Anthropic had planned to acquire AI chip startup MatX for approximately $7 billion but has since moved away from the purchase. The decision highlights the company's strategic focus on securing chip design expertise to support its growing AI infrastructure needs.
- Anthropic considered a $7 billion acquisition of MatX but dropped the deal.
- Company aims to develop proprietary AI training chips to reduce Nvidia dependency.
- Talks with other AI chip startups continue alongside expanding internal chip design team.
What happened
Anthropic initially planned to purchase MatX, an AI-focused chip startup valued around $7 billion, to accelerate its development of custom silicon for its AI models. MatX's engineering team includes former Google TPU developers, making it a strategic target for Anthropic's ambition to design specialized AI training chips.
However, the acquisition discussions were discontinued without a disclosed reason. Instead, the talks have evolved into the possibility of a partnership, while MatX seeks new funding at a lower valuation of about $4 billion. Anthropic declined to comment on the status of the deal.
Why it matters
As AI models like Anthropic's Claude family grow larger, the demand for efficient, high-performance chips increases. Developing in-house chip capabilities could reduce costs and dependency on dominant chip providers like Nvidia, whose processors face supply constraints through 2027.
By designing its own silicon, Anthropic hopes to optimize model training speeds and overall efficiency. The company is simultaneously expanding its silicon engineering team with experienced executives from Google and OpenAI and investing billions in cloud computing and chip purchases from other providers.
What to watch next
Anthropic's approach to custom chip development will be closely watched as it balances direct chip acquisitions, cloud computing contracts, and internal chip design efforts. The company's strategic decisions will impact its competitive position amid rising AI hardware demand.
Further discussions with other AI chip startups may lead to new partnerships or acquisitions as Anthropic evaluates different chip design architectures for training and inference workloads. Their upcoming IPO, expected in 2026 or 2027, will hinge on the successful scaling of their AI technology and hardware strategy.