Inside Meta’s High-Stakes Talent War for AGI Supremacy
Mark Zuckerberg’s plans to build an AGI Superintelligence team might have started in somewhat mysterious circumstances – but now, it is made up of some of the greatest AI experts in the world.
The mission is to achieve something that hasn’t been possible before – to build AI that matches or surpasses human intelligence, giving everyone a personal assistant that can help with daily life, boost creativity and empower users to achieve personal goals.
The team’s formation has triggered a talent exodus from other AI leaders, leaving the public to wonder about the implications of such powerful technology.
More urgently, it has left rival companies scrambling to understand why they are losing key staff and what, exactly, it will take to stop them.
As AI companies wonder who is next to join the team,“AI researchers now earn more than Fortune 500 CEOs while their code replaces millions of jobs,” according Jatin Modi, Global CEO at Renaissance.
In other words, Meta may be paying the highest liquid compensation to an individual contributor in the history of technology.
The bigger picture of Meta’s offerings to AI experts
Meta has rewritten the rules of technology compensation, by offering AI engineers packages worth “US$300m over four years,” according to Jatin, as the company builds its Superintelligence Labs division.
Mark Zuckerberg maintains what sources describe as a “literal list” of AI specialists he wants to recruit for Meta’s new research division.
So far, the company has sought out talent from rivals including OpenAI, Apple, Google and Anthropic.
On a deeper level, the compensation warfare additionally reflects a supply shortage.
Foundation model expertise – the knowledge required to build large language models (LLMs) that power AI systems like ChatGPT – remains concentrated among a small group of researchers who gained experience at major technology companies rather than universities.
How AI giants have responded to losing talent to Meta
Nowhere has the impact of Meta’s recruitment drive been more evident than at Apple, where a series of strategic hires has significantly weakened the tech giant’s internal AI team.
Ruoming Pang, who led Apple’s 100-person Foundation Models team responsible for powering Apple Intelligence features, joined Meta in July for a compensation package reportedly exceeding US$200m over several years.
His departure triggered a cascade of defections. Mark Lee and Tom Gunter, two senior members of his team, followed him to Meta.
In response, Google DeepMind has reportedly implemented stringent non-compete clauses to prevent researchers from joining competitors immediately.
Meanwhile, OpenAI has offered retention bonuses exceeding US$2m and equity packages worth more than US$20m to deter defections to rivals.
Inside Meta’s plan coming together
Traditional technology compensation structures have become obsolete in the AI talent market.
The new AI-focused packages are a different category entirely. Andrew Bosworth, Meta’s Chief Technology Officer (CTO), recently told employees the company has “a small number of leadership roles that we’re hiring for and those people do command a premium.”
“What bought Instagram in 2012, 13 employees and 100 million users, now buys a single mind hoping to make millions obsolete,” Jatin comments about the compensation inflation.
Meta’s recruitment drive also extends beyond individual hires to strategic acquisitions.
The company invested US$14.3bn to acquire a stake in Scale AI, primarily to secure founder Alexandr Wang as Chief AI Officer for Superintelligence Labs.
Alexandr, 28, now oversees Meta’s foundational AI research alongside the former Scale team, after becoming the world’s youngest self-made billionaire at just 25.
Meta further acquired NFDG, cofounded by Daniel Gross and Nat Friedman, to secure their services.
Daniel now leads Meta’s AI products division whilst Nat co-leads Superintelligence Labs with Alexandr.
The impact on OpenAI
Sam Altman, OpenAI’s CEO, has acknowledged Meta’s recruitment efforts while attempting to minimise its impact.
Speaking on the Uncapped podcast, he says Meta had been making “giant offers to a lot of people on our team” including “US$100m signing bonuses and more than that [in] compensation per year.
“I’m really happy that, at least so far, none of our best people have decided to take him up on that,” he says. Though he admits Meta has “gotten a few great people for sure” and “had to go quite far down their list.”
In response, OpenAI has implemented multiple tender offers allowing employees to sell shares to private buyers rather than waiting for a public offering or acquisition.
March Chen, OpenAI’s Chief Research Officer (CRO), told employees in an internal memo that executives were “recalibrating” compensation and considering additional reward mechanisms for valued staff.
Despite retention efforts, OpenAI has lost researchers including Lucas Beyer, who led the company’s Zurich office before confirming his move to Meta.
Furthermore, Pavel Izmailov, who worked on reasoning and safety research, joined Anthropic after reportedly being dismissed from OpenAI.
Jan Leike, who co-led OpenAI’s Superalignment team focused on AI safety, also moved to Anthropic in May last year, criticising his former employer for prioritising “shiny products” over building robust safety protocols.
“Missionaries will beat mercenaries,” Sam says in his staff memo.
Why Anthropic is emerging as another unexpected talent magnet
While Meta dominates headlines with massive compensation packages, Anthropic has quietly become another industry talent magnet through different strategies.
Anthropic maintains an 80% retention rate for employees hired over the past two years, according to SignalFire’s State of Talent Report.
The retention figures contrast sharply with Meta’s 64% rate and OpenAI’s 67%.
“The actual talent pool with meaningful foundation model experience is still very small,” says Deedy Das, a Principal at Venture Capital firm Menlo Ventures.
“Meta, OpenAI, Google DeepMind, Anthropic and xAI are all aggressively hiring, but the market has recognised that and compensation reflects it.”
However, Anthropic strategy to keep employees isn’t just about money.
The company also focuses on AI safety research and employee autonomy and allows researchers to publish openly whilst maintaining clearer boundaries around proprietary model development.
Neil Houlsby, formerly a Research Scientist at Google DeepMind, now leads Anthropic’s new Zurich office.
Nicholas Carlini, another DeepMind researcher, recently announced a one-year commitment to Anthropic.
Anthropic is reportedly preparing its first employee share buyback programme, allowing staff to sell shares at a US$61.5bn valuation matching the company’s recent Series E funding round.
How the AI industry is changing from compensation inflation
The talent warfare is creating systemic changes across the technology industry.
Entry-level hiring has collapsed as companies focus resources on securing experienced researchers – and among major technology companies, new graduates now account for just 7% of hires, down from 25% in 2023 and more than 50% pre-pandemic.
“The market’s hot. It’s not that hot,” Andrew says during a company meeting, attempting to counter reports that Meta offers US$100m packages to all AI hires.
“These engineers code toward humanity’s strangest bargain: our own redundancy,” Jatin concludes.
