Anthropic has officially confirmed the creation of an in-house team to design custom chips for its models. This is the company's first public step toward independent development of semiconductor solutions. According to Business Insider, an Anthropic representative said that engineers will jointly design hardware and algorithms to achieve faster and more efficient performance from Claude.
Why this matters
Moving to custom chips is not just a way to save money, but a strategic move. Instead of adapting models to standard hardware, Anthropic will be able to design microchips for the specific needs of Claude. This should speed up computations and reduce scaling costs. The company emphasizes that it does not intend to abandon its existing infrastructure. As before, AWS and Google servers, as well as Nvidia and AMD accelerators, will play a key role. Custom developments will simply become another element of a larger ecosystem.
Who they are looking for
Anthropic's published job posting mentions the creation of a "custom silicon team." The company needs engineers with experience in chip design and verification. Hands-on experience is critically important: the candidate must have personally participated in the release of semiconductor products. The requirements are strict: they need someone who has "shipped silicon," has a realistic view of timelines, and is ready to make decisions independently without relying on a large support team.
The salary for such a position ranges from $320,000 to $485,000 per year. This is comparable to the level of leading specialists in the industry and confirms the seriousness of the company's intentions.
Industry context
Anthropic has joined a group of AI labs that have decided to take chip development under their own control. In June, OpenAI announced the Jalapeño chip, developed together with Broadcom. Meta plans to put its new chip into production in September. The founder of Mistral also does not rule out the appearance of its own silicon in the future.
Designing AI chips is an expensive and labor-intensive process, but it opens up opportunities for deep optimization. Companies gain control over performance and become independent of the whims of the market. It seems that this is the new standard for major players in the field of artificial intelligence.
