The awkwardness is finally out in the open.

Microsoft spent years and billions of dollars positioning itself as OpenAI’s key partner and infrastructure provider. Now it’s pitching Wall Street on its own AI models and openly competing with the company it helped build.

This isn’t subtle anymore. During Wednesday’s earnings call, Microsoft laid out plans for homegrown AI models, custom tools, and even a competitor to Mythos, the AI system that just broke a major post-quantum cryptography algorithm. The message to investors was clear: we don’t need to rely on anyone else’s AI to keep growing.

And here’s the kicker. Microsoft made $3.2 billion from its investment in Anthropic this fiscal year. OpenAI? “Mixed bag” is how TechCrunch characterized it, which in earnings call speak means “we’d rather not talk about the specifics.”

The partnership that got too successful

This was always going to happen. Microsoft gave OpenAI the compute infrastructure and capital to build GPT-4 and beyond. In return, Microsoft got early access to the models and the ability to integrate them into everything from Office to Azure.

But partnerships built on dependencies don’t last when both sides get strong enough to go alone. OpenAI doesn’t want to be forever dependent on Microsoft’s infrastructure. Microsoft doesn’t want its AI strategy held hostage to another company’s product roadmap and drama.

So now we’re watching the slow-motion divorce play out in product announcements and earnings calls.

Why this matters more than the AI agent hype

Mark Zuckerberg also had an earnings call Wednesday, and he spent it talking about how billions of people will have personal AI agents within five years. It’s a bold prediction, and maybe it’ll happen. But Zuckerberg has been making big bets on the future since the metaverse pivot, and investors have learned to wait for results.

The Microsoft situation is different. This isn’t speculation about what might happen. It’s a fundamental restructuring of who controls the AI infrastructure that matters right now.

Microsoft isn’t just hedging its bets by investing in Anthropic alongside OpenAI. It’s building the capability to compete directly. That means developing models that can go head-to-head with GPT-4 and Claude. It means creating tools that don’t require licensing someone else’s API. It means owning the entire stack.

What happens next

OpenAI will keep building. They’re working on a “family of devices” according to president Greg Brockman, though he wouldn’t say what those devices actually are or when they’re coming. They’re trying to become a consumer brand that people interact with directly, not just an API that powers other companies’ products.

Microsoft will keep integrating AI into everything it sells while building its own models as backup. The Copilot “super app” launching later this year combines chat, coding, and agentic capabilities in one place. It’s Microsoft’s attempt to own the interface layer even if it doesn’t fully own the models underneath yet.

And Anthropic? They’re probably pretty happy right now. They’ve got Microsoft money without the baggage of being Microsoft’s primary bet. They can build Claude without worrying that their biggest investor is simultaneously trying to replace them.

The AI industry spent the last few years in a holding pattern where a handful of labs made the models and big tech companies distributed them. That era is ending. The companies with the resources to build their own models are doing it, and the partnerships that got us here are turning into competition.

It was inevitable. But watching Microsoft announce it during an earnings call, with OpenAI still dependent on their infrastructure, makes it clear just how messy this transition is going to be.