Anthropic in Talks With MatX to Accelerate Custom AI Chip Design and Lower Costs
Anthropic's Strategic Moves in AI Chip Development
By Max A. Cherney and Krystal Hu
Anthropic's Discussions with MatX
SAN FRANCISCO, Aug 27 (Reuters) - Anthropic discussed buying artificial intelligence chip startup MatX for roughly $7 billion, seeking to accelerate efforts to develop custom hardware for its fast-growing AI business, two people briefed on the matter told Reuters.
The merger talks, which a third person said have evolved into a discussion about a partnership, underscore the AI lab’s ambition to secure the resources and talent needed to build its own chips. The goal is to help Anthropic accelerate its in-house chip development.
Reuters is reporting on the discussions for the first time but could not learn why the talks are no longer active. MatX, founded by former Google tensor processing unit (TPU) engineers, is now seeking to raise new capital at a valuation of about $4 billion, one of the people said, requesting anonymity to discuss private matters.
Anthropic declined to comment on deal talks with MatX. MatX did not respond to a request for comment.
Chip Plans on Road to IPO
As Anthropic scales its Claude family of AI models, the company is looking to produce hardware that can fulfill its voracious appetite for data crunching and reduce its reliance on chips produced by Nvidia.
Anthropic, which is expected to list this year, has hired engineering and executive talent to accelerate the chip design process, which could take years. Anthropic's IPO, expected months after SpaceX went public with a $1 trillion valuation, will chase a valuation of $2 trillion, which hinges on a 2028 revenue figure of as much as $200 billion, Reuters reported earlier this month.
Anthropic said it is expanding an in-house silicon team to design custom chips that will allow Claude models to run faster and more efficiently. It also plans to keep a multi-chip approach by working with providers from Nvidia to Google.
Challenges and Benefits of Custom Chip Design
Designing chips is expensive and time-consuming. It can take a year or more to produce a viable piece of hardware, and design costs for a single generation run in the hundreds of millions of dollars.
An acquisition of an AI chip startup such as MatX would give Anthropic in-house design expertise and could help lower costs over the longer term, the people said. MatX has been working on a chip that would be useful for building large AI models in a process called training.
Anthropic plans to spend tens of billions of dollars to rent computing power from cloud providers and also plans to buy chips directly. It plans to buy $36 billion worth of Google’s AI chips, and signed a $45 billion deal to rent AI cloud computing power from Nscale. It agreed to pay SpaceX $1.25 billion per month through May 2029 for computing capacity across its data center clusters.
Discussions With Other AI Chip Startups
The discussions with MatX also suggest Anthropic may be interested in producing a training chip, while other chip startups and rival OpenAI pursue processors better suited to generating responses from chatbots, a process known as inference. Anthropic could elect to produce an inference chip as well, the sources said.
In recent weeks, Anthropic held meetings with a range of AI chip startups. It has not yet elected to make an acquisition. The meetings are an attempt by Anthropic’s engineers and executives to understand the current range of chip design approaches.
Talent Acquisition and Industry Trends
This week, Anthropic hired Google chip veteran Amir Salek as part of the company’s push into hardware. In June, the AI lab hired former OpenAI chip engineer Clive Chan who worked on the OpenAI chip unveiled earlier this year.
The leading AI labs such as Anthropic and OpenAI have become increasingly focused on custom chips, which can be tailored to their models and workloads. Through this, they hope to create significant performance and economic advantages.
Competitive Landscape and Supply Chain Considerations
At a conference this week OpenAI boasted that its first custom chip called Jalapeno outperformed a similar processor made by Nvidia. OpenAI executives noted that the chip was more energy efficient in performing inference calculations.
Making a custom chip may also help Anthropic hedge against the tight supply of Nvidia’s processors, which Nvidia said in a conference call on Wednesday would be in short supply through 2027. Google has developed TPUs, while Amazon has built its Trainium and Inferentia chips.
Anthropic was one of the first companies to run its models on hardware from several vendors including Nvidia, Google and Amazon.
The company has been seeking more computing capacity to meet surging demand for its models, including through a deal with SpaceX to use its Colossus 1 facility, which houses more than 220,000 Nvidia chips.
(Reporting by Max A. Cherney and Krystal Hu in San Francisco; Editing by David Gregorio and Christian Schmollinger)
