Palantir CEO Karp slams 2 AI rivals over token pricing model

Palantir Technologies co-founder and Chief Executive Officer Alex Karp said generative AI sales to enterprises are "structurally broken," arguing that companies are paying to lose their competitive edge.

The Palantir chief executive criticized the token-based pricing model used by OpenAI and Anthropic, saying the industry has "gone completely wrong" by charging enterprises for consumption rather than outcomes.

"Something has gone completely wrong when enterprises are paying for AI tokens and losing their competitive edge," Karp said during a live interview this week, according to reports. The CEO specifically called out the pricing structures of OpenAI and Anthropic, two of the most prominent generative AI companies, arguing that their models incentivize volume over value.

Token-based pricing charges customers based on the volume of data processed by AI models, creating what Karp described as a misaligned incentive structure. Under this model, AI providers earn more revenue when customers use more tokens, regardless of whether that usage generates proportional business value. Karp argued this dynamic leaves enterprises with escalating costs and no durable competitive advantage, as the underlying AI models are available to competitors on similar terms. The critique extends beyond pricing to the fundamental question of whether generative AI, as currently sold, delivers sustainable value to corporate customers.

The critique from one of enterprise software's most prominent voices introduces doubt about the monetization model that has fueled the generative AI boom. If corporate buyers begin questioning the value proposition of token-based pricing, it could pressure AI companies including OpenAI and Anthropic to restructure their commercial models, potentially compressing margins across the sector.

Karp's comments come as Palantir deepens its own push into the AI market with its Artificial Intelligence Platform, or AIP, which the company has positioned as a more outcome-oriented alternative to standalone generative AI tools. Unlike consumption-based pricing, Palantir charges enterprises for platform access and deployment, aligning costs with the value delivered rather than the volume of data processed. The company has reported growing demand for AIP across government and commercial clients, positioning itself as a direct competitor to the consumption-based AI providers Karp criticized.

The tension between these two models reflects a broader debate in the enterprise software industry about how to price AI capabilities. Microsoft Corp. embeds OpenAI's models into its Azure and Copilot products under a hybrid pricing structure that combines subscription fees with consumption charges. Salesforce Inc. offers its Einstein AI platform at a fixed per-user price, while Amazon Web Services provides access to multiple foundation models through its Bedrock service with per-token pricing. The variation in approaches suggests the market has not yet settled on a dominant commercial model for enterprise AI.

For investors, Karp's criticism raises questions about the sustainability of revenue growth at pure-play AI companies that rely on token-based pricing. If enterprise customers push back against consumption models, companies such as OpenAI and Anthropic may face pressure to lower prices or shift to alternative structures, potentially slowing their path to profitability. The debate also highlights the competitive positioning of Palantir itself, which has bet on a platform-based approach that Karp argues delivers more durable value to customers.

This article is for informational purposes only and does not constitute investment advice.