Anthropic Builds In-House Team to Develop Custom AI Chips
Anthropic is assembling a semiconductor team to design custom chips for its Claude AI models to improve efficiency and reduce reliance on off-the-shelf hardware.
Anthropic is establishing an in-house silicon team of designers and verification experts to create custom artificial intelligence chips specifically for its Claude AI models. The company intends to co-design hardware and software to increase performance and operational efficiency while reducing its dependence on off-the-shelf graphics cards. To streamline the process, Anthropic plans to use Claude to automate chip verification tasks.
Despite the move toward proprietary silicon, the company is maintaining a multi-chip strategy. It will continue to rely on hardware from Amazon Web Services, Google, Nvidia, and AMD to support its scaling efforts. Reports indicate that Anthropic has held discussions with Samsung Electronics as a potential manufacturing partner, possibly to leverage Samsung's zHBM technology.
This strategic shift follows a broader industry trend of AI labs developing dedicated hardware. Competitors including Meta Platforms Incorporated and OpenAI have pursued similar paths, with OpenAI recently debuting a custom inference accelerator called Jalapeño developed in collaboration with Broadcom Inc.