Meta hires another ChatGPT alumnus, and the whole industry pays the retention bill

A signing bonus is a one-time entry. A market price is not. Andrew Tulloch reportedly agreed to $1.5 billion over six years to join Meta, a figure The Next Web said would make him the most expensive individual hire in tech history if accurate ^[7]. That averages $250 million a year ^[1], and its main use to Meta's competitors is as a benchmark: once a package like that exists, every senior researcher's retention conversation starts there. Sam Altman has said Meta offered $100 million bonuses to OpenAI staff ^[8]. The counter-offers written to keep those people are a cost Meta created and someone else booked.

The Scale AI deal priced the same asset at institutional scale. Meta paid $14.3 billion for 49%, which implies a valuation near $29.2 billion ^[6][4], and the transaction delivered Alexandr Wang to run Superintelligence Labs ^[6]. How the instrument trades is visible in the Thinking Machines diaspora: five of Mira Murati's founding team went to Meta, three went back to OpenAI, and one to Elon Musk's xAI ^[17].

Ruoming Pang is the cleanest illustration of the accounting problem. Meta recruited him from Apple on a package Bloomberg valued above $200 million, and he oversaw AI infrastructure for the lab ^[15]. Seven months later OpenAI had him, according to The Information in February as relayed by Reuters ^[16]. That is roughly $28.6 million per month of service ^[3], and the counterparty acquired an infrastructure lead who had already been paid to learn Meta's stack.

Apply the 19% attrition rate Sam Jones measured, the highest of the four frontier labs he examined ^[11], to the departure count he found, and Meta's research population comes out near 4,100 ^[2]. Jones called it "the most expensive treadmill in the industry" and concluded that "hiring is the easy half" ^[13]. The traffic runs in both directions: he found Meta losing researchers to Microsoft AI, Nvidia and OpenAI while replacing them from Amazon, Scale AI and universities ^[14].

Meta can carry this. Second-quarter 2026 revenue of $60.8 billion, up 28% year over year in its investor filing ^[9], funds a payroll that does not compound into headcount. A lab financed by rounds rather than a $60.8 billion quarter has to match the same price by selling equity, which is how one company's hiring strategy becomes another company's dilution. Metz's arrival will be settled in models, not in the fact of the arrival.