Palantir posts $1.9 billion quarter as Karp questions AI lab incentives
Palantir reported second-quarter revenue of $1.9 billion, up 93% from a year earlier, and profit of $1.1 billion. CEO Alex Karp used the results and the company's shareholder letter to renew his warning that frontier AI labs may not be trustworthy partners for enterprises.
Karp wrote that some builders of large language models intend, knowingly or otherwise, to capture what he called the “means of production” of their partners. On the quarterly call, he argued that customers may pay AI providers while also transferring intellectual property, expertise, prompts, context, and orchestration into those providers' systems.
Record growth amid the AI boom
The criticism did not reflect weak demand for Palantir's products. Rapid AI adoption helped the company deliver record results. Karp noted that Palantir generated more profit in the quarter than it had recorded in total revenue during the comparable period a year earlier.
Palantir presents a different architecture for enterprise and government customers. Its AI and analysis software is model-agnostic, while organizations retain control of their data and what Karp described as their AI “exhaust”: the prompts, orchestration, and context created while systems are used.
That positioning belongs to a wider debate about enterprise AI without dependence on frontier labs as enterprises decide how much strategic control to place with providers of frontier models. Karp's argument is that the commercial terms of AI adoption should account for knowledge generated through use, not merely access to tokens or model outputs.
Competition between partners and providers
The underlying concern extends beyond Palantir. Microsoft CEO Satya Nadella has also raised questions around enterprise dependence on proprietary AI. The issue has gained attention as companies have partnered with or paid Anthropic and OpenAI while those labs have launched offerings in adjacent areas, including design tools, healthcare operations, legal services, and drug discovery.
That does not make AI labs economic villains, nor their customers victims by definition. The market is expanding quickly, and Palantir's own performance indicates that multiple business models can grow at the same time. The practical issue is whether buyers understand the allocation of data, operational knowledge, and competitive value within each agreement.
What enterprises should examine
For business leaders, the quarter reinforces the need to assess AI procurement beyond model quality and usage price. Contracts and technical designs should clarify who controls prompts, context, orchestration, derived expertise, and intellectual property; whether workloads can move between models; and whether a supplier may enter the customer's market.
Palantir's results show that enterprise AI demand is substantial. Karp's challenge is a reminder that adoption decisions also shape where the knowledge created by that demand ultimately resides, making control, portability, and competitive exposure material business considerations.