Anthropic is conducting a strong AI collaboration with Google.

Anthropic announced on Thursday that Google will utilize up to one million of its artificial intelligence chips to train Claude models. The value of this agreement is estimated to reach tens of billions of dollars, and the company aims to develop generative artificial intelligence products in a rapidly changing market environment.

This collaboration underscores the significant computational power demands for training, application, and continuous inference by artificial intelligence firms. Developers are turning to multi-billion dollar agreements to quickly provide the necessary infrastructure.

Alphabet-backed Google will provide additional cloud computing services to the startup behind the Claude chatbot. The agreement comes at a time when Google is increasing the external availability of tensor processing units, or TPUs, that were previously reserved for internal use.

Google rents TPUs through Google Cloud. Anthropic stated that they prefer these processing units due to their price-performance ratio and efficiency, and they trust their existing experience in training models with these processors.

Reuters predicted earlier in October that Anthropic would likely double and potentially triple its annual revenues next year, thanks to the rapid adoption of its enterprise products.