Meta pulls OpenAI veteran into Superintelligence Labs talent war

Meta has added another high-profile AI researcher to its Superintelligence Labs. Luke Metz, who was part of the early team behind ChatGPT at OpenAI, has joined the company. This marks a new high in the war for AI talent that is changing leading laboratories around the world.

While it may not be significant as an isolated victory in the AI sector for Metz, his move may reflect a broader trend. Prominent researchers in the field are switching between competing research labs every few months, meaning that every hire in the market could simultaneously represent a loss to a rival firm. According to reports, Metz will report to Alexandr Wang, former head of Scale AI and current head of Meta’s AI operation.

A three-lab loop in under two years

Metz’s recent career shows how quickly that revolving door is spinning. On his personal site, he describes himself as an AI researcher who was part of the initial team behind OpenAI’s “low-key research preview” that became ChatGPT. Before that, he was a research scientist at Google Brain.

He left OpenAI in 2024 for Thinking Machines Lab, the startup founded by former OpenAI chief technology officer Mira Murati, Axios reported.

That stint was short. Fortune reported in January 2026 that Metz was returning to OpenAI alongside co-founder Barret Zoph and founding-team member Sam Schoenholz. Simo said Zoph would report to her, while Metz and Schoenholz would report to Zoph. Months later, Metz is moving again, this time to the company several of his former Thinking Machines colleagues had already joined.

Meta bought its bench at record prices

It has been reported that Metz will be part of a team formed by Meta at a tremendous price. The company made an investment of $14.3 billion to acquire a 49% stake in Scale AI and appointed Wang to run the Meta Superintelligence Labs. Five of Murati’s founding team members went to Meta, three returned to OpenAI, and one will work for Elon Musk’s xAI company.

Moreover, the compensation associated with such moves is also excessive. Thinking Machines co-founder Andrew Tulloch reportedly accepted a $1.5 billion package over six years to join Meta, which The Next Web said would rank as the most expensive individual hire in tech history if accurate. Cryptopolitan reported last year that OpenAI’s Sam Altman said Meta had offered $100 million bonuses to his staff.

Meta can afford the spending. It reported second-quarter 2026 revenue of $60.8 billion, up 28% year over year, according to its investor filing.

The spending has not grown the team

Counting the number of employees reveals a more complex situation. Through his August 2026 analysis of LinkedIn Talent Insights data, recruiter Sam Jones discovered that Meta hired 778 research scientists throughout the year, but 785 left the company, thus leaving a deficit of seven employees. Meta had a 19% attrition rate, the highest of all four frontier labs examined by Jones. At the time when the research was carried out, there were 1,838 open research positions at Meta.

Jones summed up the mismatch between Meta’s recruiting push and its retention:

“Meta ran the most expensive treadmill in the industry — 778 on, 785 off — proving that in this market, hiring is the easy half.”

The traffic runs in both directions. Jones found Meta researchers leaving for Microsoft AI, Nvidia and OpenAI even as the company recruited from Amazon, Scale AI and universities. OpenAI, meanwhile, hired Ruoming Pang, who had overseen AI infrastructure for Meta’s Superintelligence Labs, The Information reported in February, per a Reuters account, seven months after Meta recruited him from Apple on a package Bloomberg valued at more than $200 million.

What Meta has shipped so far

The first model to come out of Meta’s recently opened lab, which is Muse Spark, was launched on April 8. It is currently used by Meta AI in its technologies across Facebook, Instagram, WhatsApp, and Ray-Ban eyewear. The only thing left to be seen is whether the new employees, such as Metz, will help Meta to convert the costs of recruitment into models that would give it a chance of competing with its rivals such as Google and OpenAI. Unfortunately, Meta has already spent a lot of money on hiring while keeping its research team size close to the initial team size.

The smartest crypto minds already read our newsletter. Want in? Join them.