Anthropic explores Samsung custom AI chip while hiring chip lead

Anthropic has held early talks with Samsung about building a custom AI chip, though the company has not decided what the chip would be used for, how it would fit into servers, or how powerful it should be. Anthropic also told TechCrunch a diversified hardware stack that includes chips from Google, Amazon, and Nvidia will continue to be pivotal to its compute strategy.

What exactly is Anthropic doing with Samsung?

Anthropic is in exploratory discussions with Samsung about manufacturing a bespoke AI chip, but the project remains at an early stage and lacks a detailed design. TechCrunch says Anthropic has not settled the chip's intended role, server integration, or target performance, and "had nothing further to add" on a potential Samsung partnership.

The Information is the outlet linked to the initial disclosure of the talks, and subsequent coverage notes the effort is still formative. The Decoder adds that Anthropic has been hiring chip engineers and specifically named Clive Chan, described as an early member of both Tesla's and OpenAI's custom chip teams, who is expected to build out a dedicated chip group at Anthropic.

How does this fit into Anthropic's hardware strategy?

Anthropic intends to keep a diversified hardware stack that relies on chips from Google, Amazon, and Nvidia even while it explores its own silicon. The company framed the potential chip as one element among existing suppliers rather than a wholesale shift away from third-party accelerators.

The broader context is an industry push toward custom silicon. OpenAI recently unveiled "Jalapeño," an in-house inference processor built with Broadcom, and cloud providers including Amazon and Google already offer custom-built TPUs. Samsung is already a major manufacturing partner for Nvidia and is working with Nvidia on an AI chip factory in South Korea; Samsung has also discussed chip efforts with Google.

Why does Anthropic want its own chip?

Custom chips can lower the cost of running large models and let a company tune hardware to specific workloads; that economic logic is part of why multiple AI groups are pursuing bespoke silicon. The Decoder framed the move bluntly: whoever can build and operate AI infrastructure more cheaply keeps more of the revenue. Hiring experienced chip engineers like Clive Chan signals Anthropic is preparing internal capability rather than only buying external capacity.

Anthropic's public posture balances that motive against dependency risks: the company emphasizes continued reliance on Google, Amazon and Nvidia chips while it explores alternatives.

Why it matters

A chip program would change Anthropic's leverage in procurement and could reduce exposure to global shortages or vendor concentration. It would also place Anthropic in direct technological competition with other model builders that either run or design custom inference silicon, such as OpenAI with its Jalapeño chip. If Anthropic does ship silicon purpose-built for its workloads, it could shift negotiation dynamics with cloud and chip suppliers.

What to watch

Look for formal hires and org changes in Anthropic's chip team beyond Clive Chan, an explicit chip roadmap from the company, or a fabrication or manufacturing agreement with Samsung. A public technical design or performance claim would be the clearest confirmation that Anthropic has moved from exploration to productization.