Via 247wallst.com
CEO Alex Karp escalates his war of words against OpenAI and Anthropic, arguing their models strip enterprises of data sovereignty.
Palantir CEO Alex Karp has a message for the biggest names in AI: you’re building the wrong future. And he’s backing that claim with a quarter that saw revenue climb 93% year-over-year to roughly $1.94 billion.
Following Palantir’s Q2 2026 earnings release on August 3-4, Karp sharpened his ongoing critique of what he calls “frontier AI” labs, taking direct aim at companies like OpenAI and Anthropic. His core argument is blunt: these labs are prioritizing model expansion over the things enterprises actually need, like control, privacy, and ownership of their own intellectual property.
The sovereignty pitch #
Karp accused frontier labs of attempting to “drug addict” enterprises to controlled AI futures. The implication is that companies feeding their data into closed systems run by third-party labs are slowly ceding operational autonomy.
Palantir positions itself as the alternative. The company’s platform is model-agnostic, meaning it doesn’t force customers to use any single AI provider’s models. Instead, it advocates for open-weight AI models deployed within secure, customer-owned environments. A partnership with NVIDIA for sovereign AI implementations underscores the approach, giving clients the ability to run powerful AI capabilities on their own infrastructure.
Karp introduced another term into the discourse: “tokenmaxxing.” He used it to describe what he sees as frontier labs’ tendency to push enterprises toward overly simplified, controlled interactions with their own data. According to Karp, every enterprise customer Palantir works with has expressed dissatisfaction with this dynamic.
Numbers that command attention #
Palantir’s Q2 revenue of $1.94 billion represented a 93% year-over-year jump, while GAAP net income hit $1.062 billion.
The US commercial segment saw revenue growth accelerate to approximately 149% year-over-year in Q2. Karp has been voicing these criticisms since at least June 2026, but the post-earnings commentary represents a clear intensification.
Why enterprises are listening #
When a company sends proprietary data through a third-party AI model, questions about where that data goes, who can access it, and whether it trains future versions of the model become genuinely consequential.
Palantir has deep roots in defense and intelligence work, sectors where data sovereignty isn’t a preference but a requirement. The company’s pitch for sovereign AI, where models run inside controlled environments rather than on external servers, aligns naturally with the security posture these customers already maintain.
Increasing tensions around the potential exploitation of US technology by foreign entities have made data control a national security conversation. Palantir’s model-agnostic, deploy-on-premise approach sidesteps many of these concerns entirely.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our