M&A

Anthropic drops $7B MatX buyout, pivots to chip partnership

What's the deal? Anthropic discussed acquiring AI chip startup MatX for roughly $7 billion but abandoned the deal, shifting toward a partnership to develop custom hardware, two people briefed on the matter told Reuters. MatX, founded by former GoogleDealroom has a profile for this one. Try Dealroom → tensor processing unit (TPU) engineers, is now seeking to raise fresh capital at a valuation of about $4 billion.

Why now? As Anthropic scales its Claude family of AI models, it wants hardware to match its appetite for data crunching and cut its reliance on Nvidia. Designing chips is slow and costly — a single generation can take over a year and run into the hundreds of millions of dollars.

What's the endgame? Anthropic is building an in-house silicon team to design custom chips that let Claude run faster and cheaper. MatX has been developing a chip suited to training large AI models, which would give Anthropic design expertise and lower costs over time.

Building the bench: This week, Anthropic hired Google chip veteran Amir Salek for its hardware push. In June, it hired former OpenAI chip engineer Clive Chan, who worked on the OpenAI chip unveiled earlier in 2026.

The road to IPO: Anthropic is expected to list in 2026, chasing a $2 trillion valuation that hinges on 2028 revenue of as much as $200 billion. The offering is expected months after SpaceXDealroom has a profile for this one. Try Dealroom → went public at a $1 trillion valuation.

What could go wrong? Reuters could not learn why the talks stalled. Anthropic declined to comment on the MatX discussions, and MatX did not respond to a request for comment.

The signal: Leading AI labs are racing to build custom silicon tailored to their models, chasing performance and cost advantages. Anthropic is keeping a multi-chip approach — committing $36 billion to Google's AI chips, $45 billion to rent cloud power from Nscale, and $1.25 billion a month to SpaceX through May 2029 — while OpenAI touts its first custom chip, Jalapeno, as outperforming a comparable Nvidia processor.

Read more: straitstimes.com, Reuters

Image credit: RyanDonegan

More top stories