS enjoyed an early mover advantage to cloud infrastructure-as-a-service (IaaS) yet that advantage could be in jeopardy in the AI era. These concerns were alleviated somewhat in the last earnings report as AWS revenue reported 20% YoY growth for $33 billion in Q3, marking the fastest growth since Q4 2022. Amazon also struck a seven-year cloud computing deal with OpenAI, signaling renewed momentum. 

Similar to Google, AWS is building out its custom Trainium chip and expanded its AI-focused programs, including accelerators for startups and new cloud services that help enterprises adopt generative AI. The custom chips offer 4X performance and efficiency gains for better price-performance compared to GPUs. This helps to lower AWS’ compute costs and will lead to better margins given the more competitive pricing for training and inference at scale. 

Meta ($META): The $60B Advantage+ Engine 

Meta has the fastest growth among the Big Tech companies and is an early AI beneficiary. The company is benefiting from AI-driven recommendation models that are improving advertiser ROI and increasing time spent across its family of apps, supporting stronger advertising revenue growth. The company’s Q3 advertising revenue grew by 25.6% YoY to $50.1 billion, accelerating more than nine points since Q1 and marking the fastest growth in six quarters. 

Meta’s revenue growth has accelerated from 16.1% in Q1 to 21.6% in Q2, and to 26.2% YoY in Q3.  

Source: YChartsYCharts

In terms of AI-driven revenue, Meta’s Advantage+ is outpacing OpenAI by 3X and is also offering the strongest AI revenue among the FAAMGs – in Q3, the company revealed that its end-to-end AI-powered ads platform Advantage+ has reached a $60 billion annual run rate, just 3.5 years after its launch. We recently covered this in more detail in our free newsletter, The AI Revenue Leader Nobody Is Talking About—Second Only to Nvidia Stock. 

Meta is witnessing a deceleration in margins due to rising expenses supporting its AI infrastructure. Reality Labs also continues to incur losses, recording a $4.43 billion loss from operations in Q3 2025, and its cumulative losses now total $73.04 billion. During Q3 earnings, management mentioned that total expenses “will grow at a significantly faster percentage rate in 2026” driven by infrastructure costs, incremental cloud costs and depreciation, followed by employee compensation. In December, Meta’s CEO, Mark Zuckerberg, said that they plan to cut Metaverse projects by up to 30%, which is positive.

Meta’s Q3 operating margin was down YoY and sequentially.

Meta’s Q3 capex grew by 111% YoY to $19.4 billion. The strong growth was primarily driven by investments in servers, data centers, and network infrastructure. Management also increased the 2025 capex to a range of $70 billion to $72 billion, up from the prior outlook of $66 billion to $72 billion. It represents a YoY growth of 81% from the prior year. Due to continued investments in AI infrastructure, Meta expects next year’s capex to be significantly higher than in 2025, particularly as their compute needs are higher than their expectations. Management stated in the earnings call, “As we have begun to plan for next year, it's become clear that our compute needs have continued to expand meaningfully, including versus our own expectations last quarter. We are still working through our capacity plans for next year, but we expect to invest aggressively to meet these needs, both by building our own infrastructure and contracting with third-party cloud providers.” 

Meta has the weakest balance sheet among the Big Tech companies with a net cash position of $15.7 billion. Meta has also entered a joint venture with Blue Owl Capital to fund its development of the Hyperion data center in Louisiana. Thereby, helping it to keep about $27 billion in debt off-balance sheet, where it would sit in a special-purpose vehicle tied to Blue Owl. While this approach may improve reported leverage and financial ratios, it carries inherent risks.  

Takeaway: 

Which company has the world’s most valuable data? While some will argue it’s a tie between Google, Meta, Microsoft and Amazon – I believe the contextual richness of social media data puts Meta at a slight advantage. In addition, Advantage+ is the second largest AI revenue segment on the market today; a stat this is materially underreported in the media. 

However, the biggest hurdle Meta faces is not visibly seen in the fundamentals but rather the company must convince developers to adopt its large language models, which are underwhelming on benchmarks compared to OpenAI and Anthropic's Claude. From where it stands today, it’s hard to see a path where Meta commercializes Llama. 

