Palantir CEO Alex Karp told shareholders on Monday that the companies building large language models intend to capture the means of production from their partners.
The comment came in a letter following a second quarter where the firm reported $1.9 billion in revenue, a 93% rise compared to the same period last year. Profit hit $1.1 billion, which Karp noted was more than the total revenue generated in the entire year before.
Karp, a former philosophy student with a PhD in social theory, argued that the business models of many AI frontier labs carry Marxist undertones. He suggested these entities operate like the capitalists who, in his view, historically gave rise to socialist movements.
During a conference call with Wall Street analysts, he asked if organisations were prepared to support a future where their work helps adversaries win. He questioned whether a small group living in a specific location, believing they deserve total control because of their lifestyle choices, should dictate the means of production while the rest of the country absorbs the cost of that revolution.
Karp described the situation using jargon common in defence technology, noting that his senior leadership team is entirely male. He contrasted this approach with Palantir’s model, which serves model-agnostic software to governments and enterprises. The firm allows clients to control their data and their AI exhaust, covering prompts, orchestration, and context.
Despite the harsh language, Karp is echoing a point made by other executives, including Microsoft CEO Satya Nadella. The concern focuses on a list of companies that paid for services from Anthropic and OpenAI while those labs launched their own businesses in areas like design tools, healthcare operations, legal work, and drug discovery.
Ultimately, neither side represents economic villains or heroes. The market is changing too quickly for any single outcome to be predetermined, and Palantir’s recent financial results show there is clearly room for all players.
What it means
For businesses, the choice is becoming a matter of data sovereignty rather than just capability. Companies must decide whether to rely on external labs that might monetise their internal workflows or use platforms that keep control of their own data and AI outputs in-house.
