OpenAI–Anthropic tension rises

OpenAI and Anthropic used to fight mostly through model launches. This week, the fight showed up in two very different rooms at once: investor briefings and a Washington meeting with the biggest banks. (cnbc.com, bloomberglaw.com) OpenAI sent investors a memo saying it has 1.9 gigawatts of computing capacity, compared with Anthropic’s 1.4 gigawatts. The memo also said OpenAI plans to reach 30 gigawatts by 2030, while it expects Anthropic to have about 7 to 8 gigawatts by the end of 2027. (cnbc.com, bloomberg.com) That sounds abstract, but “computing capacity” is the electricity-and-chip budget behind training and running giant artificial intelligence systems. In this business, more gigawatts usually means more chips, bigger models, and faster product rollouts. (cnbc.com, bloomberg.com) OpenAI’s memo did not read like a routine fundraising deck. CNBC reported that it described Anthropic as “operating on a meaningfully smaller curve” and said the rival company was constrained by access to compute. (cnbc.com, youtube.com) The timing matters because Anthropic has been gaining commercial traction while weighing a possible initial public offering, which is the stock-market process that turns a private company into a public one. Bloomberg reported that OpenAI’s investor memo landed as Anthropic was gaining ground and considering that step. (bloomberg.com, cnbc.com) Then the story jumped from fundraising to financial stability. Bloomberg Law reported that Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell called bank chiefs to Washington over concerns that Anthropic’s newest model could raise cyber risk. (bloomberglaw.com, bloomberg.com) According to Bloomberg’s reporting, the meeting took place at the Treasury Department on Tuesday, April 7, 2026. The officials wanted major banks to understand possible risks tied to Anthropic’s model, called Mythos, and to harden their defenses. (bloomberg.com, bloomberglaw.com) A bank cyber warning is different from a product review. It means the question is no longer only whether a model writes better code or wins more customers, but whether it could help attackers find weak points in payment systems, trading systems, or internal networks faster than defenders can patch them. (bloomberglaw.com, sherwood.news) This is also not the first leak-driven flare-up between the two companies. In March 2026, reports on a leaked Anthropic memo said Chief Executive Officer Dario Amodei attacked OpenAI’s Pentagon deal and called its public rationale “safety theatre” and “mendacious.” (cnbctv18.com, cbsnews.com) Put together, the pattern is new. OpenAI is arguing to investors that Anthropic lacks the infrastructure to keep up, while top U.S. financial officials are treating Anthropic’s latest model as powerful enough to warrant a private warning to systemically important banks. (cnbc.com, bloomberglaw.com) That is what changed this week. The OpenAI-Anthropic rivalry is no longer just a contest over chatbots and coding tools; it is now shaping how investors value the companies and how regulators think about cyber risk in the banking system. (cnbc.com, bloomberg.com)