The number had been the subject of months of investor guessing, and a new regulatory filing finally supplied it: about $24 billion. That is what Microsoft booked from OpenAI over the past year, and the filing added the caveat investors had feared along with the figure they wanted: most of Microsoft’s AI business revenue comes from that single relationship. The company’s AI growth is real, the document says in effect, and it is also dependent, and the two facts must be read together.
The dependence has been the central question about Microsoft’s AI story since the boom began. Microsoft has spent billions on OpenAI, invested in its compute capacity, and built its AI products on OpenAI’s models, and the market has rewarded the strategy with a valuation that assumes the relationship keeps compounding. The filing now quantifies what investors had been asked to take on faith. The $24 billion is a large number, larger than most standalone AI companies generate, and its existence explains why Microsoft’s cloud business has grown as fast as it has.
The composition of the figure is less clear. The filing does not break the $24 billion into its parts, and analysts have spent the days since its release trying to reconstruct the pieces: compute contracts, under which Microsoft supplies the infrastructure OpenAI trains and runs its models on; profit-sharing arrangements, which tie Microsoft’s returns to OpenAI’s revenue; and equity stakes, whose gains flow through a different part of the financial statements. The opacity is not accidental, analysts said. Microsoft and OpenAI have structured their relationship to give each side flexibility, and the structure resists simple description.
The market’s reaction to the filing was measured, a sign that the figure landed within the range of expectations. The larger issue is what the disclosure implies for the future. Microsoft’s AI revenue is not an independent stream; it is a function of OpenAI’s health, OpenAI’s spending and OpenAI’s choices. If OpenAI decides to build its own compute capacity, or to diversify its suppliers, the revenue that flows through Microsoft’s income statement would shrink, and the growth narrative that supports the stock would lose its foundation.
The relationship has already survived its share of strains. OpenAI’s leadership has publicly discussed the costs of relying on a single cloud partner, and the two companies have renegotiated their terms more than once as OpenAI’s needs changed. The current arrangement, with Microsoft supplying much of OpenAI’s compute while OpenAI’s models power Microsoft’s products, is a mutual dependence: each side needs the other, and each side knows the other knows. The filing’s $24 billion is the price of that mutualism, and it is now on the public record.
For investors, the disclosure reframes how to value Microsoft’s AI business. A revenue stream tied to one customer’s decisions is worth less than an independent stream, and the multiple the market applies to Microsoft’s AI growth will now have to account for the concentration. The counterargument, made by Microsoft’s supporters, is that the relationship is structural rather than transactional: OpenAI’s models need the scale of Microsoft’s cloud, and Microsoft’s cloud needs the demand OpenAI generates. Concentration, in that view, is a feature of the arrangement rather than a flaw in it.
The deeper question is where the AI revenue actually comes from. Microsoft sells AI products to its own customers, from Office copilots to Azure AI services, and much of that revenue is powered by OpenAI models. The filing’s disclosure suggests that a substantial share of what the company calls AI revenue originates with OpenAI’s own growth: usage of the models, the compute behind them, the enterprise deals OpenAI signs. That distinction matters because it determines how much of Microsoft’s AI story is its own, and how much is rented.
The filing’s timing adds to its significance. Microsoft reports its quarterly results later this month, and the disclosure gives analysts a fresh template for parsing the AI portion of the numbers. The company has consistently argued that its AI revenue spans multiple categories, from cloud services to productivity software, and the filing does not contradict that claim; it simply shows where the largest single source sits. For the first time, investors can measure the gap between Microsoft’s AI story as the company tells it and the same story as the numbers describe it.
Whether the relationship remains the foundation of Microsoft’s AI revenue will depend on decisions in both companies. OpenAI has said it plans to expand its own compute footprint, and its suppliers include more than one cloud provider, a fact the filing acknowledges indirectly by describing the revenue as coming from a series of agreements. Microsoft, for its part, has been building models and tools of its own, reducing its reliance on any single provider’s technology. Both companies are hedging the partnership they depend on, and the $24 billion figure is the measure of what is at stake in that careful hedging.
For now, the filing answers the question investors have been asking since the beginning of the boom: how much is the OpenAI relationship worth? The answer is about $24 billion a year, and the follow-up question is already forming: what happens if it changes? Microsoft’s position is that the relationship will deepen, that OpenAI’s growth is Microsoft’s growth, and that the dependency is a strength. The market will test that claim with every quarterly report, and the $24 billion figure will be the baseline against which the tests are scored.
