Anthropic has held discussions with AI chip startup MatX as it accelerates its custom-chip development efforts.
The acquisition talks, initially valued at around $7 billion, have reportedly evolved into discussions about a possible partnership.
Anthropic is expanding its in-house silicon team and has also been meeting other AI chip startups.
Anthropic Holds Talks With AI Chip Startup MatX To Accelerate Chip Development
Anthropic is exploring a possible partnership with MatX as it steps up efforts to develop custom AI chips and reduce its reliance on Nvidia
Summary
Advertisement
Anthropic has discussed buying artificial intelligence (AI) chip startup MatX for around $7 billion as it seeks to accelerate development of custom hardware for its rapidly expanding AI business, according to a Reuters report.
The talks have since evolved into discussions about a possible partnership. The reason the acquisition discussions are no longer active could not be determined.
MatX, founded by former Google engineers who worked on the company’s Tensor Processing Units (TPUs), is now seeking to raise fresh capital at a valuation of around $4 billion, according to the report.
Anthropic Steps Up Custom Chip Push
As Anthropic expands its Claude family of AI models, the company is looking to develop hardware that can handle its growing computing needs and reduce its dependence on Nvidia chips.
Anthropic has said it is expanding an in-house silicon team to develop custom processors that could allow Claude models to run faster and more efficiently. The company also plans to continue using chips from multiple suppliers, including Nvidia and Google.
Advertisement
An acquisition of MatX could give Anthropic chip-design expertise and help lower costs over time, according to the people cited by the news agency. MatX has been developing a processor designed for training large AI models.
AI Labs Race To Build Their Own Chips
Anthropic has held meetings with several AI chip startups in recent weeks but has not decided to acquire any of them, according to Reuters. The discussions are aimed at helping its engineers and executives understand different approaches to chip design.
The company recently hired Google chip veteran Amir Salek as part of its hardware push. In June, it also hired former OpenAI chip engineer Clive Chan, who worked on OpenAI’s custom processor.
The push comes as AI companies look to build chips tailored to their models and workloads. Custom processors could help improve performance and reduce costs, while also giving companies more control over computing capacity.
Advertisement
Anthropic is expected to spend tens of billions of dollars on computing capacity. It plans to buy $36 billion worth of Google AI chips and has agreed to a $45 billion deal with Nscale for AI cloud computing capacity, according to the news agency.
The company has also agreed to pay SpaceX $1.25 billion per month through May 2029 for computing capacity across its data centre clusters, highlighting the scale of computing resources required to support its AI models.