Behind Palantir’s cry of “beyond endurance”: fear among other companies that large-model “winners take all”

Palantir CEO Alex Karp’s fierce remarks have brought a long-simmering contradiction in the tech industry into the spotlight—large AI labs are building up momentum by accumulating customer data and decision-making leverage, while traditional enterprises are becoming increasingly worried that they could turn into “value contributors” in this AI wave.

Over the past two weeks, Karp first launched a blistering nearly 20-minute attack on CNBC. He directly accused AI labs of overstating their capabilities, pricing Token tokens too high, and claimed that every major enterprise customer he has encountered is “burning with anger” about it.

He was then followed by Palantir publishing a white paper titled “Institutional Sovereignty in the Age of AI,” which lays out 15 recommendations for enterprises and governments, aimed at preventing AI giants such as OpenAI and Anthropic from eroding its core data. These two moves quickly sparked widespread discussion across the tech community.

At the heart of this debate is a question that more and more people are loudly raising: in the AI era, whose value is ultimately captured—by the enterprises that deploy AI, or by the labs that develop the underlying models?

This issue is not only about reshaping the business landscape; it has also spilled into policy wrangling and geopolitical competition, posing a direct threat to the valuations of traditional software vendors.

Not just Karp speaking up

Karp himself admits that his position is not neutral.

Palantir’s core product is the middle layer built on top of foundation models—connecting AI with enterprise customers. This positioning gives it direct commercial interests in the power struggle between enterprises and AI labs.

In response to accusations from the outside world that he is venting his emotions, Karp said:

No. It’s the voice of the U.S. business community, delivered through me.

Notably, Karp is not the only tech executive warning about this lopsided situation.

Microsoft CEO Satya Nadella has recently written an article and repeatedly expressed similar concerns publicly. His core worry is: whether the “learning outcomes” accumulated by enterprises from using AI models can truly be retained within their own organizations.

This month, Nadella said at an event at Stanford University:

If you’re just a consumer of a foundation model, I’m not sure how you can retain enterprise value, let alone create it.

The logic of “cannibalization” by AI labs

Karp’s criticisms touched a deeper anxiety within the tech industry.

Former White House point person for AI affairs David Sacks immediately echoed the view on social media and turned his sights directly on Anthropic. Sacks wrote:

Anthropic has rolled out Claude Science, Claude Security, Claude Legal, and Claude Code—each product is moving directly into the domains previously served by companies that built applications based on its model.

Sacks added further:

The pattern is always the same: observe where value is being created, then move in directly. Lead the model layer first, and then use that position to take over the most profitable vertical markets.

This “observe-copy-expand” path leaves many companies that rely on large-model APIs feeling uneasy. For these companies, contributing data and use cases to AI labs may be providing ammunition for competitors to enter the market.

Neither OpenAI nor Anthropic has issued a public response to Karp’s criticism. Both companies’ current policies state that enterprise customer data will not be used to train their models.

An insider at an AI lab dismissed the matter, saying:

It’s foolish to respond to a Karp-style performance—he’s just putting his own interests on display.

The winner hasn’t been determined yet

The deeper backdrop to this debate is the industry’s high uncertainty about who ultimately owns AI value.

Wall Street Tech noted that this Thursday, some media reported that Starbucks is using AI to replace software previously procured from Microsoft and IBM, and immediately afterward, the two companies’ stock prices came under pressure.

This case is seen as a snapshot of how AI is accelerating the reshaping of the enterprise software landscape. Analysts pointed out that in the time it takes to make a cup of coffee, the winners and losers in the AI era could swap places at any moment. This once again underscores a harsh reality: today’s large tech companies may not be able to lock in future leadership positions.

Meanwhile, Meta announced over the weekend that it is rolling out an updated AI model and introducing a paid tier. According to Bloomberg News, during an interview, Meta CEO Zuckerberg clearly said he sees opportunities to compete on pricing:

The pricing of some other labs is extremely extreme, with very high profit margins. We believe that we can provide cutting-edge or high-quality intelligent services at a much more affordable price.

This statement further intensifies competitive pressure in the foundation model market, and also indirectly supports Karp’s criticism that AI labs’ pricing is inflated—it’s not without basis.

Risk warning and disclaimer

        There are risks in the market; invest with caution. This article does not constitute personal investment advice, and it does not take into account any particular user’s special investment objectives, financial situation, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article align with their specific circumstances. Investing based on this is at your own risk.