Palantir posted one of the strongest quarters in its history on August 4, 2026, and CEO Alex Karp used the moment to escalate a public fight with Anthropic and OpenAI, accusing frontier AI labs of hoarding enterprise data to lock in customers.
Quick facts
- Palantir’s Q2 2026 revenue hit $1.94 billion, up 93% year-over-year, beating analyst estimates of $1.81 billion.
- U.S. commercial revenue surged 149% to $764 million; U.S. government revenue grew 90% to $809 million.
- Palantir raised its full-year 2026 revenue guidance to $8.15-8.16 billion, up from a prior top end of roughly $7.66 billion.
- Net income was $1.06 billion, versus $326.7 million a year earlier; adjusted EPS of $0.41 beat the $0.28 consensus by 46%.
- CEO Alex Karp used the earnings call and CNBC interviews to accuse frontier AI labs of using enterprise customer data to build lock-in.
A genuinely unusual growth number
Per CNBC’s coverage, Karp told the network “forget consensus,” arguing no business at Palantir’s scale has grown anywhere close to this fast, and that the momentum should continue for at least another 18 months. The stock jumped double digits after hours. That’s a real result, not just favorable framing: the company has now beaten adjusted EPS estimates for nine consecutive quarters, and U.S. commercial revenue has climbed 380% since 2024 once compounding is accounted for.
Karp’s actual argument against Anthropic and OpenAI
The more pointed story is what Karp said about competitors. According to the earnings call transcript, Karp accused frontier labs of what he called trying to “drug addict” enterprises to a future the labs control, specifically naming Anthropic, and argued that businesses sending prompts and proprietary data straight to a model provider’s API are effectively handing over their competitive advantage as training data for future models. His pitch is that Palantir’s platform lets customers keep that data and workflow logic on their own infrastructure while still using whichever underlying models perform best for a given task, open or closed.
“Sovereign AI” as the actual sales pitch
Per Palantir’s own earnings release, the company describes this quarter’s growth as proof that “demand for AI sovereignty has now been unleashed.” On the earnings call, Palantir executives described a “sovereign boot camp” that drew what they called overwhelming interest from CEOs, and said the company is increasingly helping customers evaluate open-weight and closed models against their own custom benchmarks rather than generic leaderboards, specifically to avoid getting locked into one frontier provider that might not perform best for their actual workload.
Why this fight matters beyond one earnings call
Karp’s criticism lands at a moment when enterprise buyers are actively weighing exactly this tradeoff: send data to a frontier lab’s API and get the best available model performance, or keep data in-house and accept some performance gap. Palantir’s results suggest a real, growing market is choosing the second option, or at least paying Palantir to help them do both without fully committing to either. That’s a direct challenge to the business model frontier labs like OpenAI have built their enterprise pitch around.
Key takeaway
Palantir’s growth numbers are real and independently verifiable; its characterization of rivals “drug addicting” enterprises to their models is Karp’s own framing, not a neutral description. Worth watching regardless of which side of that fight you find persuasive: whether more enterprise buyers start demanding model-agnostic, data-sovereign platforms as a standard requirement rather than a differentiator.
Leave a Reply