Satya Nadella on the New OpenAI Deal: ‘We Fully Plan to Exploit It’

Satya Nadella doesn’t mince words. When asked about Microsoft’s newly renegotiated deal with OpenAI, he flat-out said: “We fully plan to exploit it.”

That “it” is the fact that Microsoft now gets to offer OpenAI’s latest models to its Azure cloud customers without paying OpenAI a licensing fee. The old arrangement had Microsoft paying a cut of the revenue from reselling GPT models. The new one flips that: Microsoft keeps all the cloud revenue, and OpenAI gets a different kind of compensation (likely compute credits on Azure, but the exact structure is still under wraps).

This is a bigger deal than most people realize. Microsoft has been OpenAI’s biggest investor and primary cloud partner for years, but the financial terms of the reseller relationship were always a bit murky. Now they’re crystal clear: Microsoft is betting that the real money isn’t in licensing AI models—it’s in being the platform that runs them.

Nadella’s choice of verb is interesting. “Exploit” usually carries a negative connotation, but in business-speak it just means “use aggressively.” And that’s exactly what Microsoft intends to do. They’re going to bundle OpenAI models into every tier of Azure they can—from enterprise-grade GPT-4 Turbo to whatever comes next—and make it frictionless for companies to adopt.

The timing matters. This deal lands right as competitors like Google Cloud and AWS are scrambling to offer their own foundation models. Google has Gemini, AWS has Bedrock with multiple model providers. But Microsoft now has an exclusive reseller relationship with the most popular generative AI models on the market, and they don’t have to split the cloud revenue.

What does this mean for OpenAI? They get continued access to Microsoft’s massive compute infrastructure and presumably some financial upside through the compute credits. But they’ve given up a direct revenue stream from cloud resale. That’s a trade-off that suggests OpenAI is more focused on building the next generation of models than on being a cloud business itself.

For customers, the message is simple: if you want OpenAI models, Azure is the cheapest place to run them. Microsoft can undercut any other provider on price because they’re not paying a per-token royalty. That’s going to put pressure on Google and AWS to either negotiate their own exclusive deals or find ways to differentiate beyond price.

I’ve seen this playbook before. Microsoft did something similar with Office 365—used their platform leverage to bundle and undercut competitors until they owned the market. The difference is that AI models are moving much faster than office suites. The model that’s dominant today might not be dominant in 18 months. Microsoft is betting that OpenAI stays ahead, or at least close enough that the platform lock-in outweighs any model advantage.

It’s a risky bet, but Nadella doesn’t sound worried. He sounds like a guy who just got the best seat in the house for the AI gold rush, and he’s not planning to share the profits.