Anthropic has agreed to pour $30bn into Google’s TPU chips and cloud services, a move that underlines the firm’s ambition to secure the computing muscle behind its rapidly expanding AI lineup. The deal grants access to roughly 5GW of new capacity, a figure matching the power output of a small nuclear reactor in the future.

The arrangement follows a separate filing that will bring an additional 3.5GW of Google hardware through a partnership with chipmaker Broadcom next year. Anthropic’s CFO Krishna Rao said the company’s annualised revenue has jumped from $9bn to $30bn in just 28 days, reflecting a surge in demand for its Claude coding agent across the market.

Google, which holds a 14% stake in Anthropic, is pushing its in‑house TPUs to compete directly with Nvidia. The deal also sees Broadcom committing to supply custom TPUs for Google through 2031, a long‑term partnership that hinges on Anthropic’s commercial success and could reshape supply chains in the AI sector.

The hefty investment mirrors similar moves by competitors such as OpenAI, which has secured capacity from Broadcom, Nvidia, and AMD. Analysts note that the circularity of tech giants acting as customers, suppliers, and investors raises regulatory eyebrows. Nonetheless, Anthropic’s commitment signals that scaling AI models will continue to command billions in hardware spending for future