CLIVE CHAN to CLAUDE FC — HERE WE GO ✅

CLIVE CHAN to CLAUDE FC — HERE WE GO ✅ The man who helped build OpenAI's custom silicon from the ground up has signed for Anthropic. 2.4 years. Employee No. 002. A 10GW accelerator program. Gone. #AILeague

OpenAI's second-ever hardware hire, the chip engineer who spent two and a half years building the GPU killer in-house, just walked across the street to the competition. Not a researcher. Not a product manager. The person who was actually designing the silicon.

Clive Chan graduated from the University of Waterloo in 2021 with a software engineering degree. 1 His pre-draft career reads like a scouting report written to impress: machine learning infrastructure at Google, liquid-rocket propulsion software at SpaceX, quantum computing at QuEra. Then Tesla's Autopilot team, three years running GPU optimization and cluster scheduling on the Dojo training infrastructure.

When OpenAI stood up its custom AI chip program in early 2024, Chan was the second hardware engineer they hired. Employee No. 002. 2 He joined in January 2024 and spent the next 2.4 years on matmul performance, roofline analysis, and the architecture of a chip program that OpenAI had been quietly building toward for years.

His read on the team he was leaving: "The density of hardware talent on that team is extraordinary, and I don't think there's a better chip design team anywhere." 1

That's the man who just signed for Claude FC.

On June 6, 2026, Chan posted a personal update on X: he'd decided to leave OpenAI and had already joined Anthropic that week. 3 The reason he gave — "I haven't been able to shake the pull to climb a new mountain from the bottom again" — is the kind of line a club chairman quotes at a press conference when a star midfielder walks out two weeks before the league final.

The timing is hard to ignore. OpenAI and Broadcom had publicly announced a multiyear partnership to deploy a 10GW AI accelerator system, with first rack deliveries targeted for the second half of 2026. 1 That program — the one Chan helped architect from the inside — is now running without him, weeks before the hardware is supposed to ship.

At Anthropic, his LinkedIn role title currently reads: "perplexity per picojoule." That phrase tells you exactly what he's been hired to solve: how much useful model output you extract per unit of energy burned. The goal is achievable either by writing better software for existing TPUs and Amazon chips, or by building custom silicon from scratch. 2 As of April 2026, Reuters reported that Anthropic was weighing whether to design its own chips, with no dedicated team yet assembled. Chan's arrival changes that calculus.

Anthropic closed a $65 billion Series H and confidentially filed for a U.S. IPO on June 1, 2026, at a reported valuation of approximately $965 billion. 4 OpenAI filed its own confidential IPO paperwork in late May at $852 billion. Both clubs are heading to the public markets at the same time, and pre-IPO equity is the most powerful signing-fee currency in the league.

Anthropic's internal retention rate over two years runs at roughly 80%. 5 OpenAI engineers are reportedly 8x more likely to leave for Anthropic than the reverse; DeepMind engineers, 11x. At current trajectory, Claude FC isn't just signing players — it's running a deliberate talent press. Chan follows Andrej Karpathy (May 2026, pre-training) through the same exit door, and both chose the same destination.

OpenAI's hardware ambitions are real and expensive. The Broadcom partnership targets 10 gigawatts of AI accelerator capacity — an infrastructure bet that dwarfs anything the league has attempted before. Losing the second employee on that program weeks before first hardware delivery is not a disaster, but it is a signal: someone who was deep inside that effort decided the opportunity cost of staying was too high.

The chip program will continue. The technical work Chan started won't disappear overnight. But in a domain where institutional knowledge of matmul performance and hardware co-design accumulates slowly, replacing an employee with 2.4 years of context is harder than replacing a researcher with 2.4 years of papers.

In 2016, Kevin Durant left the Oklahoma City Thunder — the team he'd built a Western Conference finals run with, the highest-scoring environment he'd ever played in — to join the Golden State Warriors. His stated reason was growth. His actual calculation was that he could win faster somewhere else. Observers called it a betrayal. He called it a mountain.

Chan's move is structurally similar. He's leaving what he himself described as the best chip design team in the world to go somewhere that doesn't yet have a dedicated chip team at all. That's not a transfer driven by discontent. It's a transfer driven by the kind of ambition that wants to build something from nothing, and a belief that the destination has the resources and pace to make it worth it.

"Already energized by the pace and intensity of my first days at the company," he wrote. "It's time to build." 3

#AILeague