NVDA
PLTR
The new AI middleman
For the last couple of years, AI has looked like a heavyweight title fight: OpenAI vs. Anthropic vs. whoever else can throw the biggest model punch. Palantir CEO Alex Karp is basically saying, “Cute. But the real money might be in the referee.”
Palantir’s pitch is simple: let enterprises use whichever model works best — Nvidia’s Nemotron one day, a proprietary model the next — without forcing them to hand over their data, workflows, or intellectual property like it’s the last slice of pizza at a bad office party.
Why this matters for investors
That matters because customers are getting twitchy about lock-in. If a model provider can learn from your usage and then eventually compete with you, that’s not a platform — that’s a landlord with better branding.
Palantir wants to own the orchestration layer, which means:
- customers keep flexibility across models
- security and governance become the selling point
- the value shifts from the model itself to the software that manages it
Nvidia gets a seat at the table
The article also says Palantir recently launched a platform to help U.S. government agencies securely deploy and customize Nvidia’s open-source Nemotron models. That makes Nvidia a real business counterparty here, not just a passing cameo.
Karp’s broader point is that as open-source models improve, companies may care less about worshipping one foundation model and more about who can make the whole thing usable, safe, and not a compliance nightmare.
Big picture
If Karp is right, the AI gold rush may stop rewarding only the model builders and start favoring the folks building the plumbing. That’s good news for Palantir’s “AI operating layer” story — and a reminder that in tech, sometimes the winner isn’t the star quarterback. It’s the team that sells the helmet, the playbook, and the Wi‑Fi.