Anthropic confirms in-house AI chip team as custom silicon push expands

The Claude developer says it will pursue a multi-chip strategy, keep relying on AWS, Google, Nvidia and AMD, and hire seasoned semiconductor engineers for custom silicon work.

Summary

Anthropic has publicly confirmed for the first time that it is building an internal silicon team to design custom chips for its Claude model, formalizing a move previously reported by Reuters and other outlets as AI labs seek relief from the cost and energy limits of general-purpose GPUs. The company said it will use a software-hardware co-design approach and a multi-chip strategy, with AWS, Google, Nvidia and AMD remaining central to its compute stack rather than being replaced. Anthropic has also started hiring experienced chip engineers, with one listing for its Custom Chip Team offering $320,000 to $485,000 and seeking candidates with tape-out and shipment experience. The push comes as OpenAI advances its own Broadcom-built chip effort and as competition for AI talent and compute capacity intensifies across the sector.

Terms & Concepts
  • custom silicon: Chips designed for a company's specific computing workloads rather than broad general-purpose use.
  • software-hardware co-design: An approach that develops software and chips together so each is optimized for the other.
  • multi-chip strategy: A plan to use several types of processors or suppliers instead of relying on a single chip platform.