Palantir CEO Karp Blasts Token AI Pricing as Enterprises Shift to Lower-Cost Models
Updated
Updated · CNBC · Jul 1
Palantir CEO Karp Blasts Token AI Pricing as Enterprises Shift to Lower-Cost Models
2 articles · Updated · CNBC · Jul 1
Summary
Alex Karp said enterprises are abandoning a "tokenmaxxing" approach as AI costs climb, arguing OpenAI and Anthropic's token-based model has become misaligned with customer needs.
Lower-cost alternatives are gaining traction because newer frontier models are proving more expensive than earlier versions, pushing companies to focus on return on investment and tools they can own.
Open-weight models and in-house systems are benefiting from that shift, with Karp saying technical buyers increasingly prefer custom proprietary models over broad external services.
China's faster AI progress is adding pressure on U.S. labs, Karp said, warning the industry not to underestimate how quickly Chinese models could narrow the gap.
Earlier this week, Palantir expanded its Nvidia partnership to build custom AI models for U.S. government agencies, underscoring its push toward tailored deployments.
Enterprises are fleeing high AI costs. Are open-weight models the cure, or just a different kind of expensive?
As China's AI accelerates, does America's push for open models create a critical national security vulnerability?
AI labs burn billions while their valuations soar. Is the industry's token-based business model a ticking financial time bomb?
Breaking the Token Trap: Palantir’s 85% Revenue Growth Signals Shift to Outcome-Based AI Pricing
Overview
Palantir CEO Alex Karp recently criticized the dominant AI pricing models, especially the 'tokenmaxxing' practices of OpenAI and Anthropic, just as both companies moved toward going public. His remarks have sparked industry debate about whether these frontier labs truly deliver practical, cost-effective solutions for enterprise clients. Many businesses are spending heavily on tokens for AI productivity but often find these models lacking in real-world efficiency. Karp argues that Palantir offers a better alternative, focusing on AI solutions that work reliably in production and address the real needs and costs faced by large organizations.