Anthropic assembles in-house chip design team to scale Claude

Anthropic is recruiting a dedicated team to develop proprietary semiconductors tailored to its AI needs. The company behind Claude plans to align hardware and model design so its technology can operate with greater speed and efficiency.

Building its own silicon#

Earlier reporting indicated that Anthropic had been exploring Samsung as a manufacturing partner for custom chips. Now, active job postings reveal the company is looking for engineers with chip design expertise to join a "custom silicon team."

This move reflects a growing pattern among major AI labs. OpenAI released its Broadcom-manufactured Jalapeño chip in June, built specifically for inference operations. Google DeepMind leverages Alphabet's TPU processors, while Meta is building its own MTIA accelerators.

Why the push for custom hardware#

Demand for Claude is climbing sharply, and the computational bottleneck is tightening. While Anthropic has secured access to infrastructure from AWS, Google, Nvidia, and AMD, those existing partnerships apparently cannot scale fast enough to keep pace with customer needs.

By designing its own chips, Anthropic can optimize hardware directly for its model architecture—eliminating waste and gaining a competitive edge in cost and performance. Custom silicon represents a critical lever for AI companies seeking to reduce dependence on third-party suppliers and improve their unit economics as inference traffic scales.