Technology

Microsoft Books $10 Billion in OpenAI Revenue as Azure Grows 48%

Martin HollowayPublished 6m ago5 min readBased on 7 sources
Reading level
Microsoft Books $10 Billion in OpenAI Revenue as Azure Grows 48%
source:microsoft.com

Microsoft recorded $10 billion in revenue from its commercial arrangements with OpenAI, including revenue-sharing payments, in the quarter ended September 30, 2026, according to TechCrunch. That figure is nearly 11% of the company's total quarterly revenue. Microsoft published the results on October 28, 2026, after the market closed.

Total revenue for the quarter was $92.3 billion. Analysts had expected $90.66 billion, a 16.7% year-over-year increase, according to Reuters. Operating income was $45.9 billion, up 21% year-over-year, or 22% in constant currency, which strips out the effect of exchange-rate movements.

Microsoft reported GAAP net income of $36.6 billion, up 32%, and GAAP diluted earnings per share of $4.92, also up 32%. GAAP refers to the standardized accounting rules companies use in filings. On a non-GAAP basis, net income was $36.4 billion, up 18%, and diluted EPS was $4.89, up 18%. The non-GAAP figures exclude the impact of Microsoft's investments in OpenAI, which is why the two measures diverge sharply this quarter. Readers comparing them should note that the gap comes from that exclusion rather than from operating performance.

Azure revenue rose 48% year-over-year. Microsoft Cloud revenue reached $63.9 billion, up 30% (31% in constant currency). The company reported Agents and Infra revenue of $76.8 billion, up 25%. Commercial remaining performance obligation, which is contracted work not yet delivered and often called backlog, increased 73% to $677 billion. Microsoft 365 consumer cloud revenue grew 17%, and industry solutions cloud revenue grew 10%.

The OpenAI relationship now shows up in several figures. Microsoft's revenue from OpenAI for the full fiscal 2026 year was $24.1 billion. Its receivables from OpenAI, meaning amounts owed but not yet paid, stood at $7.2 billion at the end of September 2026, up from $6 billion at the end of June 2026. Microsoft's ownership stake in OpenAI declined from about 25% in June 2026 to 24% in September 2026. Under an October 2025 agreement, OpenAI committed to buying an additional $250 billion in Azure services from Microsoft. Under an April 2026 agreement, OpenAI is to continue sharing revenue with Microsoft through 2030, subject to a cap, and is permitted to work with competing cloud providers.

Because Microsoft's fiscal year begins in July, the September quarter is the first quarter of fiscal 2027, not a continuation of the fiscal 2026 numbers reported earlier in the year.

The mechanics matter for anyone modeling the business. Microsoft's OpenAI-linked revenue arrives through a mix of Azure consumption, revenue-sharing and related commercial terms. The $250 billion commitment is a forward purchase obligation that will be recognized as Azure capacity is consumed. The $1.2 billion rise in receivables within a single quarter shows how large counterparty balances build on the books before they are collected. The stake reduction, from 25% to 24% over three months, appears in the same filing as a separate line item from the commercial revenue.

In my view, the notable feature of these results is how concentrated the dependency has become on both sides. OpenAI accounts for a material share of Microsoft's quarterly revenue, and part of Microsoft's cloud backlog depends on a single customer's commitment. The April 2026 terms allowing OpenAI to work with competing cloud providers reduce the exclusivity that once defined the arrangement, which cuts both ways for Azure's long-term growth profile.

The broader context here is that the same disclosures show the cloud business growing on its own terms. Azure's 48% growth and the 73% rise in remaining performance obligation describe demand that extends well beyond one partner, even as the OpenAI line items draw most of the attention. For infrastructure teams, the practical question is how quickly that backlog converts into deployed capacity. The results give no timetable for that conversion beyond the figures already stated.