OpenAI–Anthropic rivalry heats

OpenAI and Anthropic are competing in public more directly now, as leaked memos, revenue figures and product launches turn their rivalry into a visible fight. (techcrunch.com) The sharpest recent exchange came after Anthropic chief executive Dario Amodei wrote in a March 2026 staff memo that OpenAI’s Pentagon messaging was “safety theater,” according to reporting by The Information and TechCrunch. OpenAI and Anthropic were sparring over a Defense Department relationship after Anthropic failed to reach an agreement with the agency over unrestricted access to its systems. (techcrunch.com) At the same time, investors and industry watchers have started using annualized revenue as a running score. Bloomberg reported Anthropic passed a $19 billion run rate in early March 2026 and said on April 6 that Anthropic had topped a $30 billion run rate, up from $9 billion at the end of 2025. (bloomberg.com 1) (bloomberg.com 2) OpenAI has been putting up its own growth numbers. Bloomberg reported in February 2026 that OpenAI chief financial officer Sarah Friar said annualized revenue topped $20 billion in 2025, up from about $6 billion a year earlier. (bloomberg.com) The products behind those numbers are increasingly aimed at the same buyer: software teams that want artificial intelligence systems to write, review and manage code across many files at once. Anthropic says Claude Opus 4.6 improves coding, debugging, tool use and long-running agent tasks, while OpenAI says GPT-5.4 is built for professional work with coding, computer use, tool search and a 1 million-token context window. (anthropic.com) (openai.com) A context window is the amount of material a model can keep in view at one time, like how many pages can stay open on a desk before work slows down. Anthropic says Opus 4.6 includes a 1 million-token context window in beta, and OpenAI says GPT-5.4 also offers a 1 million-token context window. (anthropic.com) (openai.com) The contest has also moved from chatbots to full coding workstations. OpenAI’s Codex app, introduced in February 2026, is designed for multiple agents, parallel workflows and long-running tasks, and Anthropic’s Claude Code business has grown fast enough that The Information reported its annualized revenue rose from $1 billion in December to $2.5 billion in February. (openai.com) (theinformation.com) Anthropic has kept adding coding features around that push. Its Claude Code update called Code Review uses multiple agents to look for bugs, and Anthropic says Sonnet 4.6 is tuned for planning, implementation, debugging, maintenance and large codebase refactors. (theverge.com) (anthropic.com) OpenAI has matched that cadence with frequent Codex updates. The Codex changelog shows releases on April 7, April 10 and April 11, 2026, as the company keeps revising model availability and command-line tools for developers. (developers.openai.com) The money behind the rivalry is also expanding. Bloomberg reported on March 31 that OpenAI completed a $122 billion financing round at an $852 billion valuation, while Bloomberg reported Anthropic completed a tender offer in April as employees held onto shares while the company’s run-rate revenue climbed. (bloomberg.com 1) (bloomberg.com 2) What used to look like a model race now looks more like a contest over who can own the developer workflow, sell the most enterprise capacity and turn that demand into durable revenue fastest. The public messaging has gotten louder because the business stakes are now large enough to measure in tens of billions of dollars. (openai.com) (bloomberg.com)