Microsoft’s 10-K Finally Reveals OpenAI Revenue—and It Changes the AI Growth Story
In its 10-K annual regulatory filing with the U.S. Securities and Exchange Commission, Microsoft has for the first time clearly disclosed the revenue scale tied to OpenAI. For the fiscal year ended June 2026, the figure reached $24.1 billion. Notably, this number did not come from Microsoft’s Q4 earnings release on July 29. Instead, it appeared in regulatory materials filed during the same period, marking Microsoft’s first public disclosure of commercial operating revenue derived from OpenAI.
Until now, the market had only two public reference points for the overall size of Microsoft’s AI business. One was a quarter ended December 2024, when Microsoft said its AI business was on track to approach $13 billion in annualized sales. The other was a quarter ended March 2026, when CEO Satya Nadella said annualized AI revenue was moving toward more than $37 billion. However, $37 billion represents an annual run rate, not a confirmed full-year revenue number. Based on Bloomberg’s estimates using the 123% year-over-year growth disclosed in Microsoft’s March-quarter report, the analysis suggested that Microsoft’s actual AI revenue for the fiscal year ended June would be about $34 billion. Against that baseline, OpenAI’s $24.1 billion would account for more than half of Microsoft’s total AI business revenue, potentially as high as ~70%.
This revenue mix also implies that Microsoft’s AI growth narrative relies heavily on a single partner’s compute consumption and revenue sharing. A Microsoft spokesperson confirmed that the AI total revenue figure above already includes both all OpenAI-related sales and the associated revenue-share amounts.
Why AI Growth Collapses Once OpenAI Is Removed
Once OpenAI’s contribution is stripped out, the "quality" of Microsoft’s AI growth looks radically different. The data shows that if you exclude OpenAI revenue, the year-over-year growth rate of Microsoft’s AI-related commercial bookings (RPO) would fall sharply from about 84% to 25%.
This gap points to two core issues. First, Microsoft’s high-growth AI performance is highly concentrated in OpenAI. The fees OpenAI pays Microsoft mainly fall into three buckets: charges for Azure cloud computing and inference compute, costs related to AI model development, and a certain percentage of revenue sharing. All three components are directly tied to OpenAI’s operational scale and commercialization pace, rather than reflecting broad penetration of enterprise AI products.
Second, Microsoft’s own AI product commercialization has not yet generated enough scale effects. While paid seats for Microsoft 365 Copilot have surpassed 30 million and net new seats have more than doubled quarter over quarter, Copilot and broader enterprise AI service revenue still trails meaningfully compared with the ~$24.1 billion magnitude contributed by OpenAI. Growth dropping from 84% to 25%—a gap of nearly 60 percentage points—suggests Microsoft’s non-OpenAI AI business is growing roughly at the industry’s average pace, rather than matching the performance a leader should deliver.
How Single-Customer Concentration Reshapes Microsoft’s Valuation Logic
Investors’ pricing logic for Microsoft’s AI business appears to be undergoing a profound shift. Since 2026 began, Microsoft’s stock price has declined nearly 20%, leaving it at the bottom among the "Big Tech Seven." In June 2026 alone, the stock fell nearly 19%, marking its worst monthly performance since the internet bubble burst in December 2000. Even after Microsoft disclosed $24.1 billion of OpenAI-driven revenue contribution, the stock still dropped 1.1% on the day, to $487.46 per share.
The market’s central concern is concentration risk. Microsoft’s remaining performance obligations (RPO) for commercial cloud totals $625 billion, of which about 45% is tied to OpenAI. That means that in Microsoft’s future cloud revenue pipeline, nearly half of the scale is built on the assumption that a single partner will continue succeeding. KeyBanc analyst Jackson Ader pointed to a critical question that remains unanswered: within OpenAI’s $24.1 billion contribution, how much comes from revenue-share agreements versus how much comes from cloud computing or other Microsoft services. This distinction matters because if the majority of revenue comes from providing Azure compute services to OpenAI, it resembles more sustainable commercial service revenue. If it comes primarily from profit-sharing tied to an equity investment, the sustainability and predictability deteriorate substantially.
