Meta Recruits Apple’s Top AI Engineer in Talent War

Meta’s talent raid on the frontier-AI world just claimed its highest-profile Apple scalp. Bloomberg broke the story that Ruoming Pang, the distinguished engineer who ran Apple’s foundation-models team, is leaving Cupertino to join Meta’s newly formed Superintelligence Labs—the same unit that has already siphoned senior researchers from OpenAI, Anthropic and Google DeepMind.

Pang’s exit follows Meta’s June hiring blitz, when CEO Mark Zuckerberg personally lured Scale AI founder Alexandr Wang (now chief AI officer), former GitHub boss Nat Friedman and at least eleven named scientists from OpenAI and Anthropic into the Superintelligence Labs org.

Recruiters close to the talks say Meta’s standard package for tier-one researchers now starts at “low-eight-figure” RSU grants plus compute stipends that guarantee immediate access to the company’s 600,000-GPU fleet.

Apple’s own generative-AI push has lagged rivals; the company only recently weighed licensing external LLMs for Siri. Pang’s departure strips Apple of the executive who was coordinating that effort, just as Meta markets its open-weight Llama models as “community friendly”—a cultural contrast to Apple’s famously closed stack. Investors read the move as more evidence that top AI builders prefer the scale, open-research posture and fast-track publishing policies Meta now offers.

Since April, Zuckerberg has said the goal is “personal super-intelligence” for billions of users inside Instagram, WhatsApp and its forthcoming AR glasses. To get there, Meta is hoovering up anyone with frontier-model chops—even if it means paying start-up-founder money to big-tech employees. According to recruiters, Meta’s private Slack channels are brimming with Cloud TPU and OpenAI alumni discussing multi-modal agents that can plan across text, vision and robotics.

The fallout

• OpenAI lost alignment researcher Anton Bakhtin and compiler specialist Yanzhi Li in May.
• Anthropic saw Claude-evaluation scientist Sophie Krigslund depart in June.
• Apple now loses its bridge between Siri’s legacy NLP stack and future foundation models.

Microsoft and Google are responding with counter-offers reportedly topping US $5 million per year for staff-plus AI roles, but head-hunters say Meta is still out-bidding on total equity. Meanwhile, start-ups fear a brain drain that could leave them priced out of their own Series A budgets.

Meta’s Superintelligence Labs is expected to preview a Llama-5 research checkpoint at Meta Connect in September. If the company continues to peel away senior talent from rivals at this pace, the announcement could underscore a new reality: the fiercest competition in AI this year isn’t over products—it’s over people.