|
Software companies and their shareholders have watched the recent growth of Anthropic and OpenAI with unease. But the mood is shifting as customers complain about AI costs, and the old guard smells blood in the water. Last week, Palantir CEO Alex Karp and his lieutenants told his customers at an event outside San Francisco that they’re foolish to do business directly with AI firms rather than with intermediaries like his firm, which he says can help them run AI more effectively. “You‘ll go to a large language model company and learn that they don’t care about you at all,” he said. “You‘re gonna go home feeling poorer and less safe, and you buy the product, and you’re gonna pay a lot in tokens, and it's gonna be very hard to understand how it helps you.” (Spokespeople for Anthropic and OpenAI did not provide a response.) His comments come after executives at Uber, ServiceNow and Snowflake said they’re worried about runaway costs of using the latest AI tools, including Anthropic’s Claude Code. Aaron and I reported last week that some customers are finding ways to reduce costs already by using model routers and capping usage among individual employees. Software firms like Karp’s are also trying to seize the moment. They say they can play an important role in giving customers AI model options from multiple providers so that no one vendor gains leverage over them. Palantir’s head of its commercial business, Ted Mabrey, said there’s been an “explosion” in the number of customers concerned about and looking to moderate AI costs linked to token usage in the last six weeks since The Information quoted Uber CTO Praveen Neppalli Naga saying the rideshare firm blew through its full-year AI budget in the first few months of 2026 as Claude Code usage surged. To be sure, customers have complained to me that Palantir’s databases and other software are also too expensive, as they can cost tens of millions of dollars annually at large firms. (Still, Mabrey and other Palantir customers told me the software is worth the cost.) Karp said Palantir’s software and its team of specialist consultants, known as forward deployed engineers, give customers the ability to solve a given problem and “own the means of production.” FDEs develop customized AI applications powered by models from different providers, including Anthropic and OpenAI, and the customers can see how much they’re spending on tasks handled by each model powering the apps so they can easily pinpoint cost overruns. Palantir’s FDE model has been the envy of other software providers, and even OpenAI and Anthropic are hiring people with such titles. Many software firms are also pitching new products that, similar to Palantir’s software called Foundry and AIP, are essentially management dashboards to help enterprises manage their AI tools. “Everyone’s copying us,” Karp told me in passing after I mentioned the FDE hiring boom. “Quote me on that.” Some Palantir customers seem to agree with Karp’s pitch. Law firm Kirkland & Ellis, for instance, worked with Palantir to develop an AI app that helps its lawyers draft transaction documents for private fundraising deals in minutes rather than days, Erica Berthou, an executive at the law firm, said at the Palantir event. Still, Anthropic’s revenue is on track to easily surpass Palantir’s this year. Anthropic, founded in 2021, has projected $11 billion in revenue in the current quarter alone, while Palantir, which is 23 years old, has projected roughly $7.7 billion for the full calendar year, up 70% from 2025. If not for Anthropic’s rise, Palantir’s growth rate would be considered other-worldly!
| 
