Anthropic has announced plans to establish an in-house silicon team to design custom chips for its AI models. The company confirmed these plans to both Business Insider and TechCrunch after a job listing for a senior semiconductor engineer was noticed by the media. The spokesperson emphasized that Anthropic will continue to use hardware from other companies alongside its own designs as it scales up. The move is part of a broader trend among AI firms to develop custom silicon for improved performance and reduced reliance on third-party suppliers like Nvidia.
The decision comes amid growing concerns over the industry’s dependence on Nvidia for AI hardware. Anthropic’s spokesperson noted that while the company will still take a multi-chip approach, it aims to bring more silicon expertise internally. This strategy aligns with efforts by competitors like OpenAI, which recently unveiled its Jalapeño chip in partnership with Broadcom, and Google, which has long used its own hardware for models. Meta and Mistral are also reportedly exploring similar paths. Anthropic’s goal is to co-design new hardware and models simultaneously, building on past collaborations with partners but now focusing on internal development.
Anthropic’s spokesperson highlighted that the company is still in the process of hiring key team members, so any tangible benefits for users or models are expected to take time. The move is seen as a strategic step to enhance the performance of its frontier models and provide a competitive edge in a rapidly evolving AI landscape. The company’s focus on vertical integration suggests a desire to optimize both software and hardware for better efficiency and scalability. Source: arstechnica
