Microsoft CEO Satya Nadella just told companies that use AI they’re getting played. In a blog post published Sunday, covered by TechCrunch AI, Nadella warned that enterprises buying AI from proprietary labs are paying for intelligence twice. Once with cash. And again with their most valuable business secrets.
This is a striking move. Microsoft has invested in both OpenAI and Anthropic. So when the CEO of one of AI’s biggest backers starts warning customers to be wary of proprietary models, people listen.
What Nadella actually said
Nadella’s core argument, as reported by TechCrunch AI, is that the price tag on AI hides a second, bigger cost. “You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful,” he wrote. “The better you want the model to perform, the more of that knowledge you have to feed it.”
He points to something the labs call “exhaust.” That’s the trail your team leaves behind:
- The prompts your people write
- The tools your agents use
- The corrections your staff make when the model gets things wrong
That last one is the killer. Every correction, Nadella argues, gets “distilled into institutional know-how.” It’s “the kind of knowledge a competitor could never buy,” and enterprises are handing it over for free.
Why this matters
The worry isn’t new. VCs like Jason Calacanis and Palantir CEO Alex Karp have warned that big AI labs act like Trojan horses, soaking up customer data and potentially becoming competitors to the very companies they serve. What stands out here is who’s saying it now. Nadella runs the company sitting at the center of that whole ecosystem.
He also called out a double standard. AI labs claim fair use rights to train on the world’s public data, then slap restrictive terms on “distillation,” the practice of studying a model’s outputs to build a cheaper one. “I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation,” Nadella wrote. His point: model makers can’t have it both ways. In February, Anthropic accused Chinese open source models of pumping millions of prompts into Claude to improve themselves. Nadella’s argument suggests enterprises deserve that same freedom to learn from the models they pay for.
Nadella’s fix (and the Azure angle)
His solution is exactly what you’d expect from the head of a giant cloud provider. He wants companies to keep ownership of their data, prompts and feedback included, by building “proprietary learning environments” in the cloud. Which cloud? Probably the one where your data already lives. Possibly Azure.
He’s also pushing “orchestration layers.” In plain terms, that’s tooling that lets you swap between AI models from different providers instead of getting locked into one. AI “gateways” that do this are getting popular fast.
Nadella never says “open source” outright, but per TechCrunch AI, that’s the obvious subtext.
The shift already happening
The data backs him up. Idit Levine, CEO of Solo.io, says her enterprise customers are moving open source models onto their own servers after testing proprietary ones. Her pitch, in her words: “Can I take an open source model and run it on-prem? It will do almost 90% of what the big one’s doing. It will cost way less.” Her customers include T-Mobile, ADP, and SAP.
She’s not alone. Vercel and OpenRouter both report a surge in traffic to open source models. Last month, open models made up 29% of all traffic routed through Vercel’s gateway.
What to expect next
Here’s the takeaway for anyone running AI at work. Treat your prompts, corrections, and feedback as assets, not exhaust. Read the terms on customer data before you feed a model your crown jewels. And watch the on-prem open source wave, because it’s picking up speed.
When the CEO of Microsoft starts nudging enterprises away from proprietary lock-in, the trend line only points one direction. As Nadella put it: “In consuming intelligence, you are creating intelligence. And what you create should belong to you.”
More details are available in the original TechCrunch AI report.