Meta is frustrated. After plowing $45 billion into the Metaverse only to see limited adoption and mounting losses, the company placed big bets on AI with the launch of Llama in 2023. This year, it is investing some $65 billion in capital expenditures with much of that going to AI and data center initiatives.
The company has also struggled to retain AI talent as 4.3% of its AI employees departed for rival AI labs in 2024, the second-highest attrition rate in the industry behind Google’s (not including DeepMind) 5.4%, according to SignalFire’s 2025 State of Tech Talent Report. With only a 64% two-year retention rate, Meta lags behind competitors like Anthropic (80%).
A theme of exodus
Now Meta’s $14.3 billion Scale AI deal has triggered an industry backlash, with Google abandoning its $200 million annual partnership with the data-labeling startup over concerns about sharing strategic information with a Meta-aligned vendor. On June 12, Meta said it would take a 49% non-voting stake of Scale AI, which has cemented a reputation as a reliable provider of data labeling services for training large language models and other AI systems. Its customers have ranged from tech and AI companies like Microsoft, Google, Meta, and OpenAI as well as government agencies like the U.S. Army and the U.S. Air Force as well as the law firm DLA Piper.
The fallout extends beyond Google. According to Reuters report, Elon Musk’s xAI also plans to discontinue its use of Scale’s services although Reuters reported that OpenAI planned to continue working with the firm. This industry-wide exodus threatens Scale AI’s core business model, which relies on a handful of major AI labs as clients. Competitors like Labelbox and Handshake are already reporting surge in business from clients fleeing Scale AI.
Meta’s talent churn
High-profile departures have compounded Meta’s challenges. Joelle Pineau, VP of AI Research and head of FAIR (Fundamental AI Research), announced her departure in April 2025. Multiple unnamed senior researchers have left for OpenAI and Anthropic despite highly-competitive compensation offers.
The exodus stems from deeper organizational problems: Meta has been sidelining its FAIR in favor of product-focused teams, frustrating researchers seeking breakthrough innovations. The lukewarm reception of Llama 4 (and allegations of benchmark hacking) in April 2025 further damaged morale, with employees questioning whether Meta can compete with OpenAI and Google in fundamental AI capabilities.
A $14.3 billion acqui-hire and eight-figure offers?
Meta’s $14.3 billion investment for a 49% stake in Scale AI represents the culmination of an unprecedented recruitment strategy. Sources familiar confirm that the primary driver for the massive investment was to secure Scale AI CEO Alexandr Wang to lead Meta’s new “superintelligence” unit. At this price, it may be the most expensive talent acquisition in tech history, effectively doubling Scale AI’s valuation to $29 billion.
Faced with AI brain drain, some media reports suggest that Mark Zuckerberg is abandoning Meta’s traditional hiring playbook. The company identified approximately 50–100 elite AI researchers for aggressive pursuit, with Zuckerberg personally leading the charge through methods unprecedented for a Fortune 500 CEO, according to a report from The Register.
“I got an email from Mark personally,” one AI researcher told The Register. “And he said, ‘I have an offer for you.’ Wow, and the offer was crazy” — at least $10 million per year. The researcher, who requested anonymity, said the offer didn’t even specify what the job entailed. Other AI labs are paying similar salaries.
Other media reports suggest Meta is willing to shell out nine-figures for some candidates. In a recent podcast episode of “Uncapped with Jack Altman” featuring Sam Altman, the OpenAI CEO said, “Meta thinks of us as their biggest competitor.” He added that that a “bunch” or people at OpenAI had received “giant” offers from Meta to “a lot” of OpenAI employees. “$100 million signing bonuses” and even higher annual compensation packages.
Zuckerberg’s tactics reportedly include sending top AI researchers personal emails and invitations to join a WhatsApp group called ‘Recruiting Party. He is also said to host recruitment meetings at his private residences in Lake Tahoe and Palo Alto. Zuckerberg is also said to.be physically rearranging Meta’s Menlo Park headquarters so new AI hires sit near his office while targeting senior researchers from top labs.
Notable successes include recruiting Jack Rae, a principal researcher from DeepMind, to lead pre-training efforts. Yet not all attempts succeed—Google countered Zuckerberg’s pursuit of DeepMind CTO Koray Kavukcuoglu by promoting him to SVP reporting directly to CEO Sundar Pichai.
The talent shortage is expected to persist until at least 2027. Bain & Co. projects the US could face a 700,000 worker reskilling need by 2027, while global estimates suggest an 85 million talent shortage by 2030 across all tech sectors.
