Palantir Technologies reported $1.9 billion in revenue for the second quarter of 2026 on August 3, nearly double what it earned the same quarter a year earlier. Profit was $1.1 billion. Wall Street analysts had expected revenue of about $1.81 billion (Benzinga). Palantir's own forecast, issued the previous quarter, had projected revenue between $1.797 billion and $1.801 billion (Palantir IR). The company also raised its full-year 2026 growth forecast to 82%, up from 71% (Palantir IR).

This growth has been speeding up for several quarters. In Q1 2026, total revenue grew 85% compared to a year earlier. In Q4 2025, revenue grew 70% (Palantir IR). For all of 2025, revenue was $4.475 billion, up 56%.

But the biggest story was not the numbers. It was what CEO Alex Karp said about Palantir's competitors.

Karp's argument is about what happens when a company uses AI built by someone else. Think of it like hiring a contractor who watches how you work, learns your trade secrets, and then sets up a competing business next door. In his Q2 2026 shareholder letter, Karp wrote that "There are Marxist overtones and undertones to our business" (TechCrunch; Palantir). He was saying that the companies building the most advanced AI systems intend, knowingly or not, to take control of the productive capacity of the businesses they claim to help.

On a call with Wall Street analysts, Karp asked whether companies would accept a future where their work helps their adversaries win, and only a small group of people ends up controlling the country's means of production. He accused companies of paying for "token self-gratification at real cost" — meaning they are paying AI labs to absorb their intellectual property and expertise into the labs' own systems (TechCrunch).

This is not a new complaint from Karp. On June 10, 2026, he said companies are "unhappy" with the major AI labs and believe the labs only care about "tokenmaxxing" — pushing as much text as possible through their systems (CNBC). Taken together, Karp's comments over the past two months make one argument: when a company feeds its private data into an AI provider's system, that provider can learn from it. Over time, the AI provider becomes a gatekeeper the company cannot easily work around.

Palantir says its software is the solution. The company sells AI tools to governments and businesses that let customers keep control of their own data and the traces their AI use leaves behind — the questions they ask, the logic they apply, the context they provide (TechCrunch). Palantir's 2025 annual report states the company builds technology to foster AI accountability (Palantir IR). The idea is simple: instead of sending private data to a third party who might retain or learn from it, Palantir's platform keeps that information under the customer's control. The AI model underneath can be swapped out like a replaceable part.

Karp also predicted that Palantir would generate $15 billion to $18 billion in free cash flow within two years (Yahoo Finance). He made that prediction on July 25, before the Q2 results. It fits the pattern: revenue has roughly doubled year-over-year for three straight quarters.

The broader context is a market where businesses are moving from testing AI to deploying it in earnest, and the question of who controls the AI pipeline is becoming a real concern for the people writing the checks. Karp's argument is self-serving — Palantir sells the alternative he is pushing. But the tension he points to is genuine. When a company sends its private data and reasoning through a third-party AI system, the provider can learn how that company thinks about its own problems. Over time, that is a transfer of know-how that is hard to measure and impossible to undo.

Worth flagging: Palantir's senior leadership team is entirely male (TechCrunch). A company selling itself as the trustworthy guardian of enterprise AI is making that pitch with a leadership team that has no gender diversity. That may not affect the technical merits of what they argue, but it sits uneasily alongside the governance Karp is promising.

Palantir's Q1 2026 financial filing also reveals that the company owns shares in publicly traded companies, valued at market price each reporting period (Palantir IR). That detail adds a layer of financial complexity that has nothing to do with the company's core business of selling software.

The real question Karp is raising, stripped of the rhetoric, is practical: when you send your data to someone else's AI, what leaves your control, who can learn from it, and how hard will it be to switch away later? Palantir's growth suggests a significant part of the market is already deciding to keep control rather than hand it to the AI labs. Whether that is the right choice depends on what a company is trying to do. But the trade-off — convenience now versus keeping your own knowledge — is one every business adopting AI at scale is now working through.