Anthropic to Use Google’s AI Chips Worth Tens of Billions to Train Claude Chatbot
In a landmark deal underscoring the escalating race for AI computing power, Anthropic has expanded its partnership with Google, securing access to as many as one million of the tech giant’s AI chips, valued at tens of billions of dollars, to train future versions of its flagship chatbot, Claude.
The agreement grants Anthropic more than one gigawatt of computing capacity—set to come online in 2026—for training and deploying advanced iterations of its Claude model using Google’s custom-designed Tensor Processing Units (TPUs). These chips, traditionally reserved for Google’s own AI operations, have now become a key strategic asset in Anthropic’s push to compete with industry leaders like OpenAI and Anthropic’s major investor, Amazon.
Why Anthropic Chose Google’s TPUs
Anthropic said it selected Google TPUs for their superior price-performance ratio, energy efficiency, and compatibility with the company’s existing infrastructure.
The startup has already been using TPUs for training and serving its Claude models, and the new deal cements Google as its primary infrastructure partner moving forward.
This partnership also deepens Anthropic’s reliance on Google Cloud, which will provide additional computing, storage, and networking services for the company’s AI workloads.
The Billion-Dollar Compute Race
The deal is the latest example of soaring demand for AI computing power as leading developers seek to build models capable of reasoning, learning, and interacting more like humans.
For context, Alphabet’s TPUs offer an alternative to the supply-constrained Nvidia GPUs, which currently dominate the AI hardware market.
Industry executives estimate that one gigawatt of compute—roughly equivalent to the power consumed by one million homes—can cost about $50 billion in infrastructure, hardware, and energy.
By comparison, OpenAI has reportedly signed multi-year deals valued at over $1 trillion to secure up to 26 gigawatts of computing capacity for future AI systems, leveraging Nvidia and AMD chips.
Anthropic’s Growth and Vision
Founded by former OpenAI researchers, Anthropic has positioned itself as an AI safety-first company, focusing on alignment, trustworthiness, and enterprise-ready solutions.
Its Claude models—used by companies including Notion, Quora, and Cursor—have been praised for their balanced reasoning and ethical AI behavior.
Reuters recently reported that Anthropic’s annualized revenue run rate is projected to more than double or even triple in 2025, driven by the growing adoption of Claude Pro, Claude for Teams, and enterprise APIs.
Amid intense competition and massive infrastructure costs, this deal strengthens Anthropic’s technical foundation while giving Google a stronger foothold in the global AI infrastructure market.
