Barret Zoph co-founded Thinking Machines Lab with Mira Murati, left before the startup launched, briefly joined OpenAI, and is now at Google. The whole arc took under a year. This is not gossip. It is a structural diagram of how AI talent actually moves: from startup to hyperscaler, with the intermediate stop at a rival hyperscaler serving as leverage. The startup was the credential. The credential was always for Google.
The Revolving Door as Infrastructure
OpenAI's data center chief departure, flagged by Fast Company as a bad sign, fits the same pattern. The people who build the most critical infrastructure at AI companies are also the most portable. Data center architecture and model training expertise are skills that every hyperscaler needs and will pay to absorb. The talent market is functioning less like a labor market and more like an acquisition pipeline without the acquisition price. Google, Microsoft, and Amazon essentially have standing offers for anyone who exits a frontier AI company. The startup ecosystem creates the training ground. The hyperscalers harvest it. TurboFund's investor signals dashboard tracks how VC attention shifts when key founders exit startups like this.
Nvidia, Salesforce, and the AI P&L
This week's Nvidia and Salesforce earnings beat adds a financial layer: the AI infrastructure trade is printing money for the companies that sell the picks and shovels, while the companies that use the picks and shovels to mine for gold are hemorrhaging talent. A 2025 paper in Strategic Management Journal by Greve and Cobb found that executive departures from startups in concentrated industries accelerate consolidation toward incumbents. The revolving door isn't a symptom of instability in AI. It is the stability mechanism, keeping the hyperscalers staffed and the frontier companies perpetually re-founding themselves.