Google’s $40B Anthropic Bet

Google and Anthropic said Friday that Google will invest up to $40 billion in the artificial intelligence startup, beginning with $10 billion now. (cnbc.com) The remaining $30 billion is tied to performance milestones, according to CNBC’s report on the agreement announced April 24. TechCrunch reported the package includes both cash and computing capacity. (cnbc.com) (techcrunch.com) That structure matters because training and running large language models is now as much a data-center business as a software business. Anthropic’s Claude models need chips, networking and cloud capacity in volumes that few companies can supply. (techcrunch.com) (anthropic.com) Google is not just writing a check. The deal deepens Anthropic’s use of Google Cloud and Google’s Tensor Processing Units, the in-house chips Google sells to outside customers through its cloud business. (cnbc.com) (blog.google) Anthropic had already moved more of its newest work onto Google’s platform this month. On April 7, Google Cloud said Claude Mythos Preview was available in private preview on Vertex AI for select customers. (cloud.google.com) Anthropic also said two weeks ago that it was expanding its partnership with Google and Broadcom for more compute, while still training and running Claude across Amazon Trainium, Google Tensor Processing Units and Nvidia graphics processors. (anthropic.com) The company added that Amazon remains its primary cloud provider and training partner, and said it continues to work with Amazon Web Services on Project Rainier. That means Google’s new commitment strengthens Anthropic’s options rather than making Google its only infrastructure backer. (anthropic.com) For Google, the investment spreads its artificial intelligence bets beyond its own Gemini models and gives Google Cloud a closer tie to one of the biggest outside model makers. CNBC said the agreement comes as big technology companies compete to lock up access to leading AI labs. (cnbc.com) For Anthropic, the immediate effect is simpler: more money and more machines. In the current AI market, those two inputs increasingly decide which labs can keep shipping larger models and serving more enterprise customers. (techcrunch.com) (anthropic.com)