In addition, Microsoft and OpenAI revised their cooperation agreement in April 2026. The revised terms cap the total revenue-share amount OpenAI pays Microsoft at $38 billion, extending through 2030. This means that even if OpenAI’s revenue continues to grow at a rapid pace, Microsoft’s revenue-share take will still be constrained by a hard ceiling. That cap creates an institutional constraint on Microsoft’s long-term AI growth ceiling.
Why Microsoft Struggles to Break Free From OpenAI Dependence
Microsoft is not unaware of the risks of relying on a single source. In recent years, it has taken multiple steps to reduce dependence on OpenAI, including investing in competitor Anthropic PBC and providing compute support, as well as accelerating the development of its own AI models. A lead executive overseeing Microsoft’s AI model business has stated publicly that the company is "paying Anthropic a significant amount of money" with the goal of reducing and ultimately eliminating that cost. Within core office products, Microsoft has also begun substituting OpenAI and Anthropic models with internally developed MAI models in applications such as Excel and Outlook.
However, these efforts are unlikely to change the fundamental revenue-structure reality in the near term. OpenAI remains Microsoft’s largest single customer for Azure. While Microsoft said in its fourth quarter that about $51 billion in incremental commercial bookings came from customers outside AI startups, over the entire fiscal year, OpenAI still contributes most of the annual bookings growth.
The underlying reason is the compute lock-in effect. Training and running large-scale models at OpenAI consumes massive Azure compute resources. That compute consumption directly converts into revenue for Microsoft’s AI business. As long as OpenAI maintains its current iteration pace and user scale, Azure demand from OpenAI will not disappear. If Microsoft were to proactively reduce the compute it supplies to OpenAI, it would not only sacrifice near-term revenue, but may also push OpenAI toward other cloud providers—an outcome Microsoft would be least willing to accept.
Can High Capital Expenditure Sustain Long-Term AI Returns?
Microsoft’s investments in AI infrastructure are substantial. In Q4 2026, Microsoft’s capital expenditures reached $41 billion, up 69% year over year. Total capital spending for fiscal year 2026 exceeded $145 billion, with 88 additional data centers added during the year. Microsoft expects 2026 calendar-year capital expenditures of about $190 billion. After an accounting methodology change (extending data center useful life from 15 years to 25 years), reported spending falls to roughly $175 billion.
The logic for returns on this massive capital base depends on expectations that AI compute demand will keep rising. Azure revenue surpassed $100 billion for the first time in fiscal year 2026, with quarterly year-over-year growth of 43%, the fastest pace since early 2022. Yet when roughly 70% of AI business revenue comes from a single customer, the payback cycle and risk exposure of capital spending must be reassessed.
If OpenAI’s business cadence shifts—whether due to reduced model training demand, a move to other cloud providers, or slower growth in its own revenue—Microsoft’s massive data-center cluster built to stockpile AI compute capacity would face underutilization risk. About two-thirds of capital spending goes into short-cycle assets like CPUs and GPUs. The depreciation cycle for these hardware assets is far shorter than for data center buildings, which means Microsoft faces more urgent pressure to recover investments in compute equipment.
What the "Winner-Takes-All" Setup Means for Crypto AI Projects
Microsoft’s deep linkage with OpenAI is a textbook example of the current AI industry’s winner-takes-all dynamics. Top large-model companies concentrate capital, compute, and users, while cloud providers lock in compute demand by partnering closely with those leading model firms. Under this setup, resources and growth concentrate among just a handful of companies.
For the crypto AI track, this structure offers a two-sided lesson. On one hand, decentralized AI initiatives aim to break exactly this centralized pattern. By building AI infrastructure on distributed compute networks, open-source models, and cryptoeconomic incentives, they seek to avoid dependence on a single centralized entity. The concentration risk exposed by Microsoft’s reliance on OpenAI also provides a real-world endorsement of decentralized AI’s value proposition.
