Microsoft and OpenAI’s AGI clause is gone, and that changes everything

The Microsoft-OpenAI relationship has been called a lot of things — partnership, alliance, even a “situationship.” Now it’s just a business deal, and a less weird one at that.

On Monday morning, Microsoft announced a handful of changes to the deal that has defined both companies’ AI strategies for years. The headline: OpenAI is no longer locked into Azure as its exclusive cloud provider. The old language said OpenAI would “ship first on Azure” — now it says that’s only the case unless Microsoft “cannot and chooses not to support the necessary capabilities.” That’s a big caveat, and it opens the door for OpenAI to run workloads on AWS, Google Cloud, or anyone else.

But the quieter, more significant change is the death of the AGI clause.

For those who haven’t been following the fine print, the original deal had a weird provision: once OpenAI achieved artificial general intelligence — defined as a system that can outperform humans at most economically valuable work — Microsoft would lose access to that technology. It was a kind of off-ramp, a way to ensure that the most powerful AI wouldn’t be owned by a single corporation. It also gave OpenAI’s board an escape hatch if they ever felt Microsoft was getting too cozy with dangerous tech.

That clause is now gone. Officially dropped. No more AGI trigger. No more special treatment for the hypothetical superintelligence.

I’ve always found that clause fascinating, mostly because it assumed AGI would be a clear line you cross, like a finish line. Anyone who’s worked on AI long enough knows it’s not that neat. We’ve been arguing for years about whether GPT-4 counts as a spark of AGI, and the answer is always “it depends on who you ask.” The clause was probably unenforceable in practice, but it served as a symbolic guardrail.

Now it’s just a standard cloud partnership. Microsoft remains the primary cloud partner, but OpenAI can serve all its products to customers across any cloud provider. That’s a huge shift for enterprise customers who didn’t want to be locked into Azure just to use GPT-4 or DALL-E.

The timing makes sense. OpenAI has been aggressively courting enterprise customers, and many of those companies are already on AWS or GCP. Forcing them to also sign up with Azure was a non-starter for a lot of deals. Now OpenAI can go where the customers are.

Microsoft, meanwhile, gets to keep its preferred access to OpenAI’s models for its own products — Copilot, Azure AI services, and so on. They’re not losing the crown jewels. They’re just no longer the only game in town.

What does this mean for the AGI debate? Honestly, not much. The clause was always more of a philosophical statement than a practical mechanism. If OpenAI actually builds something that qualifies as AGI, do you really think Microsoft would just politely step aside? The legal battles would be legendary. Dropping the clause now just removes a potential headache that was never going to work cleanly anyway.

But it does signal something: both companies are now treating this as a straightforward commercial partnership, not a grand mission to safely birth superintelligence. The idealism of the early OpenAI days — the nonprofit structure, the capped-profit model, the AGI off-ramp — has been slowly eroded. This is just the latest step.

Is that a bad thing? Depends on your perspective. For enterprise customers, it’s great. More flexibility, less lock-in. For the people who worried about unchecked AI development, it’s another guardrail removed. For Microsoft and OpenAI shareholders, it’s probably a smart business move.

I’ll be curious to see how this plays out with regulators. The FTC and EU have been circling both companies for a while. Removing the AGI clause might actually simplify antitrust concerns — it’s harder to argue they’re a monopoly if OpenAI isn’t tied exclusively to Azure.

Either way, the old deal is dead. The new one is cleaner, more pragmatic, and a lot less weird. I’m not sure that’s a good thing.