Palantir and Nvidia Launch Air-Gapped AI Stack as Token Billing Cracks Enterprise Budgets
Palantir and Nvidia introduced an air-gapped “sovereign AI operating system” built on Blackwell Ultra hardware and open Nemotron models, while Palantir CEO Alex Karp argued that token billing is harming enterprise budgets. Speaking to CNBC on July 1, 2026, Karp called token-based pricing a “wealth tax,” describing a system where companies pay for consumption, receive little return, and may effectively share proprietary data with the providers charging for usage. His remarks came amid reports of budget strain, including TechCrunch confirmation that Uber exhausted its entire 2026 AI coding budget within four months and Microsoft ending Claude Code access for thousands of engineers in its Windows and Office division effective June 30. Karp questioned why providers charge per token if models can deliver major business value. He also noted that claims about enterprises losing intellectual property via APIs are not independently verified, and the article summarizes OpenAI and Anthropic policies that, by default, do not use customer inputs to train models.






