Palantir's U.S. commercial revenue jumped 149% year-over-year to $764 million, driven by a push among companies to reduce their dependence on large AI model makers. The surge highlights this year's central battle in enterprise software: whether businesses should buy AI from model developers like OpenAI or build independent platforms.
Total revenue rose 93% to $1.1 billion in the second quarter. U.S. government revenue grew 90% to $809 million. The stock rose 29.5% after the results and hit $175.23 on August 10, up 87% from its June low of $107. Some analysts expect shares could surpass the all-time high of $207 set last year.
What "AI sovereignty" means for software buyers
CEO Alex Karp used the word "sovereignty" to describe the shift away from model providers. Companies are running their own AI on their own data, using Palantir's platforms to control how models work with corporate information. Karp argues the alternative - paying per "token" to OpenAI, Anthropic, or Google for access - exposes companies to data leaks and wasted spending.
Karp said in a CNBC interview: "They want to addict us to a future that they believe they can control. They basically argue that we have to move toward a future where they own nothing, where there's no profitability in the business, where there are no jobs, and where our adversaries win."
The SaaSpocalypse fear hit software companies earlier this year when Anthropic launched Claude Code and OpenAI released Codex. Those tools let companies build software without buying existing platforms. ServiceNow and Snowflake integration to the model makers instead. Palantir went the other direction, arguing its customers should run AI on their own infrastructure.
Karp said on CNBC: "Every company we deal with, all of the largest and most important government institutions in the world, are asking, 'Why should we pay for something that isn't useful because of tokenmaxxing, and why should we lose proper control of our business?'"
The defense business provides cover for a commercial push
Palantir's biggest customer remains the U.S. Department of Defense, which uses the Maven Smart System for target identification. U.S. government revenue was $809 million. That base lets Karp attack model makers while his company remains insulated from software market shifts. The defense business also drew criticism when Palantir helped Immigration and Customs Enforcement and the Department of Homeland Security access personal records.
The company has expanded into media. USA Today partnered with Palantir to gauge reader demand. Politico, Business Insider, and Germany's Bild signed deals. Reuters also partnered with Palantir. Unions at USA Today criticized the deal because of Palantir's immigration enforcement work.
Why this matters for sales professionals
The AI sovereignty trend changes what you sell and how. If Palantir's story holds, enterprises will push back against paying model providers for token-based services. That means more companies will buy AI platforms that keep data in-house. For sales pros in software or services, the skill of selling a privacy-by-design architecture and clearly explaining token billing risks will be the answer when buyers ask about AI. For AI Learning Path for Vice Presidents of Sales, that means understanding how sales demos of your platform should surface the governance, not just the capability.
Your membership also unlocks:
