NBIS
PLTR
AI, but make it an asset
Palantir is leaning hard into a thesis that sounds a little like the anti-Uber of AI: don’t just keep paying per prompt, own the thing you’re building. In its latest partnership with Nebius, Palantir is backing an approach where eligible customers can run open models on dedicated infrastructure while keeping control of their compute, data, and trained models.
Why this matters
CEO Alex Karp has been swatting at the “pay-per-token” model for months, arguing that companies are basically mailing their best operational know-how to someone else’s black box and renting it back. With Nebius, Palantir gets a real-world way to support the opposite: AI systems that improve inside a customer’s own walls instead of drifting off into a third-party subscription void.
- Enterprises keep ownership of proprietary data and trained models
- Usage costs may shift from pure consumption toward building durable assets
- Palantir’s pitch becomes about workflow sovereignty, not just software
The investor angle
This is not the kind of partnership that instantly adds a new revenue mountain. But it does reinforce the story Palantir has been selling: it’s not just another AI vendor, it’s trying to define how companies buy and control AI in the first place. If that framing sticks, Palantir could end up with a lot more leverage than a typical infrastructure middleman.
Big picture: if AI turns from “renting brains by the minute” into “owning the brain you trained,” Palantir wants to be the company handing out the blueprints.