Anthropic is recruiting engineers to design custom silicon chips for its artificial intelligence models, according to recent job listings. The company aims to co-design this hardware with its software to improve speed and efficiency, following reports that it previously considered Samsung as a manufacturing partner. This move addresses the growing strain on external infrastructure as demand for Claude services increases. While Anthropic maintains agreements with Nvidia, AMD, Google, and AWS, relying solely on third-party hardware limits its ability to scale operations effectively. Other major players have already adopted similar strategies, with OpenAI using Broadcom chips, Google DeepMind relying on TPUs, and Meta developing its own accelerators. The shift signals a broader industry trend where builders of large language models seek greater control over their underlying technology to meet rising computational needs.
- Target role is within the custom silicon team
- Previous partner consideration included Samsung
- Current hardware partners include Nvidia and AMD
