Developer Tools

Anthropic builds custom chips for Claude to cut Nvidia reliance

Anthropic hires silicon team as it joins OpenAI, Google in designing bespoke AI hardware.

Deep Dive

Anthropic has confirmed plans to design its own semiconductor chips to power its Claude models, joining OpenAI, Google, Meta, and others in reducing reliance on Nvidia’s hardware. The company posted job listings for senior silicon engineers and technical program managers, signaling a push toward custom silicon development. While Anthropic will still use multi-chip architectures and may collaborate with partners like Samsung, the move aims to enhance performance through vertical integration—co-designing hardware and models together.

This strategic shift addresses two critical challenges: Nvidia’s dominance in AI infrastructure and the competitive need for optimized hardware-model pairings. Anthropic’s approach mirrors rivals like OpenAI, which recently unveiled its Jalapeño chip in partnership with Broadcom. However, Anthropic’s initiative is still in early stages, with key hires and development timelines yet to be finalized. The long-term goal includes leveraging bespoke silicon to improve efficiency and potentially enable deployment on edge devices or smaller models.

Key Points
  • Anthropic is hiring a custom silicon team to design chips for Claude, reducing Nvidia dependency.
  • The company plans a multi-chip approach, possibly partnering with Samsung, and co-designing hardware with models.
  • This move aligns with OpenAI, Google, and Meta, aiming for better performance and competitive edge.

Why It Matters

Custom hardware could lower costs, improve performance, and reduce reliance on Nvidia for AI providers and developers.

📬 Get the top 10 AI stories daily