Anthropic confirmed on August 5, 2026 that it is building an in-house team to design custom AI chips optimized for its Claude models, joining a growing club of frontier labs racing to control their compute stack from the transistor upward.
A ‘custom silicon team’ for Claude
Business Insider was first to report the hiring push, which Anthropic then confirmed to TechCrunch. The company is recruiting engineers with experience across hardware and software design and told employees the goal is to “co-design hardware and models” so Claude can run faster and cheaper. Job listings describe a new custom silicon team without setting a public tape-out date.
Samsung, Broadcom and the vertical-integration race
The Information reported last month that Anthropic is in talks with Samsung Electronics about manufacturing a future chip on the Korean giant’s 2-nanometer process, potentially tapping its advanced packaging as well. The move follows OpenAI’s June debut of the Broadcom-built Jalapeño inference chip and mirrors longstanding programs at Google (TPU) and Meta (MTIA). Anthropic will keep buying capacity from AWS, Google Cloud, Nvidia and AMD while its own design ramps.
Why now
Anthropic’s revenue reportedly cleared $30 billion this year and demand for Claude keeps outrunning available accelerator capacity across cloud providers. The company also has close context on the memory crunch highlighted by Samsung and Broadcom’s $200B AI infrastructure MoU, and it sits alongside compute megadeals like Amazon’s $50 billion OpenAI investment, Core Scientific’s 2.5 GW AMD partnership, and Nvidia’s $5B bet on Safe Superintelligence — each reinforcing why frontier AI labs need to own the silicon roadmap.
Reporting based on coverage from TechCrunch, Business Insider and The Information.