Apple’s AI ambitions have hit a stumbling block as one of its star executives bolts for Menlo Park. In the latest move of a talent war that’s showing no signs of cooling, Ruoming Pang—who ran Apple’s in‑house “foundation models” group—has decided to join Meta, lured by a package reportedly worth millions of dollars per year.

Meta’s CEO, Mark Zuckerberg, has turned recruitment into a core part of his strategy for catching up in AI. Over the past few months, Meta has dangled eye‑watering offers to engineers from OpenAI, Anthropic, and Scale AI—some rumored to include signing bonuses in the tens of millions. The recent formation of “Meta Superintelligence Labs,” headed by former Scale AI CEO Alexandr Wang, underscores how seriously the company is pursuing artificial general intelligence (AGI), even as skeptics question whether current methods will ever deliver true AGI.

Related /

• OpenAI and Google are walking away from Scale AI
• OpenAI CEO: Meta’s recruiting strategy won’t beat innovation
• Zuckerberg’s AI plan B: hire the CEO after failing to buy the startup
• Meta hires Scale AI CEO and buys 49% stake in $14.3 billion deal

Pang’s departure marks a particularly stinging defeat for Apple. He joined the company in 2021 from Alphabet, and since then has overseen a team of roughly 100 engineers building the core models that power Apple Intelligence features like Siri’s contextual understanding and on‑device summarization of emails and messages. These models were also expected to underpin newer additions—personalized notifications, content summaries, and animated “Genmoji” stickers—that Apple previewed at WWDC 2024 but has struggled to ship reliably.

Insiders say discussions at Apple around possibly licensing third‑party technology—specifically models from OpenAI or Anthropic—for the next big Siri upgrade unsettled Pang’s group. Rumors of a hybrid Siri, partly powered by external providers, reportedly left long‑time team members feeling sidelined and uncertain about the future of their own work. As news of these behind‑the‑scenes talks broke, whispers began circulating that other engineers were quietly fielding competitive offers, and that Apple risked losing more talent if it didn’t shore up its internal roadmap.

The timing could hardly be worse. Apple’s AI push has already faced public setbacks: promised “Apple Intelligence” Siri enhancements slated for late 2024 have been delayed until 2026, and a federal lawsuit accuses the company of false advertising over features that never arrived. Having invested heavily in promoting next‑generation Siri, Apple now finds itself needing to reassure both consumers and employees that its home‑grown AI remains central to its strategy.

By contrast, Meta has doubled down. In June, the company invested $14.3 billion in Scale AI at a $29 billion valuation, positioning itself to leverage Scale’s data infrastructure and talent base. Within days, Meta was scooping up three researchers from OpenAI—reportedly at Zuckerberg’s direct request—and adding them to the nascent superintelligence team. Just this week, Pang became the latest high‑profile recruit.

Meta’s gamble is that assembling an elite squad of AI experts will accelerate its work on large‑scale models and next‑generation applications—ranging from its standalone AI chatbot to AR glasses that overlay real‑time intelligence onto the physical world. For shareholders and engineers alike, the message is clear: come work at Meta, and you might help build the future of intelligence.

Apple’s software chief Craig Federighi and Mike Rockwell—who led the Vision Pro headset project—are now jointly overseeing the company’s AI efforts. Their immediate challenge is twofold: rebuild confidence within the foundation models group and set a clear, competitive direction for Siri and related services. Analysts predict a phased approach, with partial feature rollouts via iOS 18.x and a fuller revamp in 2026 when Apple can bundle external and internal models seamlessly.

In the meantime, Apple must stem the tide of defections. Its legendary culture of polished product releases and tight integration has become a double‑edged sword: engineers want to work on cutting‑edge breakthroughs, but the company’s cautious, quality‑first ethos can slow experimentation. As Meta’s bidding war intensifies, Apple faces a critical decision: open its doors wider to outside models, or double down on the teams still committed to a wholly “Apple‑made” AI.

In the war for AI talent, reputation and mission matter as much as money. And today, that mission looks a lot clearer at Meta than it does in Cupertino.