Anthropic hires custom silicon team to design Claude chips

Anthropic confirmed the existence of its custom silicon team on August 5, 2026, and is actively recruiting to take its first chips from initial specifications to full production. According to job listings posted on the company's careers portal, the AI startup is seeking a Silicon Engineer and a Technical Program Manager (TPM) for Silicon. The Silicon Engineer role, based in San Francisco, New York, and Seattle, offers a salary range of $320,000 to $485,000 per year. Meanwhile, the TPM position, located in San Francisco and New York, carries a salary band of $365,000 to $435,000 per year. These hires will help Anthropic co-design hardware alongside its Claude models to boost operational speed and efficiency.

Rather than building a massive, self-contained semiconductor division, Anthropic is opting for a lean, partner-heavy program by design. The internal team will retain design authority while coordinating a broad network of external contractors, including ASIC design services, intellectual-property vendors, foundries, and packaging firms. Though no manufacturing partners have been officially named, reports indicate Samsung has been discussed as a potential foundry. The TPM candidate must have at least eight years of experience and have guided at least one chip through tapeout and production. The engineer role will span eight key domains, including front-end design, pre-silicon verification, physical design, design for test, analog and mixed-signal, technology and foundry, design infrastructure, and packaging.

For AI practitioners and hardware engineers, this development highlights a growing industry trend where software and hardware are built in tandem. By tailoring silicon specifically to the mathematical and memory demands of Claude, Anthropic aims to optimize performance while maintaining a multi-chip approach that still utilizes third-party hardware. Furthermore, the company expects these teams to utilize AI-assisted engineering, effectively using Claude to help design the very chips it will run on. This co-design strategy could ultimately lower API costs and improve latency for developers relying on Anthropic's models.