Nadella joins Palantir in warning that AI labs absorb enterprise expertise

AI labs are mining enterprise data, and now it is not just Palantir saying it.

Nadella’s warning

On July 16, 2026, Microsoft CEO Satya Nadella warned that companies using frontier AI models risk handing their know-how to the labs behind them. In his telling, entire industries could see their expertise absorbed and turned into competing products.

He also offered a test for what real control looks like: you should be able to swap out a generalist model without losing the company expertise built into your systems. If your institutional knowledge only exists inside someone else’s model, it is not really yours anymore.

The Karp prelude

Nadella’s comments landed two weeks after Palantir CEO Alex Karp went on CNBC and called frontier labs’ token pricing insane, accusing them of “tokenmaxxing”: charging enterprises for tokens while absorbing their proprietary data.

Palantir did not stop at rhetoric. The company followed with a 9-point AI sovereignty manifesto and, together with Nvidia, a Sovereign AI stack for running open models on your own infrastructure, air-gapped from the outside world.

From rant to buying criterion

The argument is picking up allies. David Sacks called Karp’s warning exactly right on the All-In podcast, and Chamath Palihapitiya backed the math behind it.

What started as one CEO’s rant is turning into a procurement checklist. Sovereign AI, on-premise deployment, and data retention treated as treasure are becoming line items in enterprise vendor conversations, not fringe positions. When Microsoft’s CEO and Palantir’s CEO converge on the same warning from opposite corners of the industry, buyers listen.

The open question for every enterprise now sounds simple and cuts deep: would you trust a frontier lab with your company’s playbook? The vendors betting on “no” just got their strongest talking points yet.