The artificial intelligence infrastructure battle is only getting more intense, folks. AI company Anthropic announced Monday that it is expanding its use of cloud services and custom AI chips from Google Cloud, locking in access to massive computing power to train and run its growing lineup of models.
The deal centers on Google’s Tensor Processing Units, or TPUs, which are specialized chips built for AI workloads. Under the expanded agreement, Anthropic will gain access to multiple gigawatts of TPU capacity, an enormous amount of compute that will begin coming online starting in 2027. Those chips will be delivered through Google Cloud infrastructure, with hardware supplied through Broadcom.
Anthropic says the additional compute will help support development of its foundation models, AI agents, and enterprise software built around its Claude family of AI systems. The company has been scaling aggressively as demand for generative AI services continues to grow across industries.
Beyond the raw compute, Anthropic is also leaning further into Google’s broader cloud ecosystem. The company is using tools like BigQuery for data analysis, Cloud Run for application deployment, and AlloyDB for database workloads. According to the announcement, thousands of companies already access Anthropic’s Claude models through Google Cloud.
Those customers include firms such as Coinbase, Cursor, Palo Alto Networks, Replit, and Shopify.
The move highlights the increasingly tight relationship between major AI labs and cloud providers. Training cutting-edge AI models requires massive computing clusters and staggering amounts of electricity, meaning startups like Anthropic often partner closely with infrastructure giants that can supply both chips and global-scale data centers.
Google clearly wants a big seat at the AI table, and deals like this show the search giant is willing to deploy its custom silicon to compete with rivals such as Microsoft and Amazon Web Services, both of which are also investing heavily in AI hardware and infrastructure.
At the same time, Anthropic continues to pursue a multi-cloud approach, but this announcement makes one thing clear: Google’s TPUs are becoming a major pillar of the company’s future AI ambitions.
