The Three People Who Knew How to Spend 100 Billion on AI Infrastructure Just Left OpenAI for Meta
The three former OpenAI infrastructure leaders who built Stargate are now at Meta. What they know how to do is harder to build than any data center.

image from grok
Three senior OpenAI executives who held the complete execution chain for the $100 billion Stargate data center program—Peter Hoeschele (strategy/operations), Shamez Hemani (engineering execution), and Anuj Saharan (supply chain logistics)—departed on the same day (April 9) to join Meta's compute infrastructure division. The departure represents a significant brain drain for OpenAI, which is simultaneously pivoting from its 'build-to-own' Stargate model to renting cloud capacity from providers, while Meta is committing $115–135 billion in capital expenditures for AI infrastructure over 2026–2027. The hires give Meta institutional knowledge of large-scale AI infrastructure execution that no amount of capital can easily replicate.
- •The three executives held complementary roles covering the entire $100B infrastructure execution chain—strategy, engineering, and supply chain—making their simultaneous departure uniquely impactful.
- •Meta's lack of an existing cloud business to monetize infrastructure spending makes this talent acquisition critical; they need the operational expertise that OpenAI spent years developing.
- •OpenAI's shift from building its own data centers to renting cloud capacity aligns with its revised commercialization timeline ahead of a potential IPO, likely explaining why the expertise became expendable internally.

