Anthropic's $7 Billion Chip Plan Takes a New Turn

This article first appeared on GuruFocus.

Anthropic seems bent on reducing its reliance on externally sourced AI processors, even if a big purchase doesn't materialize.Reuters reported that the developer of Claude had been in talks to acquire AI chip startup MatX for almost $7 billion, although the discussions then pivoted to a possible cooperation.MatX was formed by ex-Google tensor processing unit engineers and is building chips to train massive AI models. The business is currently raising further money at a valuation of around $4 billion. Anthropic's appeal is simple.The corporation might get more control over cost and performance by building custom silicon, as demand for Claude increases. Anthropic has already recruited veterans from Google and OpenAI and grown its internal silicon team.Meanwhile, the corporation is spending much on outside computers. Anthropic aims to spend $36 billion on Google AI chips and has signed a $45 billion cloud-computing deal with Nscale, highlighting the huge cost of securing enough infrastructure to train and run frontier models.Another route for Anthropic to develop out more of that tech on its own might be a MatX relationship. The acquisition may have hit a snag, but Anthropic's push into bespoke chips certainly hasn't.