OpenAI cuts a deal with Microsoft to sell on AWS — and it’s a bigger win than it looks

OpenAI just pulled off something that would’ve been unthinkable a year ago: it got Microsoft to greenlight selling its products on Amazon Web Services.

That’s the headline takeaway from a new agreement between the two companies, buried under the usual corporate language about “strategic alignment” and “mutual growth.” But if you squint past the press release boilerplate, this is actually a pretty big deal.

Let me back up. OpenAI has been tied to Microsoft’s cloud infrastructure since the early days — Azure was its exclusive compute provider, and Microsoft poured billions into the partnership. That exclusivity was always a double-edged sword. On one hand, it gave OpenAI access to essentially unlimited compute. On the other, it meant any customer who wanted to use GPT-4 or DALL-E 3 had to do it through Azure, not AWS. That’s a real limitation when AWS still owns roughly a third of the cloud market.

So what changed? Money, mostly. OpenAI is reportedly in the middle of a $50 billion funding round — yes, you read that right — and part of that valuation hinges on showing investors it can scale revenue beyond the Microsoft ecosystem. Selling on AWS opens up a massive new distribution channel. Enterprise customers who are locked into AWS contracts or simply prefer it can now access OpenAI models directly, without having to spin up an Azure account.

Microsoft, for its part, gets a revenue-share cut from those AWS sales. That’s a pragmatic move. Rather than trying to force every OpenAI customer onto Azure — which was never going to work for everyone — Microsoft takes a slice of the pie regardless of where the compute happens. It’s the same logic that made Microsoft put Office on iOS and Android years ago: if you can’t beat ‘em, take a cut.

I’d be lying if I said this doesn’t carry risks for Microsoft. AWS is its biggest cloud competitor, and giving OpenAI the freedom to sell there could erode Azure’s differentiation. But Microsoft seems to have decided that owning a piece of OpenAI’s growth is worth more than protecting Azure’s exclusivity. That’s a bet on the model provider winning, not the infrastructure provider.

There’s another angle here that’s getting less attention: regulatory pressure. The FTC and EU have been sniffing around the Microsoft-OpenAI relationship for months, worried it’s too cozy. By loosening the exclusivity clause, Microsoft can argue that OpenAI operates independently and that the market remains competitive. It’s a smart defensive move that also happens to make business sense.

What does this mean for the rest of us? If you’re a developer or a company building on OpenAI’s APIs, this is good news. More competition among cloud providers should mean better pricing and fewer lock-in headaches. And if you’re already on AWS, you can now use OpenAI models without adding another vendor to your stack.

I’m curious to see how Google responds. It’s been pushing its own Gemini models hard on GCP, and now it faces a more aggressive OpenAI presence on AWS. The cloud AI arms race just got a lot more interesting.

This deal isn’t going to make headlines the way a new model launch does, but it might matter more in the long run. Distribution is everything in enterprise software, and OpenAI just doubled its addressable market overnight.