Broadcom extends chip agreements with Google and Anthropic

Broadcom has announced that it plans to manufacture upcoming versions of AI chips for Google, alongside an enhanced agreement with Anthropic, enabling the AI startup to access about 3.5 gigawatts of computing capacity using Google's AI processors.

Broadcom's shares went up by 3% in after-hours trading, while shares of Google's parent company, Alphabet, remained steady.

This securities filing highlight shows the booming need for setups capable of handling generative AI technologies. 

Anthropic's fame has increased this year, with its Claude app topping the list of free US apps in Apple's App Store in February after issues between the company and the Pentagon surfaced.

During last month's earnings call, Broadcom CEO Tan remarked, "For Anthropic, we're beginning 2026 on a strong note with one gigawatt of TPU compute. 

Looking forward to 2027, we anticipate the demand will surpass three gigawatts of compute."

Following the earnings call, Mizuho analysts headed by Vijay Rakesh projected that Broadcom would garner $21 billion in revenue from Anthropic's AI in 2026 and $42 billion in 2027. The filing on Monday did not specify a dollar figure.

At the same time, Broadcom is also teaming up with Anthropic competitor OpenAI on creating specialised silicon for AI. 

Presently, both model creators mainly utilise graphics processing units from Nvidia via cloud service providers like Amazon, Google, and Microsoft. 

OpenAI also plans to utilise six gigawatts of AMD's GPUs, with the initial gigawatt scheduled to be available in the year's latter half.