Anthropic hires Amir Salek, who ran Google's TPU business until 2022, to join its compute team as part of a push to develop its own chips
Anthropic PBC has hired Amir Salek, a founder of the custom chip program at Alphabet Inc.'s Google, as the AI lab lays the groundwork for a push into making its own semiconductors.
Context & Ripple Effects
Anthropic had already confirmed a dedicated in-house silicon team to co-design chips and Claude using a multi-chip approach. Hiring Amir Salek adds TPU-business experience to that effort as it moves from team formation toward execution.
The chip push sits alongside Anthropic’s existing dependence on external capacity: its Google and Broadcom TPU-capacity agreement supplies next-generation compute while Anthropic develops its own hardware capability.
First-order effects
- Anthropic’s compute team gains a leader with experience from Google’s TPU program, strengthening its ability to organize a custom-semiconductor effort for Claude.
- Anthropic can pursue internal chip design without displacing its current Google and Broadcom capacity arrangements, which remain its immediate source of large-scale compute.
Second-order effects
- Google and Broadcom retain Anthropic as a major compute customer in the near term, but Anthropic’s internal design capability gives it more leverage over the mix of hardware it uses over time.
- The hire intensifies competition for specialized AI-chip leadership between model developers and cloud platforms, where talent spans chip design, supply coordination, and workload requirements.
Third-order effects
- If model labs continue building hardware teams while contracting for outside capacity, AI infrastructure is likely to settle into heterogeneous stacks rather than a single-provider compute model.
- Custom-chip expertise becomes a strategic moat for AI developers: the firms able to co-design models and silicon can exert more control over performance, capacity planning, and supplier dependence.
The trend: AI model developers are pairing external compute contracts with in-house silicon teams to gain greater control over the hardware stack.