Google is set to deliver up to one million of its in-house artificial intelligence chips to Anthropic, marking one of the largest hardware commitments yet in the AI sector.
The multiyear agreement, reportedly valued in the tens of billions, gives the fast-growing AI company a huge boost in computing capacity while reinforcing Google’s role as both investor and infrastructure backbone.
The chips, known as tensor processing units or TPUs, are custom-built by Google to accelerate AI model training and will begin rolling out in 2026.
Their deployment will add more than a gigawatt of processing power, a scale that highlights the escalating costs of developing and operating advanced language models.
Anthropic, which builds the Claude series of chatbots, relies on both Google Cloud and Amazon Web Services to train its systems.
Alphabet Inc (NASDAQ:GOOG)-owned Google has already invested about $3 billion in the company, while Amazon.com Inc (NASDAQ:AMZN) has pledged up to $8 billion.
Krishna Rao, Anthropic’s finance chief, said the latest deal would “help us continue to grow the compute we need to define the frontier of AI.”
The move strengthens Google’s cloud credentials and positions its TPUs as a rival to Nvidia’s dominant AI chips, further intensifying the global race for processing power.