Anthropic builds custom silicon team for Claude chips

Anthropic has begun recruiting a dedicated custom silicon team to design proprietary semiconductors tailored for its Claude artificial intelligence models. The initiative, confirmed by an Anthropic spokesperson following recent job postings for a silicon engineer and a technical program manager, marks a shift toward in-house hardware development. While the company intends to maintain a multi-chip approach by continuing to utilize third-party processors alongside its own designs, reports suggest Anthropic has explored a manufacturing partnership with Samsung to bring its silicon to life.

This move aligns Anthropic with a broader industry trend of major AI developers seeking independence from Nvidia, whose market dominance has created supply bottlenecks. For instance, OpenAI recently collaborated with Broadcom to develop its own custom inference chip, named Jalapeño. Meanwhile, Google and Meta already deploy proprietary hardware to run their workloads, and Mistral is reportedly investigating a similar path. By co-designing hardware and software side-by-side, Anthropic hopes to achieve the tight vertical integration necessary to optimize performance and efficiency.

For AI practitioners and software developers, this hardware transition could eventually yield faster, more cost-effective API access to Claude models. As the market shifts toward smaller, open-weight models capable of running on edge devices, proprietary silicon could help Anthropic maintain a competitive edge by lowering operational costs for its frontier models. However, because the chip design team is still in its formative hiring stages, developers should not expect to see the practical benefits of this custom hardware in their workflows for some time.