On the other hand, crypto AI projects face their own economic-model challenges. Decentralized compute networks must compete with centralized cloud services on cost, latency, and trust. If they cannot achieve sufficient scale on both the compute supply and demand sides, decentralized AI projects may still hit a growth ceiling. Microsoft’s case shows that even among the most resource-rich tech giants, sustainable AI growth still depends heavily on the diversification of customer structure—an insight that also applies to crypto AI projects.
Structural Risks in the AI Sector, Seen Through the Microsoft Case
The fact that Microsoft’s AI revenue is highly concentrated in OpenAI reveals a deeper structural feature of today’s AI industry: value distribution and risk distribution across the value chain are not symmetric. Cloud providers carry much of the capital burden for building compute infrastructure, yet growth in AI business revenue depends heavily on the commercialization progress of a small number of major model companies.
This structure can generate rapid growth in an industry upcycle. Microsoft’s AI business grew from a $13 billion annual run rate to more than $37 billion in roughly three quarters. But during an industry downturn or volatility period, the negative effects of concentration are amplified. If growth slows among leading model companies or if partnership terms change, the cloud providers bound to those model firms will face a sudden drop in revenue growth.
Microsoft’s response—investing in multiple model providers, developing models in-house, and adjusting the revenue-share cap with OpenAI—aims to diversify risk. But the real-world effectiveness of these measures still needs time to prove itself. For the entire AI sector, Microsoft’s case provides an important analytical framework: when assessing the growth prospects of any AI-related business, the degree of diversification in revenue sources and customer concentration should be treated as core metrics—on par with revenue scale.
Summary
In Microsoft’s fiscal year 2026, its annualized AI business revenue reached $37 billion, with OpenAI contributing about $24.1 billion—roughly 70%. After excluding OpenAI, Microsoft’s AI-related commercial bookings growth rate fell sharply from 84% to 25%. This revenue structure highlights how dependent Microsoft’s AI growth is on a single source: approximately 45% of remaining commercial cloud RPO is tied to OpenAI; the stock is down nearly 20% over the year; and market doubts about the quality and sustainability of the growth have continued to deepen. Microsoft has attempted to diversify risk through investments in Anthropic, in-house MAI models, and revisions to revenue-share cap arrangements with OpenAI. However, in the near term, the revenue structure is unlikely to change materially at its core. For participants in the AI sector, Microsoft’s case suggests that beyond growth scale, the health of revenue diversification and the controllability of customer concentration are key dimensions for evaluating long-term AI business value.
FAQ
Q: Is Microsoft’s $37 billion AI revenue an actually collected figure?
The $37 billion represents the AI business annual run rate disclosed by CEO Nadella in Microsoft’s March 2026 quarter, not the confirmed full-year revenue. Based on Bloomberg’s estimates using 123% growth, Microsoft’s actual AI revenue for fiscal year 2026 is about $34 billion. The $24.1 billion figure is the actual revenue from OpenAI disclosed by Microsoft in its 10-K filing.
Q: What exactly does the $24.1 billion OpenAI pays Microsoft include?
Under the cooperation agreements, OpenAI’s payments to Microsoft mainly consist of three parts: fees for Azure cloud computing and inference compute, costs related to AI model development, and a certain percentage of revenue sharing.
Q: Is there a cap on the revenue share between Microsoft and OpenAI?
Yes. In April 2026, Microsoft and OpenAI revised their cooperation agreement, capping the total revenue-share amount OpenAI pays Microsoft at $38 billion. The arrangement runs through 2030.
Q: How did Microsoft’s stock perform in 2026?
Since 2026 began, Microsoft’s stock price has fallen nearly 20%, placing it at the bottom among the "Big Tech Seven." In June 2026 alone, it fell nearly 19%, marking its worst monthly performance since December 2000. After the disclosure of the $24.1 billion OpenAI revenue contribution, the stock still fell 1.1% on the day.
Q: What steps is Microsoft taking to reduce dependence on OpenAI?
Microsoft has invested in Anthropic PBC and provided compute support, replaced OpenAI and Anthropic models with internally developed MAI models in products such as Excel and Outlook, and adjusted the revenue-share cap terms with OpenAI.
