Palantir CEO Alex Karp: Companies Reluctant to Transfer Alpha to OpenAI or Anthropic
Palantir CEO Alex Karp stated that companies are no longer willing to transfer core alpha to LLM providers like OpenAI or Anthropic. He questioned the token-based charging model of LLMs, pointing out that if true value is created, a revenue-sharing model should be adopted instead of fixed fees, emphasizing that customers are unwilling to pay the true costs. Karp believes that Palantir's ontology application layer and computing business can achieve similar effects to cutting-edge models without transferring corporate alpha. Source: Public Information
ABAB AI Insight
As a co-founder of Palantir, Alex Karp has previously criticized the business models of large models, and his recent comments continue his stance on data sovereignty, similar to the emphasis on control of sensitive data in the company's early government contracts. Palantir builds proprietary knowledge graphs through its ontology layer, leveraging platform resources to help clients retain data control, motivated by concerns over LLM privacy risks, differentiating competition, and enhancing high-margin application revenue. The trend towards privatized deployments, similar to Snowflake or Databricks, indicates that corporate AI deployment is transitioning from cloud-based general models to local/proprietary ontologies. Essentially, this reflects a concentration of capital: companies refusing to let alpha flow out drives data and application layers to concentrate on platforms, with middleware like ontology localizing model capabilities, reducing dependency risks while enhancing pricing power and cash flow for vendors like Palantir. ABAB News · Cognitive Laws 1. Token charging is easy, but transferring alpha is difficult; corporate data sovereignty cannot be relinquished. 2. True value should be shared, not fixed token fees. 3. Ontology retains alpha, while general LLMs only sell computing power.