Anthropic explores custom chip with Samsung
Anthropic’s Samsung chip discussions show how frontier AI labs are trying to control more of the compute stack.
Frontier AI labs are increasingly thinking like infrastructure companies.
What happened
Anthropic is in discussions with Samsung around a possible custom AI chip. The details remain early, including how the chip would be used and what kind of performance profile it would target.
The discussion points to a broader effort by AI labs to reduce dependence on standard compute supply chains and shape hardware around their own workloads.
Why it matters
Compute is one of the biggest constraints in AI. If frontier labs can influence chip design, they may improve cost, performance and availability for model training and inference.
For Anthropic, a custom chip strategy would also make sense as AI models become more expensive to run and competition for high-end accelerators stays intense.
The bigger picture
The AI race is no longer only about models. It is also about chips, data centres, power, networking and capital access. The more AI labs move into hardware, the more they start to look like full-stack infrastructure companies.