Palantir's Karp Slams OpenAI and Anthropic Over Token Pricing
CEO Alex Karp argues that soaring token costs are pushing enterprises toward open-weight AI models, signaling a structural shift in the market.
Palantir CEO Alex Karp has launched a pointed critique of the dominant commercial AI model providers, singling out OpenAI and Anthropic for what he describes as a fundamentally broken pricing structure. In Karp's view, the relentless escalation of token costs is not a minor friction point for enterprise customers — it is a warning signal that something has gone "completely wrong" with how frontier AI companies have chosen to monetize their products.
The phenomenon Karp is targeting, sometimes called "tokenmaxxing," refers to the incentive model where providers benefit financially when applications consume more tokens — the units of text that large language models process and generate. Critics of this structure argue it creates a perverse dynamic: the more capable and verbose an AI system becomes, the more expensive it grows to operate, often at rates that outpace the productivity gains businesses actually realize.
Read more Caidya and Simbec-Orion Merge to Build Full-Cycle CRO Platform →
The practical consequence, according to Karp, is that cost-conscious enterprises are increasingly pivoting toward open-weight models — AI systems whose parameters are publicly released and can be run on a company's own infrastructure without per-token licensing fees. This shift has been quietly accelerating across industries as CFOs scrutinize AI line items and demand clearer returns on deployment spending.
Karp's comments carry strategic weight given Palantir's own positioning. The company has built its enterprise AI platform around operational efficiency and decision-support tools for government and commercial clients, placing it in direct competition with the broader ecosystem of API-dependent AI applications that rely on OpenAI or Anthropic's hosted models. His critique is as much a market thesis as it is a broadside — efficiency-first AI architectures, he implies, will outlast token-hungry alternatives.
Whether Karp's framing resonates across the enterprise technology sector remains to be seen, but the underlying tension he identifies — between frontier model capability and the economics of scaled deployment — is a genuine fault line in the AI industry's next phase of growth. Continue reading at US Top News and Analysis.
