Report: Google in talks to buy roughly $40B stake in Anthropic

Google isn’t just buying a piece of Anthropic. Google is trying to lock in a place in the AI supply chain — chips, cloud, and model distribution all at once. The April 24 announcement matters because the headline number, up to $40 billion, is only half the story. The other half is compute. Anthropic needs absurd amounts of it, and Google wants that demand running through Google Cloud and Google’s own TPUs. ### What actually happened? Google said it will invest up to $40 billion in Anthropic, with $10 billion going in now and as much as $30 billion more tied to performance milestones. The deal extends a relationship that already made Google an investor, cloud partner, and hardware supplier to the Claude maker. ### Why is this bigger than a normal startup round? Because Anthropic is no longer a normal startup. In February, Anthropic raised $30 billion at a $380 billion post-money valuation and said run-rate revenue had reached $14 billion. By April 6, it said run-rate revenue had already surpassed $30 billion. That is the kind of growth curve that turns infrastructure from a support function into the whole game. ### So is this really a cloud deal? Basically, yes. Anthropic said back in October 2025 that it planned to expand its use of Google Cloud technologies to as many as one million TPUs, in a deal worth tens of billions of dollars and expected to bring well over a gigawatt of capacity online in generation TPU capacity starting in 2027. The money and the machines fit together. ### Why do TPUs matter so much here? Google’s TPUs are its in-house AI chips — the alternative to renting mostly Nvidia-based capacity. If Google can get a frontier lab like Anthropic to train and serve major workloads on TPUs, Google isn’t just selling cloud time. It is proving its custom silicon can carry one of the better answers to the Microsoft-OpenAI and AWS-Trainium stories. ### But isn’t Amazon Anthropic’s main partner? That’s the catch. Anthropic keeps saying Amazon remains its primary cloud provider and training partner for mission-critical workloads. On April 20, Anthropic and Amazon expanded their collaboration again, with Anthropic committing more than $100 billion over 10 years, yet not locking Anthropic up exclusively. ### Why would Anthropic want it that way? Because dependence is dangerous. Anthropic trains and runs Claude across AWS Trainium, Google TPUs, and Nvidia GPUs, and it keeps telling customers that this mix gives better resilience and flexibility. Think of it like a power-hungry factory buying electricity from multiple grids. If one supplier gets big. ### What does Google get besides upside on the equity? A giant anchor tenant. Anthropic said Claude is available on AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry, but the infrastructure underneath still has to live somewhere. If more Claude training and inference land on Google hardware, Google captures spending that might otherwise flow to Nvidia-heavy clouds or to Amazon’s Trainium stack. ### Bottom line This is the AI market maturing into something more industrial. The flashy part is the $40 billion. The important part is that frontier labs and cloud giants are now financing each other’s scale. Anthropic brings demand. Google brings chips, capital, and distribution. Whoever controls that loop gets more than a stake in a startup — they get leverage over how enterprise AI gets built.