From a traditional investor’s perspective, this was not an entirely trustworthy business model, as it relied on government contracts. Palantir therefore sought to convince investors that its services could also be applied in the private sector to analyze complex corporate problems.

With the advent of generative AI, the market’s understanding of Palantir changed. The allure of intelligence work faded. Suddenly, Palantir looked more like a conventional provider of data processing for AI. Its shares skyrocketed from $20 to an all-time high of $198, a gain of nearly 900%. No wonder Palantir won over investors and gained enormous popularity after turning many of them into millionaires.

It seemed that the story was not going to end anytime soon. The shares were severely overvalued and ranked among the most expensive on the market.

The turning point came in 2026. Palantir fell as low as $106 in June, representing an annual decline of more than 40%. No one doubted that the shares were undergoing a correction. Even at that price, a price-to-earnings ratio of around 100 left the valuation exceptionally high. The bottom was nowhere in sight.

Politics Is Factored Back Into the Price

The decline in Palantir’s popularity among investors had several deeper causes beyond its expensive fundamentals. The first was the company’s significant involvement with the White House. Palantir’s services were used in Venezuela, Iran and the Gaza Strip, as well as in the deportation of migrants. It was not merely that the company supplied the technology, but that it also strongly defended these policies. Neither Palantir’s management nor its employees can plausibly claim that providing technology for such purposes is politically neutral.

The company’s objectivity was supposed to rest on a balance between Thiel and Karp. But Karp underwent a political about-face. As recently as 2024, he said he would vote against Trump. Since then, Karp has declared Palantir to be “anti-woke” and, above all, has taken pride in the fact that its technology is used to kill America’s enemies.

Of course, the question remains: who defines and determines who is an enemy of the United States? That shift has had consequences. The company must now contend with the departure of engineers who do not share these values. In the long run, staff can certainly be replaced. Yet internal tensions are not the only concern for investors.

The second challenge is far more serious, as it threatens the very essence of the company. Michael Burry articulated the objection well when he noted that Anthropic would soon be eating Palantir's lunch. The objection is almost childishly simple, but it goes to the heart of Palantir’s business model: what, exactly, will Palantir know that the language models themselves do not?

Karp responded by saying that there will always be a layer between the data and the language model. He is right about that, but it misses the point. This layer will exist, yet companies are already developing alternatives that will compete directly with Palantir at very low prices. Language model developers could also acquire a company offering such a solution and integrate it themselves. Karp's answer was therefore not particularly convincing.

And so he pressed the point further, warning of the risk that enormous power could become concentrated in the hands of the companies managing AI models. Here, too, he is right: the risk is real and will only grow as these companies become more powerful. But it is a double-edged sword. How much oversight does Palantir itself have over the data it handles?

Is this not like the thief shouting, “Catch the thief”? AI sovereignty is important, but will Palantir really ensure it? If so, we would not be seeing both the German and French intelligence agencies walk away from the company. The latter recently opted for the domestic French company ChapsVision precisely because of concerns over data control and sensitivity.

Great Numbers, an Uncomfortable Question

Palantir nevertheless managed to fend off these attacks on the very core of its business with impressive second-quarter 2026 results. The company raised its full-year revenue outlook to approximately $8.15bn after reporting staggering growth for the quarter.

Palantir’s revenue rose 93% year on year to $1.94bn, as both its government and commercial businesses accelerated in the United States. Profit increased 55% to $1.06bn. The company therefore continues to expand at a remarkable pace while maintaining massive margins.

The shares jumped by tens of percent on the news. Even this significant gain, though, was not enough to put them in positive territory for the year. Palantir remains down about 7%.

The rally undoubtedly pleased shareholders, but it has once again put Palantir in a bind by pushing its valuation to extremely high levels. That is a major risk.

It is not that the company is unable to deliver strong financial results or solid fundamentals. The greater concern stems from questions about the long-term sustainability of its business model. Given that the current valuation already assumes an almost flawless future trajectory, this is a significantly bigger problem than Palantir’s present financial performance.