Instead, in the near-term, Meta’s opportunity lies in its ability to monetize its own platforms. This approach has proven to be unusually effective, given the revenue we are seeing from Advantage+. Over time, however, Meta will need to bring additional AI-driven applications to market to keep pace with competitors that are beginning to monetize not only infrastructure, but the full software stack built on top of it. 

Apple ($AAPL): Record Services Revenue 

Apple has the slowest revenue growth among Big Tech companies. The company’s Q3 revenue grew by 7.9% YoY to $102.5 billion. Revenue is expected to accelerate to 11.4% YoY to $138.5 billion in Q4. The services revenue achieved a record of $28.8 billion in Q3, growing 15.1% YoY.

Apple’s revenue growth has accelerated from 5.1% in Q1 to 9.6% in Q2 and decelerated to 7.9% YoY in Q3. 

Source: YChartsYCharts

The company’s Q3 operating margin improved 50 basis points YoY and 170 basis points sequentially to 31.7%. The improvement in margins was primarily driven by favorable product mix, cost discipline, and a higher contribution of margin from better-margin services revenue.

Apple’s Q3 operating margin was up YoY and sequentially. Source: YCharts 

Takeaway: 

Apple is not an infrastructure player, and thus given we are in the infrastructure phase of AI, one of the world’s most valuable companies is largely sitting this one out. While Apple benefits from exceptional margins, a fortress balance sheet, and strong cash generation, those strengths are not translating into leadership in the current phase of the AI training market. 

Our firm is focused on identifying AI’s biggest winners, and from that perspective, the opportunity cost of owning Apple over companies supplying AI infrastructure, energy, and on the brink of monetizing LLMs is significant. Apple may serve as a defensive allocation, but in my opinion, it does not offer the same asymmetric upside as the companies building—and monetizing—the backbone of the AI economy. 

This may change once more AI applications are in the hands of consumers, but for now, Apple is not part of our AI investment focus. 

Conclusion 

Based on this analysis, Google stands out as having the strongest Big Tech AI positioning today, given its ability to monetize AI directly through Search while simultaneously accelerating Cloud growth. To add to this, Google will see improved unit economics with custom silicon with a path to expand margins despite elevated capex. 

At the same time, we continue to see Big Tech broadly underperform AI hardware suppliers – even Google. For example, while many are celebrating Google’s stock performance last year, stocks like AMD and Micron effortlessly outperformed the 2025 Big Tech winner – as did dozens of lesser-known suppliers. This is easily visible in the QQQs returning 20.2% last year compared to SMH at 48.7%.  

In our view, the most compelling value creation in today’s AI market remains on the supply side, particularly among underappreciated hardware and infrastructure players. Rather than Big Tech, I/O Fund is focused on where AI demand is more broadly flowing, which companies are translating spend into profits, and how to position ahead of the next leadership shift. Viewed through a longer-cycle lens—especially around when monetization becomes meaningful—the expectation that Big Tech will lead as early as the first half of 2026 appears premature relative to other, more attractive areas of the market.

Next Wednesday, I’m dropping my Top 15 List of AI Stocks for Q1 2026 – this quarterly report ranks the companies I believe will define the next year and whose fundamentals are on fire. Notably, Big Tech did not rank high enough to make the list. The 42-page report is for I/O Fund premium members. Sign up now.Top 15 List of AI Stocks for Q1 2026 – this quarterly report ranks the companies I believe will define the next year and whose fundamentals are on fire. Notably, Big Tech did not rank high enough to make the list. The 42-page report is for I/O Fund premium members. Sign up now.

Royston Roche and Damien Robbins, Equity Analysts 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:

• Palantir Stock 2026 Forecast: Is Its High Valuation Sustainable?
• Top 10 Tech Stocks of 2025: How the AI Trade Defied the Skeptics
• Nvidia & Beyond: I/O Fund’s Best Free AI Stock Research in 2025
• AI Stocks & Nvidia: I/O Fund’s 2025 Tech Media Highlights