Palantir Posts $1.9B Quarter as Karp Calls AI Labs ‘Marxist’

Palantir reported second-quarter 2026 revenue of $1.9 billion, a 93% jump over the same quarter last year, and $1.1 billion in profit. CEO Alex Karp used the shareholder letter to go further than the numbers, accusing AI frontier labs of effectively extracting enterprise intellectual property through paid token usage. He called the dynamic "Marxist," arguing that a small group of AI builders intends to capture the means of production from the very partners paying them.

What happened

Metric	Value
Q2 2026 revenue	$1.9 billion
Revenue growth (year-on-year)	93%
Q2 2026 profit	$1.1 billion
Prior year comparison	More profit in one quarter than total revenue in Q2 2025

Palantir published its Q2 2026 results on August 3, and the numbers were hard to argue with. Revenue came in at $1.9 billion, up 93% over Q2 2025. Profit reached $1.1 billion, which Karp noted in the shareholder letter exceeded Palantir’s total revenue from the same period the previous year.

Alongside the financial results, CEO Alex Karp, who holds a PhD in social theory, used the letter and the analyst call to lay out a pointed critique of the AI industry. He wrote that “others, including many of those building large language models, intend, knowingly or otherwise, to capture the means of production of their purported partners.”

What exactly is Karp’s argument?

The core claim is that when enterprises pay AI labs for token usage, they are effectively funding those labs to absorb enterprise knowledge, workflows, and expertise into the lab’s own models. Karp described it on the analyst call as “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.”

He argued the motivation is not purely commercial. In his framing, AI labs believe they are morally superior and are therefore entitled to expand into enterprise territory, competing directly with the customers who fund them. The shareholder letter described this tension as having “Marxist overtones and undertones.”

Karp positioned Palantir differently: its software is described as model-agnostic, meaning it is not tied to a single AI provider, and it gives organisations control over their “AI exhaust,” his term for the prompts, context, and orchestration data that flows through AI systems. That data stays with the customer rather than feeding a lab’s training pipeline.

Why it matters

Karp is not alone in raising this concern. According to TechCrunch, Microsoft CEO Satya Nadella has made similar observations. The underlying anxiety is real: a growing list of companies have paid for or partnered with Anthropic and OpenAI, only to watch those same labs launch products in design, healthcare, legal, and drug discovery that compete with those partners.

For any business currently spending on AI APIs, the question Karp is surfacing is worth taking seriously. Are you training a competitor? The answer depends on the provider’s terms, how your prompts and outputs are used, and whether the vendor has a financial incentive to productise what it learns from your usage.

Our coverage of Amazon’s AI budget overruns showed how quickly token costs compound at scale. Add the IP-leakage concern on top of that, and the case for scrutinising AI vendor contracts becomes stronger.

Our take

Karp has a flair for provocation, and calling the AI industry “Marxist” is designed to get attention. The rhetoric about “vegetables” and “war fighters” on the analyst call is pure defense-tech theater. But strip the jargon away and the structural point is legitimate: some AI labs are being paid by enterprises while simultaneously building products that compete with those enterprises. That is a real conflict of interest, not a philosophical abstraction.

The irony is that Palantir’s record quarter happened precisely because AI adoption is booming, including at the same enterprises Karp says are being exploited. The market is growing fast enough that everyone is winning for now. The question is what happens when growth slows and the labs need new revenue lines. That is when the IP argument gets sharper.

If you are evaluating AI integration for your business, the vendor’s data use policy deserves the same scrutiny as the price. Not because every lab is acting in bad faith, but because the incentives Karp describes are structural, not personal.

What to do about it

1. Read the data use terms of any AI API you are currently paying for, specifically what the vendor can do with your prompts and outputs.
2. Ask whether your AI vendor competes, or plans to compete, in your industry vertical.
3. Evaluate model-agnostic middleware or on-premise options if your prompts contain sensitive IP, customer data, or proprietary processes.
4. Track your token spend against business outcomes so you can quantify what you are paying and what you are getting. Our workflow automation work often surfaces this kind of hidden cost early.

The honest takeaway: Karp’s language is loud, but the data governance question underneath it is one every enterprise AI buyer should answer before signing the next contract.

Source: TechCrunch · AI