A $402 Billion Company Just Made Your Argument
Palantir's CEO told national television that companies using OpenAI and Anthropic are paying to give away their most important secrets. Then the company deployed open-weight NVIDIA models in air-gapped environments and reported they beat frontier models on production tasks within 24 hours. The sovereignty argument this community has been making from Discord just got made from a $402 billion earnings call.
The AI Layer (Cycle 2)
This content is for educational purposes only and is not financial, investment, or trading advice. All company data referenced is drawn from public sources.
This community has been making a specific argument for a long time: when you feed your data into someone else's model, you are giving away the thing that makes your work valuable. You run local models, you self-host, you choose open weights over API calls, because you believe that controlling your own intelligence matters more than convenience.
Last week, the CEO of a $402 billion company said the same thing on national television, and the market rewarded him for it.
Palantir reported Q2 earnings that beat expectations across every metric. Revenue hit $1.94 billion against a $1.81 billion estimate. EPS came in at $0.41 versus $0.34 expected. U.S. commercial revenue grew 149% year over year. The company raised its full-year guidance by $498 million. Adjusted free cash flow was $1.22 billion in a single quarter. The market responded by pushing the stock up 37% from pre-earnings levels.
But the number is not the story. The story is what CEO Alex Karp said about the frontier AI labs. He argued that companies using OpenAI and Anthropic are "paying to give away their most important secrets" and that the resulting data becomes "the training data embedded in the foundations of all future models." Palantir's CRO added that customers are "contributing to the commoditization of their own businesses." Their answer is open-weight models deployed on customer infrastructure, where the customer retains the weights, the reasoning traces, and the metadata.
Palantir then launched an engine for NVIDIA's open-weight Nemotron models in air-gapped environments. Within 24 hours of integration, a standard Nemotron model without post-training reportedly beat frontier models on five production tasks inside Palantir's stack.
The question worth raising on the next call: The sovereignty argument this community has been making from Discord is now being made from a $402 billion earnings call. Does that validate the thesis, or does it mean the thesis is about to get co-opted by enterprise platforms that will charge a premium for the same principles you are building for free?
Data Sources: Kavout Fundamental Analyst, Kavout News Sentiment (PLTR), August 2026. Deep Value Investing, "Palantir Q2: Karp Takes On The Two AI Labs," Seeking Alpha, August 10, 2026.
Produced by Mike Hernandez