Palantir, Microsoft’s biggest warning gets real for AI stocks

In July, Microsoft (MSFT) CEO Satya Nadella surprised AI stock investors with an uncomfortable way to think about what they were buying into.

As I wrote for TheStreet in “Microsoft CEO adds fuel to Palantir CEO’s AI warning,” Nadella argued that companies weren’t merely paying for tokens. Moreover, they were feeding outside models the prompts, corrections, and institutional knowledge that eventually make those systems useful for their users.

“You essentially pay for intelligence twice, once with money, and again with something even more valuable,” Nadella wrote in his July essay.

Days later, that warning became a central part of Palantir’s bull case. Now, two months on, Palantir and Nvidia, two of the biggest names in the AI space, are starting to act on it.

Palantir CEO Alex Karp pushed the argument further, which I covered in “Palantir CEO escalates Microsoft’s AI warning.”

During the company’s August earnings call, Karp warned businesses that they could potentially end up aiding an outside AI provider build “a competitive business that doesn’t require your business or your people.”

That sovereign-AI thesis has become a critical part of Palantir’s bull case, alongside powerful commercial growth that has driven the stock up around 51% in August alone, as Jim Cramer recently explained.

That said, the debate’s now moving beyond speeches. Reuters reports that Palantir and Nvidia are now taking concrete steps around how frontier AI models interact with sensitive corporate information.

Palantir is drawing a harder line on AI data control

The Palantir-Anthropic tension is becoming a lot less about abstract AI safety and more about who gets to keep data, for how long, and under what terms.

Reuters, citing The Information, says Palantir pressed Anthropic for an irrevocable zero-data-retention guarantee before allowing Anthropic models to be made accessible through Palantir’s software. 

Clearly, Palantir is looking for contractual certainty that sensitive prompts usage data and outputs cannot later be retained, repurposed, or exposed through policy changes.

More Wall Street:

• Wall Street’s AI trade faces its biggest valuation test
• The next Wall Street shift is already underway
• Wall Street sends strong 4-word verdict on the stock market

The dispute between the two businesses has intensified after Anthropic switched up the policy for its Fable model in June, which allowed companies to retain usage logs for 30 days to defend it against attacks.

That change led to a customer pushback. At the same time, OpenAI faced similar scrutiny over allegations that user data could potentially be used in model-related work, according to the report.

More Wall Street:

• Wall Street sends strong 4-word verdict on the stock market
• Wall Street’s AI trade faces its biggest valuation test
• The next Wall Street shift is already underway

Moreover, Palantir isn’t alone in tightening controls. Nvidia has reportedly limited Anthropic models to less-sensitive tasks while using its own Nemotron models for internal work. At the same time, Booz Allen has barred Anthropic’s commercial model from specific proprietary cybersecurity activities.

Overall, for Palantir, that solidifies the case that AI sovereignty is perhaps becoming an operating requirement.

Palantir’s valuation leaves little room for the AI story to break

For Palantir stock investors, the Anthropic dispute is strategically useful, but the pricing overhang is still a big issue for them to contend with.

Palantir stock has surged over 50% in August after another blowout quarter, according to Yahoo Finance, before giving away part of those gains in September.

Even after the pullback, though, the company is trading at a market cap of nearly $400 billion and trades around 143-times trailing earnings and 75 to 82 times forward earnings, depending on the estimate set as per Seeking Alpha data.

Clearly those numbers leave very little room for any missteps. So far, Palantir has delivered though.

Q2 sales surged 93% to $1.94 billion, U.S. commercial revenues were up 149%, and adjusted operating margins shot up to 62%. Moreover, its management expects nearly $8.15 billion in 2026 sales, which implies 82% growth. 

Nevertheless, the latest Anthropic development strengthens part of Palantir’s growth thesis.

If clients insist on similar protections in droves, then Palantir’s role becomes a lot more valuable. The thing with Palantir is that it isn’t just selling access to AI, but it’s looking to become the layer that effectively determines how models interact with sensitive corporate data. 

That also creates a fascinating counterpoint to the emerging AI slowdown narrative. Calls from Anthropic CEO Dario Amodei and others to slow down frontier-model development rattled AI stocks as investors immediately translate slower capability gains into sluggish infrastructure spending and lower future growth.

For Palantir, a more circumspect AI environment could cut both ways.  

A genuine collapse in enterprise AI spending will be painful for Palantir, as its valuation assumes years of strong growth. However, greater scrutiny around safety, data ownership, and model governance will strengthen Palantir’s demand. 

Enterprises might become a lot less interested in blindly opting for AI models and more focused on controlling how models touch proprietary information. The Anthropic development is important because it adds real-world valuation for the sovereign-AI problem Karp has been selling investors.

Related: Elon Musk sends strong signal for SpaceX, Nvidia stocks