Anthropic custom silicon AI hardware OpenAI IPO

Anthropic Snaps Up OpenAI's Second-Ever Chip Engineer Ahead of Rival IPOs

9 June 2026

Anthropic has hired Clive Chan, who was the second hardware employee in OpenAI's custom chip program, as both companies prepare for IPOs. The move comes as Anthropic is reportedly exploring the development of its own AI chips, which could reduce its reliance on Google TPUs and Amazon hardware while improving profit margins. Chan's exact role at Anthropic is unclear, but his expertise in custom silicon design could help the company build a dedicated chip team.

Clive Chan, who claims to be the second person ever hired onto OpenAI's custom chip programme, has defected to Anthropic. The timing is notable: both companies are heading toward public listings, and Anthropic is reportedly sizing up whether to build its own silicon.

Chan was candid about his respect for what he left behind. In a public LinkedIn post, he praised the hardware talent density at OpenAI as something he'd not seen matched anywhere, and predicted the chips being built there would become central to the development of AGI. High praise. Then he handed in his notice.

At OpenAI, Chan was involved in building custom chips from the ground up and had a hand in the company's strategic partnership with Broadcom, a relationship that has apparently been bumpy due to disagreements over production costs and questions about OpenAI's creditworthiness.

What exactly Anthropic wants him for is anyone's guess. His LinkedIn job description reads "perplexity per picojoule", which is technically about squeezing maximum model performance out of minimum energy. That could mean optimising software stacks for existing GPUs and TPUs, or it could mean designing bespoke chips altogether. Both are plausible. Neither Anthropic nor Chan has been more specific.

The business case for custom silicon is fairly straightforward. Anthropic currently runs Claude on Google TPUs and Amazon hardware. That works, but it's expensive and it hands a degree of control to suppliers. Purpose-built chips, optimised for Anthropic's specific inference workloads, would reduce that dependency and improve margins over time. As AI shifts from research novelty to infrastructure commodity, margin efficiency starts to matter enormously.

Reuters reported in April that Anthropic was in early discussions about chip design, with no dedicated team in place yet. Chan could be the first brick in that wall, and he's arriving with hard-won knowledge from one of the few organisations that has actually attempted to build AI silicon at this scale.

Before OpenAI, Chan spent around two and a half years at Tesla's Autopilot division working on custom ML training chips, covering software framework bring-up, datacentre co-design, and energy-efficient number formats. He joined OpenAI in January 2024.

Anthropic recently signed a long-term agreement with Google and Broadcom as part of a stated commitment to put $50 billion into US computing infrastructure. Whether that deal is a precursor to in-house chip development or just a hedged supply arrangement remains to be seen. Either way, hiring someone with Chan's background sends a reasonably clear signal about the direction of travel.