|
Anthropic and OpenAI have leaned heavily into serving businesses with AI agents that automate white collar tasks and software that helps businesses manage these agents. The two AI developers are even competing to sign up private equity firms to create joint ventures that would sell AI to the PE firms’ portfolio companies, my colleagues scooped last week. But Palantir, a 22-year-old software firm, is arguably in a better position to deliver AI agents and related software to businesses. Palantir offers so-called end-to-end software, helping customers link and organize data they store in systems such as Snowflake, Salesforce, and SAP, and to develop AI agents that can utilize that data. That’s no small feat, given that the traditional enterprise apps don’t love the idea of being treated as merely a source of data that other companies’ AI agents can tap. Many of these traditional enterprise firms say they want to charge money for that privilege. Palantir’s advantage goes beyond software. Seemingly everyone who is anyone in enterprise software has explicitly or implicitly acknowledged the benefits of Palantir’s “forward deployed engineer” model in the AI era, given the challenges that still remain in making AI work right, and safely, at a large scale. FDEs are essentially in-house consultants that work with customers to build custom versions of Palantir software, known as Foundry, so it can work better for their needs. (In case you’re counting, the list of FDE admirers includes Salesforce, ServiceNow and Snowflake, plus OpenAI.) Palantir doesn’t develop its own large language models the way OpenAI and Anthropic do. But its software enables customers to use the best proprietary and open source models to do everything from automated monitoring of supply chains, workforces and inventory to providing data tools for managing sales. In other words, Palantir offers so-called data integration services like what Snowflake has; consulting services that everyone is trying to emulate; and an “AI action layer” similar to what OpenAI and Anthropic (and many other firms) want to sell to businesses that want to use AI agents. For instance, Centrus Energy, which operates the only U.S.-owned uranium enrichment facility, is beginning to use AI agents developed with Palantir’s software and Anthropic models to help make changes to employee shifts if staffing issues come up; find alternative suppliers if a shipment is delayed; and create regulatory compliance reports to submit to the Nuclear Regulatory Commission, according to Patrick Brown, senior vice president of field operations at Centrus. The company estimates that using the software could save his company $300 million, though it didn’t explain how it calculated that figure or the time frame it pertains to. “Agents aren’t replacing our people, they’re extending their capabilities… so then the experts can put their attention on our most challenging problems,” Brown said in a presentation at Palantir’s conference for customers last week. Reduction In Force That may be, but in some of our conversations with Palantir customers, the main cost savings they said they enjoy from unifying all of their data sources is cutting employees that were previously responsible for managing different piles of data. That has led to serious savings, some of them said. Palantir’s work with Ondas, a drone provider, and World View, an aerospace firm that operates stratospheric balloons, epitomize its advantage over other enterprise apps that sell AI products. Both firms provide surveillance services to companies and governments. In one example, they say they are planning to use AI agents with Palantir’s software to automatically trigger a work order in pipeline operators’ enterprise resource planning apps in the event of a methane leak. In other words, the Palantir system sits on top of and accesses the enterprise apps these businesses use. Eventually they hope to replace traditional enterprise apps with a Palantir-led system that harnesses their stored data. “I'd like to have a pure Palantir enabled solution without having to have other SaaS tools that we maintain,” said World View CEO Ryan Hartman. While government contracts in the U.S. and abroad still account for the majority of Palantir’s revenue, the company’s commercial segment has exploded in the past year as businesses scrambled to use AI into their products and workflows. Palantir’s revenue from selling its software to businesses in the U.S. grew a staggering 109% last year after jumping 54% in 2024. Overall commercial revenue, including sales to firms in the rest of the world, climbed 60% last year, 30 percentage points faster than its growth from that segment in 2024. For comparison, enterprise software firms including Salesforce, SAP, and Adobe each grew revenues around 10% last year. In some senses, OpenAI has a bigger enterprise business than Palantir’s, thanks in part to ChatGPT subscriptions to professionals and organizations. But OpenAI leaders have to rapidly expand its enterprise business to justify the company’s $800 billion private valuation, especially as Anthropic revenue surges from selling models to businesses. (Palantir’s public market capitalization is around $370 billion.) Earlier this year, OpenAI announced Frontier, a product to help businesses create multiple “AI coworkers” and assign them different tasks that involve pulling data from various enterprise applications. An OpenAI spokesperson said they believe Frontier is different from Palantir’s Foundry software because it offers pre-built or custom agents. By all accounts, though, Foundry does this too. The spokesperson said OpenAI customers could connect their Foundry software to Frontier, meaning that they don’t have to pick one or the other. Why OpenAI and Anthropic Are Courting Private Equity Firms As we noted earlier, Anthropic is in talks with Blackstone and other private equity firms about creating a joint venture to sell the Claude maker’s AI to companies funded by the investment firms. Now OpenAI says it is talking with a different set of private equity firms, namely TPG and Bain, about forming a joint venture of its own. It’s not hard to view these discussions as the opening chess moves in what promises to be an intense battle between these firms for large corporate AI deals. Both of them already work with consulting firms like Accenture and Capgemini to sell AI to enterprises. Teaming up with private equity is a logical next step, as those PE firms control technology spending decisions for dozens of software providers that should be the biggest buyers of AI. Based on our discussions, OpenAI and Anthropic have much to learn about the intricacies of selling to large customers. Selling AI to portfolio companies of PE firms could help them gain experience and be better positioned to do business with other types of customers that, unlike the PE-owned ones, are entirely in charge of their own tech spending.—Kevin McLaughlin
| 
