Anthropic Gains Ground
Enterprise AI spend is tilting toward Anthropic, with Reuters reporting the company may have narrowed the revenue gap with OpenAI as firms prioritize deployable products over flagship model claims. That shift is linked to Anthropic’s push into managed agents and easier integration paths, and has attracted high-profile investor commentary calling it a competitive threat to established analytics and data-ops vendors. ( )
Reuters reported on April 8 that Anthropic may have pulled close to OpenAI in the revenue race, a sharp turn in a market that looked lopsided a year ago. One recent estimate put Anthropic at a $30 billion annualized revenue run rate, while OpenAI was around $24 billion to $25 billion. (reuters.com) (neowin.net) The surprise is not that Anthropic built a strong model. The surprise is that companies buying artificial intelligence software seem to care more about getting it wired into daily work than about who wins benchmark charts. (reuters.com) (ramp.com) Ramp, which tracks business purchasing through corporate card and bill-pay data, said in its March 2026 update that nearly one in four businesses on its platform now pays for Claude, Anthropic’s assistant. A year earlier, that figure was one in 25, while OpenAI’s adoption fell 1.5% in the same month, Ramp said. (ramp.com) That kind of shift usually happens when a product stops being a demo and starts being plumbing. In February, Anthropic held an “Enterprise Agents” launch focused on finance, legal, sales, and product teams using Claude inside existing workflows instead of in a separate chatbot window. (anthropic.com) Anthropic then pushed further on April 8 with Managed Agents, a hosted service for long-running tasks. The pitch was simple: let companies use Claude for multi-step work without having to build and maintain all the scaffolding themselves. (anthropic.com) (wired.com) That scaffolding is the unglamorous part of artificial intelligence at work. It includes the software that keeps an agent running, connects it to internal tools, watches for failures, and scales it when more employees start using it. (anthropic.com) Anthropic’s bet is that if it owns more of that layer, it stops being just a model vendor and starts looking more like the operating system for office tasks. That is why its February push included plug-ins and workflow tools aimed at functions like finance, engineering, and design. (anthropic.com) (techcrunch.com) That also explains why investors are suddenly comparing Anthropic to software companies outside the model race. Business Insider reported that Michael Burry argued Anthropic was “eating Palantir’s lunch,” framing Claude’s enterprise traction as a threat to vendors that make money from data pipelines, analytics, and decision tools. (businessinsider.com) Palantir sells software that helps governments and corporations organize data and run operations. If Anthropic can drop agents into those same workflows and make them useful with less custom work, some of that budget can move away from classic data-ops software and toward the model company itself. (businessinsider.com) (anthropic.com) OpenAI is still huge, and Reuters said both companies are racing toward possible public offerings with different strengths. But April’s signal was that enterprise buyers may be rewarding the company that makes artificial intelligence easier to deploy, not just harder to beat on a leaderboard. (reuters.com) (ramp.com)
