Anthropic is significantly expanding its use of Google Cloud technologies by incorporating up to one million Tensor Processing Units (TPUs), in a move valued at tens of billions of dollars, to enhance its AI research and product development capabilities. This expansion, expected to bring over a gigawatt of capacity online in 2026, is part of Anthropic's strategy to meet the growing demand from its 300,000 business customers, which include a rapidly increasing number of large accounts. The expansion aims to support more comprehensive testing, alignment research, and responsible large-scale deployment, as well as maintain Claude's cutting-edge status in the industry. Anthropic employs a diversified compute strategy, utilizing Google's TPUs, Amazon’s Trainium, and NVIDIA’s GPUs, to advance Claude's capabilities while maintaining strong partnerships, including a continued collaboration with Amazon on Project Rainier. The announcement comes amid Anthropic's international growth, with a new office in Seoul, and reaffirms its commitment to American AI leadership.