Broadcom confirms it will make future versions of Google’s AI chips; says: Will draw on ...

Broadcom confirmed this week that it has agreed to manufacture future version of Google’s artificial intelligence chips, cementing its role in powering the next wave of generative AI. Along with this, the company also announced an expanded partnership with Anthropic, which will gain access to about 3.5 gigawatts of computing capacity built on Google’s tensor processing units (TPUs). After the announcement the shares of Broadcom also rose 3% in an extended trading following the disclosure.AI giant and maker of Claude, Anthropic has seen explosive growth this year. The company also reported annualised revenue surpassing $30 billion, up from $9 billion at the end of 2025. It now counts more than 1,000 enterprise clients spending over $1 million annually — double the number just two months ago.“This groundbreaking partnership with Google and Broadcom is a continuation of our disciplined approach to scaling infrastructure,” said Anthropic CFO Krishna Rao. “We are building the capacity necessary to serve exponential customer growth while enabling Claude to define the frontier of AI development.”Most of the new infrastructure will be located in the US, Anthropic added.Broadcom already helps Google produce its TPUs, which are critical for training and running large AI models. On an earnings call last month, CEO Hock Tan said Broadcom had delivered 1 gigawatt of compute for Anthropic in 2026 and expects demand to surge past 3 gigawatts in 2027.Analysts at Mizuho estimate Broadcom could generate $21 billion in AI revenue from Anthropic in 2026 and $42 billion in 2027, though Monday’s filing did not specify dollar amounts.Broadcom is also working with OpenAI on custom silicon. Both OpenAI and Anthropic currently rely heavily on Nvidia GPUs through cloud providers like Amazon, Google, and Microsoft. OpenAI has additionally committed to tapping six gigawatts of AMD GPUs, with the first gigawatt expected later this year.