Anthropic to Develop Its Own Chips for Claude [Global AI Briefing]

[Financial News] Anthropic has set up an in-house team to design custom chips for its artificial intelligence (AI) model Claude.
Following OpenAI and Meta Platforms, global AI giants are now moving in unison to design their own chips for AI models, intensifying competition in custom semiconductor design.
On the 5th local time, Anthropic said it is "building an in-house silicon team to design custom chips for Claude," formally confirming for the first time its previously reported chip development plan.

Anthropic said the custom chip project will proceed through a software-hardware co-design approach. The company plans to develop the chip architecture and the Claude model at the same time so that each can influence the other.
In a job posting for its chip design team, Anthropic said it is looking for staff to support "first silicon prototype validation and debugging." The move suggests the company is building a team aimed not just at performance testing, but at actual mass-production chip design. Salaries are in the range of $320,000 to $485,000, and the company is seeking experienced professionals across the full spectrum of semiconductor design and verification.
Anthropic stressed that its chip design plan does not mean it is abandoning its existing supply chain.
The company said it will continue using hardware from AWS, Google, NVIDIA, and AMD while maintaining a multi-chip strategy that includes in-house chip design. Its own chips, it added, are simply an additional track within that strategy.
In fact, Anthropic has already secured about 3.5 gigawatts (GW) of custom TPU capacity starting in 2027 through long-term contracts with Google and Broadcom Inc.
After OpenAI unveiled its inference-only chip "Jalapeño" with Broadcom in June, Meta Platforms and Anthropic have joined the race, meaning all three frontier AI companies are now competing to design their own custom semiconductors.
Industry observers say AI companies are directly entering custom semiconductor design as a strategy to improve the efficiency of AI models. The goal is to make AI systems run faster and more efficiently as user demand surges.
Experts also say the strategy reflects an effort to cut inference costs per token by aligning chip architecture with AI model design.
[email protected] Lee Gu-soon Reporter