Anthropic builds custom AI chips to speed up Claude
Anthropic is hiring engineers to design chips for faster AI training and inference.
Anthropic, the maker of the Claude AI models, is assembling a team to design its own custom AI chips to accelerate performance and reduce reliance on third-party hardware. The company confirmed the move to TechCrunch after Business Insider first reported the initiative, noting plans to co-design hardware and models for faster, more efficient AI workloads. Anthropic has already partnered with AWS, Google, Nvidia, and AMD for cloud computing, but sees internal chip development as critical to scaling Claude amid surging demand.
The hiring push targets engineers experienced in chip design for Anthropic’s new 'custom silicon team,' as revealed in a job listing. This strategy mirrors moves by competitors like OpenAI, which unveiled its Broadcom-built Jalapeño chip for inference tasks in June. Google DeepMind has long used Alphabet’s TPU chips, while Meta develops its MTIA accelerators. Anthropic’s potential partnership with Samsung, per a prior report by The Information, underscores the industry’s shift toward proprietary hardware to gain a performance edge.
- Anthropic is hiring chip design engineers for a new 'custom silicon team' to build AI-optimized hardware
- The move aims to co-design chips and models for faster, more efficient AI workloads, reducing reliance on partners like AWS and Nvidia
- Anthropic follows OpenAI (Jalapeño chip), Google (TPUs), and Meta (MTIA) in developing proprietary AI accelerators
Why It Matters
Custom AI chips could give Anthropic a competitive edge in performance and cost efficiency for scaling Claude globally.