Anthropic, the San Francisco‑based AI research lab, was deep in negotiations to buy MatX, a fledgling chip startup, for an estimated $7 billion, according to Reuters. The figure, roughly three‑quarters higher than the $4 billion valuation at which MatX is currently raising capital, sparked interest across the semiconductor and venture‑capital worlds.

MatX, launched in 2023 by Reiner Pope – who oversaw AI software for Google’s Tensor Processing Units – and Mike Gunter, a lead designer on the TPU hardware, focuses on processors built specifically for large‑language‑model training. The company’s pitch is simple: a chip optimized for a single workload can outperform Nvidia’s general‑purpose GPUs by a wide margin. It plans to ship volume units starting in 2027.

In February, MatX closed a $500 million Series B round led by Jane Street and Leopold Aschenbrenner’s Situational Awareness fund, with participation from Marvell, Spark Capital and the Collison brothers. The funding round set the company’s valuation at about $4 billion, a number that now serves as a benchmark for the proposed acquisition price.

Anthropic’s interest in MatX aligns with its broader push to develop in‑house silicon capability. Earlier this year the lab announced an internal chip team and began recruiting engineers who have already shipped silicon. It is also in talks with Samsung about manufacturing a custom part, signaling a multi‑pronged approach to securing hardware for its AI models.

Financial backing for Anthropic’s hardware ambitions is massive. Broadcom is reportedly seeking more than $60 billion in debt to fund chips destined for Anthropic, while AMD has poured $5 billion into the company and committed to a two‑gigawatt deployment. These figures dwarf the $7 billion price tag that surfaced for the MatX deal, suggesting that the acquisition would have been a modest addition to Anthropic’s overall financing plan.

Why the purchase fell through is not clear. Reuters could not pinpoint a specific trigger, and its reporting relies on unnamed insiders rather than any formal filing or public announcement. Both Anthropic and MatX declined to comment on the matter when approached.

Shifting from an outright purchase to a partnership could reflect strategic caution. An acquisition would have handed Anthropic a ready‑made design team but also introduced an immediate integration challenge at a time when the lab is racing to get chips into data centers. A supply agreement, by contrast, could deliver the needed silicon without reshaping Anthropic’s organization.

For MatX, remaining independent may be advantageous. Its Series B investors likely did not back the startup to become a single customer’s internal department. Continuing as an independent supplier could make it easier to raise future rounds at a $4 billion valuation, rather than being tied to one buyer whose acquisition fell through.

The episode underscores a broader shift in the AI hardware landscape. Companies that once relied on Nvidia’s GPUs are now seeking specialized accelerators, and the fastest route to an alternative lies with the few teams that have already designed such chips. Recent funding activity supports this trend: Etched, a similar accelerator startup, raised $500 million at a $5 billion valuation, and a cluster of related startups collectively secured about $1.6 billion across five rounds this year.

At present, no formal partnership has been announced, and the only confirmed fact remains the one Reuters put on record: a $7 billion acquisition was contemplated and then abandoned. The story highlights how quickly valuations and deal dynamics can evolve in the high‑stakes race to build the next generation of AI hardware.

This article was written with the assistance of AI.
News Factory APP - agentic news to boost your SEO & AEO.