Anthropic Launches Custom Silicon Team to Reduce Nvidia Dependency and Compete with OpenAI

Summary

Anthropic is building an in-house custom silicon team to design chips optimized for running Claude and other AI models, marking a significant strategic shift for the AI company. The initiative was confirmed by an Anthropic spokesperson after Business Insider spotted job listings for senior semiconductor engineers and technical program managers on the company's careers page. This move mirrors efforts by competitors including OpenAI (which recently announced its Jalapeño chip), Google, Meta, and Mistral, reflecting a broader industry trend toward vertical integration in AI infrastructure.

The company's primary motivation is to reduce its reliance on Nvidia, which currently dominates the market for AI training and inference hardware. By designing chips tailored specifically for Claude's architecture, Anthropic aims to achieve better performance and cost efficiency while maintaining independence from Nvidia's supply chain constraints. An Anthropic spokesperson clarified that the company will still employ a "multi-chip approach," using hardware from other manufacturers alongside its custom designs as it scales, and will continue its co-design partnerships with external hardware partners.

The strategic move also reflects broader competitive pressures in the AI industry, where custom silicon can provide performance advantages that are difficult for competitors to match. Anthropic plans to co-design hardware and software models in parallel, bringing silicon expertise in-house to optimize new model architectures for its custom chips. However, the company is still in the early hiring phase, meaning actual silicon products are likely years away from deployment.

• OpenAI, Google, Meta, and Mistral are pursuing similar custom silicon strategies, signaling an industry-wide shift
• Anthropic is currently in early hiring stages; significant products are likely years away