Anthropic's rapid commercial growth

Anthropic says demand for Claude has pushed its annualized revenue past $30 billion, up from about $9 billion at the end of 2025. (anthropic.com) The company said on April 6 that more than 1,000 business customers now spend over $1 million a year on Claude, up from more than 500 when Anthropic announced its Series G fundraise in February. (anthropic.com 1) (anthropic.com 2) Anthropic is adding supply to match that growth. It said it expanded its partnership with Google and Broadcom for multiple gigawatts of next-generation compute and called it its biggest compute commitment so far. (anthropic.com) Claude is Anthropic’s family of artificial intelligence models and products for chat, coding, and workplace tasks. The company has been widening that lineup with Claude Code, Claude Cowork, and industry products such as Claude for Financial Services. (anthropic.com) (claude.com) (anthropic.com) That business push now reaches regulated customers. Anthropic said last week that JPMorganChase, Google, Microsoft, Cisco, CrowdStrike, Palo Alto Networks, and other launch partners are using Claude Mythos Preview in defensive security work, with access also extended to more than 40 additional organizations that build or maintain critical software infrastructure. (anthropic.com) Mythos is a general-purpose language model that Anthropic says is its most capable frontier model so far, with stronger scores than Claude Opus 4.6 on many benchmarks. Anthropic’s April 7 system card and research post focus on cybersecurity testing and describe the model identifying thousands of high- and critical-severity vulnerabilities in controlled evaluations. (anthropic.com) (red.anthropic.com) Bloomberg reported on April 10 that Wall Street banks including Goldman Sachs, Citigroup, and Bank of America had gained or expected access to Mythos, while JPMorganChase was the only bank named publicly in Anthropic’s initiative. Silicon Republic reported the same day that the preview was being tested by banks and other regulated users in the United States, Canada, and Britain. (bloomberg.com) (siliconrepublic.com) At the same time, some Claude users say the product has become less reliable. VentureBeat reported on April 13 that developers and power users have complained about weaker performance from Claude Opus 4.6 and Claude Code, while Anthropic leaders said the company has been tuning settings across user segments and disputed the idea that the core model had simply been downgraded. (venturebeat.com) Anthropic has also changed how some heavy users pay. Starting April 4, Claude Pro and Max subscription limits no longer covered third-party agent harnesses such as OpenClaw, shifting that usage to separate pay-as-you-go billing instead. (thenextweb.com) The company is also spending to pull more enterprises into production. Anthropic said on March 11 that it would invest an initial $100 million in the Claude Partner Network to fund training, technical support, and joint go-to-market work for firms helping customers deploy Claude. (anthropic.com) The result is a company trying to scale on two fronts at once: selling Claude deeper into large organizations while adjusting model access, pricing, and controls for the users who stress the system most. (anthropic.com) (venturebeat.com)