Alex Karp just picked the fight that could reshape how every enterprise buys AI. Palantir’s CEO went back on CNBC this week to hammer the frontier labs, and according to Hacker News, his line landed hard: “We have people trying to drug addict us to a future they believe they control.” The target was specific. Anthropic, OpenAI, and the model makers Karp says want a world “where we own nothing, where our businesses aren’t profitable, where none of us have jobs.”
Strip away the theatrics and there’s a real argument underneath. Karp is betting that companies won’t hand their data and their edge to a handful of labs. He wants enterprises to control their own models and run an application layer, like Palantir’s, on top of their own stack. Keep the data in-house. Keep the value in-house.
What’s actually being fought over
This is a battle over who owns the means of production in AI. The labs say your data is safe and never trains their models. OpenAI even says businesses have to explicitly opt in. Karp isn’t buying it. His pitch to customers: “Why would we tokenmaxx and pay people for something that’s not useful and then not control the means that allow us to advance our business while keeping the value of the business inside?”
What stands out is who’s nodding along. This isn’t a lone contrarian.
- Chamath Palihapitiya said Karp “deserves a medal.”
- Satya Nadella posted that “the current regime does precisely the transfer Karp and companies fear.”
- Jamie Dimon, without even hearing the comments, said companies are scrutinizing spend and hunting for return on investment.
When the CEO of Microsoft, a company that has poured billions into OpenAI, publicly validates the critique, the ground is shifting.
Why it matters now
Three forces are converging. Enterprises are getting nervous about lock-in. Open-weight models, the kind you can download and run yourself, are getting good enough to matter. And cheap Chinese models, some accused of distilling U.S. systems, are undercutting the whole premium-API business.
Karp’s take on the distillation fight is the sharpest part. He won’t call China unfair, because the frontier labs did the same thing at scale. “How do you think the models got their value? They distilled all the value of IP everywhere, including enterprise, everywhere,” he said. In other words, the labs built their moats on everyone else’s data. Now everyone else wants their moats back.
In late July, Palantir signed an open letter with Nvidia, Microsoft, and others defending open-weight models as essential to national security. That’s the coalition forming around “AI sovereignty” as Karp calls it.
The numbers behind the noise
Karp can talk this loud because the results back him. Palantir posted 93% overall revenue growth year over year, and U.S. commercial revenue jumped nearly 150% in Q2. When a founder is growing that fast while attacking the biggest names in the industry, people listen. His track record here is short but loud. He was the first executive to publicly go after the labs’ token pricing back in early July, and that comment went viral and pulled Nadella, Dimon, and Palihapitiya into the conversation within days.
The next 1-3 years
Expect the enterprise AI stack to split. Frontier models for the frontier stuff, open-weight and self-hosted models for anything touching proprietary data, compliance, or national security. The pure closed-API play, where you rent intelligence and ship your data out the door, starts to look risky for regulated industries and governments.
Practical takeaways if you’re running AI decisions:
- Audit your data flow. Know exactly what leaves your walls when you call a model.
- Test open-weight models now. Llama-class and Chinese options have closed the gap faster than most planners assumed.
- Push vendors on ownership terms in writing, not just security promises. Karp’s whole point is that assurances aren’t control.
- Separate the workloads worth premium API money from the ones you should run yourself.
Karp is loud, brash, and selling his own solution, so read the pitch with that in mind. But the trend is bigger than him. Control is becoming the new currency in enterprise AI, and the labs are on defense for the first time. More detail on his comments and the full earnings picture is available at the original source.