Microsoft reports fast-growing AI revenue every quarter. New disclosures suggest most of that growth traces back to a single customer, and that customer is also one of Microsoft’s biggest investments.

What he said

Tech industry critic and newsletter writer Ed Zitron dug into Microsoft’s latest financial disclosures in a post on his newsletter, Where’s Your Ed At. Microsoft disclosed that for fiscal year 2026 (the twelve months ending June 30, 2026), it recorded $24.1 billion in revenue from its commercial arrangements with OpenAI, including revenue-sharing payments, plus $6.0 billion in accounts receivable from OpenAI as of that date, meaning money OpenAI owed Microsoft but hadn’t yet paid.

Bloomberg estimated Microsoft’s total AI-specific revenue for the same year at roughly $34 billion, based on the 123% annual growth rate Microsoft reported in March. By that math, OpenAI alone would account for around 70% of Microsoft’s AI revenue, and Zitron notes the $24.1 billion figure works out to more than 7% of Microsoft’s entire company-wide revenue for the year, not just its AI business.

Zitron’s read on what that means: “We are now four years into the AI bubble, and Microsoft has little to show for it other than one very large and very unsustainable company that requires near-infinite resources to keep paying its cloud compute bills.”

Who he is

Zitron is the CEO of the media relations firm EZPR and writes Where’s Your Ed At, a newsletter focused on tech industry criticism. He also hosts Better Offline, a podcast about the technology industry produced with iHeartRadio. He has been one of the more consistently skeptical voices on AI industry economics over the past two years, and this piece continues that line of reporting rather than starting it.

What he gets right, and where it’s overstated

The financial figures themselves come straight from Microsoft’s own disclosures, plus Bloomberg’s published estimate for the AI-specific revenue split, so the core numbers are solid ground: Microsoft’s AI revenue is heavily concentrated in one customer, and that customer’s spending is itself a two-way relationship, since Microsoft has also invested in OpenAI. That concentration is a legitimate, quantifiable thing for anyone tracking Microsoft’s AI business to know about.

Where Zitron’s framing runs ahead of the data is in treating concentrated revenue as proof the strategy has already failed. Microsoft’s AI business growing quickly because of a large, closely tied partner is not automatically evidence the partnership is doomed. It is evidence of concentration risk, the danger that comes from depending heavily on one relationship rather than many. Those are different claims, and the disclosures support the first one much more clearly than the second.

The piece also doesn’t cover what happens if OpenAI’s revenue mix shifts, since OpenAI sells through multiple channels and cloud providers, not exclusively through Microsoft. Without that context, the 70% figure describes today’s structure, not a permanent one.

Why it’s notable

Public narratives about AI industry revenue tend to talk in aggregate: how much a company’s “AI business” is worth, full stop. This piece uses a specific disclosure, the actual dollar figure Microsoft reported for one named customer, to show how much of that aggregate number is really one relationship. That’s a meaningfully different and more precise picture than most AI revenue coverage offers, and it’s backed by numbers Microsoft itself put on the record, not estimates or leaks.

What it means for builders

If your product runs on Azure OpenAI Service (Microsoft’s cloud offering that lets you call OpenAI’s models through Microsoft’s own infrastructure instead of going straight to OpenAI), or you’ve built infrastructure decisions around the assumption that Microsoft and OpenAI’s partnership is stable and permanent, this is worth tracking as a real dependency, not just industry gossip. A revenue relationship this concentrated on both sides means changes to OpenAI’s pricing, availability, or business terms could ripple into Microsoft’s own AI roadmap faster than a more diversified customer base would allow.

The practical step isn’t to panic or migrate off Azure OpenAI today. It’s the same advice that applies to any single-vendor dependency: know your exit path. Understand what it would take to move a production workload to a different model provider if pricing or terms changed with little notice, and revisit that plan periodically rather than assuming today’s arrangement is fixed.

End of article