Palantir just reported its best quarter ever — $1.9 billion in revenue, up 93% year-over-year, with $1.1 billion in profit. U.S. commercial revenue grew 149% year-over-year. The company raised its full-year revenue guidance to 82% growth. By any measure, it was a blowout.

Then CEO Alex Karp stepped to the microphone and called the AI industry “Marxist.”

It sounds like a headline grab. But if you actually listen to what he’s saying, there’s something real in there for any business currently navigating the AI vendor landscape.

What Karp Actually Said

Karp’s argument is about control. His position: the major AI frontier labs — OpenAI, Anthropic, Google DeepMind — build powerful models, but their model-as-a-service approach means enterprises are fundamentally handing over their most sensitive data and workflows to organizations whose primary interest is training the next generation of models.

He frames this as a form of intellectual property collectivism. You put your proprietary data in, their model learns, and the shared model that comes out is no longer just yours. Whether or not you agree with the political framing, the underlying concern is something enterprise IT and legal teams think about constantly: data leakage, model contamination, and loss of competitive advantage.

Palantir’s pitch — and the reason Karp can be this aggressive — is that Palantir’s software sits on top of models but keeps client data isolated. The AIP platform is built around the idea of AI sovereignty: enterprises get the capability without surrendering the data.

The Numbers Behind the Argument

What makes Karp’s commentary more than just provocative noise is that the financials back it. When U.S. commercial revenue is growing 149% year-over-year, something is clearly resonating with enterprise buyers.

That growth is happening against a backdrop where most enterprise AI projects still fail to make it from pilot to production. CIOs are simultaneously being pressured to ship AI results and increasingly nervous about what they’re signing up for with hyperscaler AI services. Palantir’s pitch — operationally proven, data-sovereign, built for regulated industries — is landing.

The guidance raise to 82% full-year growth suggests this isn’t a one-quarter event. It’s compounding.

What Enterprises Are Actually Buying

The practical story here is about what enterprise AI buyers want in 2026, and it’s getting clearer:

Operational control. They want to know exactly what data their AI is touching, where it goes, and who can access the outputs. Chatbot wrappers built on shared API infrastructure do not satisfy this requirement for most large organizations.

Domain-specific capability. Generic models get them to 70%. The final 30% — the part where AI actually generates enterprise value — requires context about their specific data, workflows, and decision-making. Palantir’s software-plus-service model addresses this.

Accountability. As the EU AI Act enforcement era begins (August 2, 2026 enforcement went live), enterprises in regulated industries need audit trails, explainability, and documented human oversight. Building that on top of a consumer-grade API is genuinely difficult.

What This Means for Business

Karp’s “Marxist” framing will generate plenty of LinkedIn hot takes and industry commentary. Ignore the label. Focus on what the underlying question means for your organization:

When you adopt an AI vendor’s model via API, what are you actually agreeing to? Who trains on your data? What happens to the patterns your business’s information contributes to the broader model?

These aren’t hypothetical questions. They’re the questions your legal team, your board, and increasingly your enterprise customers will ask. The businesses that can answer them clearly — because they chose vendors or deployment models that preserve data sovereignty — are going to have an easier time scaling AI internally and selling AI-enhanced services externally.

Palantir’s quarter doesn’t make them the right choice for every enterprise. But Karp’s willingness to make this argument publicly, backed by numbers that show the market is rewarding it, is a signal about where enterprise AI procurement is heading.

The era of “just use the API and figure out governance later” is ending. Enterprises that plan their AI stack with data control at the center will have a structural advantage over those that have to unwind messy vendor dependencies later.

The EDNA Perspective

At Enterprise DNA, we’ve seen this pattern play out in the data world for years. The organizations that built genuine data capability in-house — clear ownership, documented lineage, governed access — were the ones who could actually move fast when AI came along. The ones who outsourced everything to black-box platforms are now scrambling to understand what data they even have.

AI is following the same arc. The governance decisions enterprises make today, when the AI market is still maturing and vendor lock-in is relatively soft, will determine how much leverage they have in three years when AI is infrastructure.

Palantir’s $1.1 billion profit quarter is a market signal. What it’s signaling is that enterprises are starting to pay for AI with data sovereignty built in — not just AI capability.