You probably think Microsoft is the undisputed king of AI. Satya Nadella’s face beams from every earnings call. Azure is booming. Copilot is everywhere. The narrative is simple: Microsoft bet on OpenAI early, and that bet is printing money.
But here’s the thing about empires built on borrowed land: the lease can expire.
I’ve been digging into the fine print of Microsoft’s latest filings. Buried in the regulatory language, a single figure screams louder than any press release: roughly 70% of Microsoft’s AI revenue is tied directly to OpenAI. Not to Azure. Not to Copilot. Not to a diversified portfolio of AI bets. To one company—one that Microsoft does not own, one that is actively building its own infrastructure, and one that has every incentive to walk away.
Let that sink in. The world’s second most valuable company is running a AI business where seven out of every ten dollars come from a single, unowned partner. That’s not a moat. That’s a concentration risk wearing a hoodie.
I’ve spent years watching tech giants build dependencies. I’ve seen Apple’s reliance on Foxconn, Amazon’s on AWS (wait, that’s their own), but this is different. This is like a landlord building a mall and then discovering the only tenant is a startup that might open its own store tomorrow.
Here’s the twist: Microsoft’s AI success is actually OpenAI’s success. When OpenAI grows, it needs more compute. More compute means more Azure revenue. That’s the partnership story. But the deeper truth is that OpenAI’s growth also gives it leverage. The more successful OpenAI becomes, the more it can negotiate with Google Cloud, AWS, or even its own custom chip project. And if OpenAI ever decides to bring compute in-house, Microsoft doesn’t just lose a customer—it loses 70% of its AI revenue narrative.
Don’t take my word for it. Look at the filings. Read the risk factors. Microsoft explicitly warns that its AI revenue is “concentrated” on a single partner. In corporate speak, that’s the equivalent of screaming “fire” in a smoke-filled room.
For investors, this is a wake-up call. The stock price of Microsoft has been buoyed by AI hype. But the emperor’s new clothes are made of API calls to a company that might one day decide to rent its own servers. Enterprise buyers, you’re locking into Copilot, but Copilot’s brain is leased. If the relationship sours, your entire AI strategy could be stranded.
So here’s my position: Microsoft’s AI leadership is not structural—it’s relational. And relationships, especially in tech, can change faster than a quarterly earnings report. The question isn’t whether OpenAI will leave Microsoft. The question is whether Microsoft can build a real moat before the lease is up.
Right now, they’re a toll collector on a highway that someone else is paving. If you’re betting on Microsoft’s AI future, you’re betting on a partnership that’s one bad board meeting away from collapse.
FAQ
Q: Does Microsoft actually own OpenAI?
A: No. Microsoft has invested billions and has a revenue-sharing agreement, but it does not control OpenAI. OpenAI remains an independent company that can choose its compute partners.
Q: What happens if OpenAI leaves Azure?
A: Microsoft would lose roughly 70% of its AI revenue. The stock would take a massive hit, and Copilot’s underlying models would need to be retrained or replaced, disrupting enterprise customers.
Q: Isn't Microsoft diversified with its own models like Phi?
A: Phi is a small model for specific tasks. The massive revenue—and the market’s AI narrative—still hinges on OpenAI’s frontier models. Diversification is a work in progress, not a current reality.