Anthropic poaches OpenAI's second-ever chip engineer as both companies race toward IPOs

The detail worth sitting with is Chan's specific lineage: he didn't just work on chips at OpenAI, he was embedded in the Broadcom partnership that defines OpenAI's current inference silicon roadmap. Anthropic isn't just buying generic hardware expertise, it's acquiring someone who knows exactly how a direct competitor structures its foundry relationships.

This hire lands one week after Anthropic's confidential S-1 submission to the SEC (covered across multiple outlets on June 1), and the timing is unlikely to be coincidental. Public investors scrutinizing an AI lab's S-1 will ask hard questions about compute costs and margin trajectory. Demonstrating proprietary silicon capability, or at least credible progress toward it, is a direct answer to that line of questioning. OpenAI is running a parallel track, as its Michigan Stargate infrastructure announcement from the same week shows, but that bet is on scale of third-party and owned data center capacity rather than custom silicon design. Anthropic appears to be threading a different needle: smaller physical footprint, tighter inference optimization, and now the talent to execute on custom chip architecture before the IPO window closes.

Watch whether Anthropic discloses a custom ASIC program or a new foundry partnership in its public S-1 filing when it drops. If Chan's hire is followed within six months by a tape-out announcement or a named silicon partner, the IPO-chip strategy connection is confirmed. If neither surfaces before the offering, this was talent acquisition for future optionality rather than near-term differentiation.