Anthropic is building an in-house team to design custom chips for its Claude AI models. The company is seeking engineers with experience in chip design for its custom silicon team.
The goal of the initiative is to co-design hardware and models to make the technology run faster and more efficiently. Anthropic has not provided a timeline for its chip plans or stated if it intends to manufacture the chips itself.
The Information reported that Anthropic is scouting Samsung as a potential partner for building custom chips. Previous remarks from Anthropic regarding 'logic chips' had sparked speculation concerning a potential Samsung partnership.
Anthropic currently employs a multi-chip strategy, having established deals with Amazon Web Services (AWS), Google, Nvidia, and AMD to access AI computing hardware.
Industry competition and development trends
Other major technology firms are also developing proprietary hardware. OpenAI unveiled its Jalapeño chip, built with Broadcom, which is designed for inference workloads and is expected to undergo mass deployment by the end of 2026. Google DeepMind relies on Alphabet's Tensor Processing Unit (TPU) chips to power its AI models, while Meta has been developing its own MTIA accelerators for AI workloads.
Estimated cost of designing an advanced AI chip, according to industry sources, due to the need for skilled engineers and defect-free manufacturing.
