By Stuart Kerr, Technology Correspondent, LiveAIWire
The AI talent arms race reached a genuinely unprecedented figure in mid-2025, when OpenAI CEO Sam Altman revealed on the podcast Uncapped that Meta had offered some of his top researchers signing bonuses of 100 million dollars, with total compensation packages running higher still. Altman said none of his best people had taken the offer, but the number itself marks a new ceiling for what a single AI researcher can command, and it is far from an isolated case. Meta CEO Mark Zuckerberg has personally led recruitment for a new Superintelligence Labs division, reportedly offering one researcher, Andrew Tulloch of Thinking Machines Lab, a package worth as much as 1.5 billion dollars over six years. Tulloch turned it down.
The scale of these numbers reflects genuine scarcity rather than simple corporate excess. There are only a small number of researchers globally capable of building and improving frontier AI models, and every major lab, OpenAI, Google DeepMind, Meta, Anthropic, and a growing list of well-funded startups, is competing for essentially the same narrow pool. Meta’s own investment strategy makes the logic explicit: the company paid up to 14.3 billion dollars for a 49 percent stake in Scale AI specifically to bring the startup’s CEO, Alexandr Wang, into its leadership team.
Who Is Actually Winning the Retention War
Compensation numbers grab headlines, but retention data tells a more useful story about where AI talent actually wants to work. SignalFire’s 2025 State of Talent Report, tracking over 650 million professional profiles, found that Anthropic retains 80 percent of employees hired at least two years ago, the highest figure among major AI labs. Google DeepMind follows at 78 percent, while OpenAI trails at 67 percent, roughly on par with large established tech companies like Meta at 64 percent. The same report found engineers are eight times more likely to leave OpenAI for Anthropic than the reverse, and the ratio against DeepMind runs even higher, at roughly eleven to one in Anthropic’s favour.
Why Money Alone Isn’t Winning This
Anthropic’s retention advantage is notable precisely because the company is not the highest payer in the field. SignalFire’s analysis attributes the edge to researcher autonomy, a flatter organisational structure without forced management tracks, and a working culture that appeals specifically to researchers who want intellectual independence over the largest possible paycheck. That does not mean compensation is irrelevant. It means that once pay reaches a certain extraordinary threshold, the differentiator for where elite researchers choose to stay becomes culture and mission alignment rather than incremental salary increases.
What This Means If You’re Trying to Break Into AI
For anyone hoping to enter the field, the talent war at the very top tells a misleading story about the job market as a whole. SignalFire’s data shows new graduate hiring at Big Tech companies has fallen by more than half compared with pre-pandemic levels, with new grads now accounting for just 7 percent of hires, down 25 percent from 2023 alone. The extreme compensation packages making headlines apply to a small number of already-established researchers with rare, proven expertise, not to entry-level roles. For early-career candidates, the practical path into AI increasingly runs through demonstrable project work, open-source contributions, and specialised technical depth rather than credentials alone, since companies are explicitly prioritising proof of capability over potential.
The Human Cost Behind the Numbers
The intensity of this competition carries real organisational costs beyond payroll. Google DeepMind has reportedly enforced six-to-twelve month noncompete clauses on some departing researchers, paying full salary while they sit out before joining a competitor, a retention tactic aimed squarely at slowing the flow of institutional knowledge to rivals. The pattern of senior researchers clustering around specific leaders rather than specific companies has also intensified. When OpenAI’s former CTO Mira Murati departed to found Thinking Machines Lab, nineteen colleagues followed her, illustrating that loyalty in this market increasingly attaches to individuals and missions rather than employer brands.
None of this suggests the AI talent war is close to cooling. Every major lab remains capital-rich enough to keep offering extraordinary packages, and the underlying scarcity of researchers capable of building frontier models has not eased. What the retention data does suggest is that money alone does not determine who wins this competition long term. The labs combining serious compensation with genuine researcher autonomy, as our own coverage of how AI is reshaping the broader workplace has found in other contexts, are proving more durable than those relying on cash alone to keep talent in place.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.