Anthropic has officially begun assembling a dedicated in-house team to design custom AI chips, a move confirmed through a company statement to Business Insider and a new job listing for a lead silicon engineer. The initiative marks a pivot toward vertical integration, aiming to co-design hardware alongside its Claude models for faster, more efficient inference at scale.
The push for custom silicon
Until now, Anthropic leveraged compute from AWS Trainium, Google TPUs, Nvidia GPUs, and AMD for its AI workloads. However, surging demand — reflected by a run‑rate revenue exceeding $30 billion in April, up from roughly $9 billion at the end of 2025, and over 1,000 customers spending more than $1 million annually — has driven the company to seek its own chip designs. A spokesperson emphasized that Anthropic would continue using chips from all existing partners while building its own layer, ensuring supply‑chain flexibility.
Team details and ambitions
The job listing seeks engineers skilled in front‑end design, pre‑silicon verification, physical design, analog and mixed‑signal work, and packaging, with a salary band of $320,000 to $485,000. Candidates must demonstrate “direct personal contribution” to shipped semiconductor designs and be prepared for “consequential calls without a large organization behind them.” The listing also mentions supporting “first‑silicon bring‑up and debug,” indicating that Anthropic expects to tape out a physical chip rather than merely evaluate third‑party options. Reports from earlier this year suggested the company was in talks with Samsung as a potential manufacturing partner, though no factory or release date has been confirmed.
Industry context
Anthropic joins a growing list of frontier AI labs pursuing custom hardware. In June, OpenAI unveiled its Jalapeño inference chip built with Broadcom, claiming up to 50% cost reduction. Meta is accelerating its own MTIA accelerators, while Google DeepMind has long used Alphabet’s TPUs. Even smaller startups are exploring specialized silicon to reduce dependence on Nvidia’s dominant GPUs. For Anthropic, owning a chip layer not only lowers long‑term compute costs but also strengthens recruitment by offering engineers the chance to design cutting‑edge hardware from scratch.
