Anthropic Is in Talks with Samsung to Build Its Own AI Chips

Anthropic is in early-stage discussions with Samsung to manufacture custom AI chips using the Korean company's 2nm process, per TechCrunch. The talks were first reported by The Information, and they signal that the maker of Claude is moving from quietly exploring the idea to actively pursuing it. NVIDIA still holds an estimated 74% of the AI chip market, but nearly every major US AI company is now working to reduce that dependency.

The custom-chip arms race

OpenAI launched its Jalapeño processor — built with Broadcom — just last month. Google has its TPUs, Amazon has Trainium, and Meta recently struck a deal with Samsung for its MTIA chip reportedly worth $7.3 billion, according to SammyFans. Anthropic is now following the same playbook.

The company hired Clive Chan from OpenAI's chip team — a clear sign the project has moved beyond casual exploration. Anthropic hasn't publicly confirmed the Samsung partnership, but a spokesperson told TechCrunch that its "diversified hardware stack, including chips from Google, Amazon, and Nvidia, will continue to play a key role in its compute strategy." That framing matters: the custom chip is a hedge, not a replacement for existing supplier relationships.

Specifics remain thin. Anthropic hasn't settled on what the chip will be used for — training, inference, or something else — nor on performance targets or a manufacturing timeline.

Samsung as the TSMC alternative

Samsung's appeal here isn't accidental. TSMC dominates advanced chip manufacturing, but tight capacity at its most advanced nodes has pushed several big AI spenders to look elsewhere. Samsung is actively building out its foundry customer list: alongside the Meta and Anthropic discussions, it is also in talks with Google over a chip codenamed Icefish, and already manufactures chips for Tesla.

The Korea Herald has covered Samsung's push to position its foundry division as a credible rival to TSMC for next-generation AI silicon. Landing Anthropic would add another marquee name to that effort.

What this means

For consumers, none of this changes anything immediately — Anthropic's products run on the same cloud infrastructure either way. But the broader shift matters: as AI companies build their own chips, they gain more control over cost and performance, which eventually shapes what AI tools can do and at what price. NVIDIA's dominance isn't going anywhere fast, but it now has serious competition from the very companies buying its hardware.