Palantir just posted one of the loudest quarters in enterprise software, and its CEO used the moment to call the AI industry Marxist. According to TechCrunch AI, Alex Karp wrote in Palantir’s shareholder letter that frontier AI labs are too untrustworthy for enterprises, arguing that model builders “intend, knowingly or otherwise, to capture the means of production of their purported partners.” It’s a provocative line. It’s also mostly wrong about what’s actually happening in the market.
Let me start with the number that matters. Palantir reported $1.9 billion in revenue for Q2, up 93% year over year, plus $1.1 billion in profit. As Karp himself put it, that’s “more profit in a single quarter than we did in total revenue in the same period the year before.” The skyrocketing use of AI didn’t corner Palantir. It supercharged them. So the framing of AI labs as an existential threat sits awkwardly next to results that prove the opposite.
The real fear Karp is naming
Strip away the jargon and there’s a legitimate worry underneath. Karp’s argument, as detailed in TechCrunch AI, is that when you pipe your proprietary data and workflows into someone else’s model, you’re training your future competitor. His words on the analyst call were blunt: “You are paying for the right for them to migrate your IP, your know-how, your expertise to their model, so that they can build a competitive business that doesn’t require your business or people.”
That concern isn’t fringe. Microsoft CEO Satya Nadella has voiced versions of it too. And the pattern is real: companies partnered with or paid Anthropic and OpenAI, then watched those same labs launch products in design, healthcare, legal, and drug discovery. When your vendor becomes your rival, the partnership math changes.
Where the argument breaks down
Here’s what stands out. The “they’re colonizing your enterprise” story assumes a zero-sum market. Palantir’s own quarter shows it isn’t. AI is growing so fast, and shifting so quickly, that there’s clearly room for labs, platforms, and integrators to all win at once. TechCrunch AI put it plainly: none of these companies are economic villains or heroes, any more than other for-profit companies are.
Calling competitors Marxist while running a defense-tech company that sells to governments is also a strange rhetorical choice. It’s positioning dressed up as philosophy. Karp studied social theory, and he knows exactly what the word does: it turns a business-model disagreement into a moral crusade. That’s marketing, not analysis.
What this means for practitioners
The useful signal here isn’t the insult. It’s the data-control question every buyer should now be asking. Palantir’s actual pitch is model-agnostic software that lets organizations keep their data, prompts, orchestration, and context in-house. That’s a real differentiator, and it’s worth taking seriously even if you find the Marxist framing overheated.
A few practical takeaways:
- Read your AI contracts for training rights. Know exactly whether your prompts, outputs, and fine-tuning data can feed a vendor’s general models. This is the single most important clause.
- Separate the model from the workflow. Model-agnostic architecture lets you swap providers as prices and capabilities shift. Lock-in is the actual risk, not ideology.
- Watch where your vendor is expanding. If a lab you rely on is launching products in your industry, treat that as a competitive signal and plan a fallback.
- Don’t buy the binary. The market isn’t labs-versus-everyone. Most enterprises will run a mix of foundation models and control-focused platforms for years.
Why it matters now
Karp’s letter lands as the industry debates a genuine tension: how much should you trust the company that both sells you intelligence and competes for your customers? That question is real and getting sharper. But the answer isn’t to pick a moral side. It’s to structure deals so you keep control of what makes your business yours.
Palantir’s quarter is the strongest evidence against Karp’s own doom framing. The AI market is expanding fast enough for many winners, and the smart move is architecture that keeps your options open, not rhetoric that picks a villain. More details are available in the original TechCrunch AI report.