Meta poaches top OpenAI researchers as Zuckerberg assembles 'Superintelligence' team
Jun 30, 2025
Key Points
- Meta recruited approximately 20 senior researchers from OpenAI, Google DeepMind, and Anthropic with reported offers up to $100 million, forcing OpenAI's Chief Research Officer Mark Chen to send a defensive internal memo promising to recalibrate compensation.
- Zuckerberg's $100 billion annual free cash flow and voting control of Meta shares let him spend tens of billions on talent without quarterly earnings pressure, a structural advantage OpenAI lacks as a capital-dependent company already losing $7 billion annually.
- Meta is branding the effort as a 'Superintelligence' lab rather than a ChatGPT competitor, signaling Zuckerberg knows direct consumer AI clones fail and is instead building models for internal ecosystem features and B2B applied AI.
Summary
Meta is raiding OpenAI's research team with aggressive offers large enough to force OpenAI into a public defensive posture. Mark Chen, OpenAI's chief research officer, sent a forceful internal memo over the weekend promising to fight the talent poach, recalibrate compensation, and retain staff without sitting idle. The memo, leaked to Wired, came days after Mark Zuckerberg announced a new "Superintelligence" lab led by Alex Wang as chief AI officer and Nat Friedman as co-lead. Meta has successfully recruited approximately 20 senior researchers from OpenAI, Google DeepMind, and Anthropic.
Signing bonuses and structural problems
Zuckerberg is offering $100 million signing bonuses to some OpenAI staffers, according to Sam Altman. Dylan Patel has questioned whether those are true day-one cash bonuses or total stock-based packages vesting over time. Either way, the offers are large enough to create friction within OpenAI. Researchers who stay loyal get nothing, while those who threaten to leave get paid premium rates. Peers suddenly see each other getting massive retention packages, creating perverse incentives.
OpenAI's cash position
OpenAI is already losing $7 billion annually, as Bill Gurley noted, and now faces pressure to recalibrate compensation across the board to match Meta's bids. Chen's memo promised to pursue "creative ways to recognize and reward top talent" and maintain "high personal standards of fairness." That constraint makes blanket raises harder to justify.
Meta's financial position
Zuckerberg has nearly $100 billion in annual free cash flow and controls enough Meta shares to block activist pressure for earnings-per-share growth. He can spend tens of billions on talent and strategic bets without quarterly earnings concerns, a position similar to early-stage Bezos. OpenAI, by contrast, raises capital from external investors who eventually want returns. Losing money at scale while your research team gets poached compounds the problem.
Historical precedent
Zuckerberg executed a similar playbook in 2010–2011, aggressively poaching mid-level and senior engineers from Google with tens of millions in stock packages to build Facebook's ads business. Google responded by raising compensation across the board. The current raid is larger and more public, but the mechanism is familiar. What differs is that Meta is now doing this not as an upstart fighting a sclerotic incumbent, but as a $1.3 trillion company with fortress economics attacking a younger, faster-moving rival.
Meta's strategy
Zuckerberg is marketing the effort as a "Superintelligence" lab rather than competing head-to-head with ChatGPT. This suggests he likely knows Google's direct-clone approach to Gemini failed to gain traction despite 100 million users. ChatGPT remains so dominant it has become a generic term. Meta's bet is to build models to power features across its own ecosystem and B2B applied AI, not to win consumer AI mindshare.