Silicon Rebellion: Anthropic Recruits Hardware Engineers to Build Custom AI Chips

Anthropic is putting together a specialized team to design its own custom silicon for artificial intelligence workloads. Business Insider first broke the story, and Anthropic later confirmed the move. The company behind the Claude model wants to build in-house hardware to make its software run much faster and cut down energy costs across its data centers.

To pull off this hardware shift, Anthropic is actively hiring experienced chip design engineers for a newly formed custom silicon group. Reports from last month indicated that Anthropic approached Samsung as a potential manufacturing partner to fabricate these physical chips. By co-designing hardware directly alongside its software models, Anthropic aims to optimize processing power specifically for its proprietary algorithms.

The decision to build custom processors comes as public demand for Claude skyrockets. Every major artificial intelligence firm is racing to secure enough processing infrastructure to keep up with user traffic and model training. Up to this point, Anthropic relied heavily on third-party cloud providers and hardware manufacturers. The firm signed major infrastructure deals with Amazon Web Services, Google, Nvidia, and AMD to secure access to graphics processors and cloud capacity. However, as user demand climbs, buying off-the-shelf chips from external suppliers creates clear bottlenecks and drives up operational costs.

Anthropic is far from the only software lab taking control of its hardware supply chain. Competitors are pouring massive resources into proprietary processor designs to break their reliance on third-party suppliers. In June, OpenAI revealed its Jalapeño chip, built in partnership with Broadcom to handle specific software inference workloads. Google DeepMind relies on Alphabet custom Tensor Processing Units to power its model infrastructure, while Meta builds out its own MTIA accelerator family to handle internal workloads across its data centers.

Designing custom silicon lets companies optimize chip architecture for specific software tasks rather than using general-purpose processors. Standard graphics chips handle a wide range of tasks, which often wastes power and compute performance. Custom chips streamline memory handling, lower electrical demands, and boost output speeds specifically for large language models.

Building physical processors takes years of research, heavy capital investments, and careful manufacturing coordination. Still, control over hardware design gives major tech players a huge long-term edge. As Anthropic builds out its engineering team, designing custom chips will help the company secure its infrastructure, lower operating costs, and keep pace with rival software labs in the race for raw processing power.