Apple’s AI brain drain isn’t a one-off headline anymore, it’s starting to look like a pattern—and a pretty awkward one, given how central AI has suddenly become to the company’s future pitch. In the span of just a few weeks, Apple has lost at least four more AI researchers and a senior Siri executive to rivals Meta and Google DeepMind, right as it leans on Google’s Gemini models to power the next generation of Siri.

The latest departures read like a who’s who from Apple’s foundation model efforts. Bloomberg reports that researchers Yinfei Yang, Haoxuan You, Bailin Wang, and Zirui Wang have all exited, with You and Bailin Wang heading to Meta and Wang joining Google DeepMind. Yang, meanwhile, has left to start a new company, suggesting Apple isn’t just competing with Big Tech for this talent but also an increasingly attractive startup ecosystem built around generative AI. These are not generalist engineers; they’re people working on the underlying AI models that sit at the heart of Apple’s “Apple Intelligence” strategy and Siri’s planned reboot.

The optics get even messier when you zoom in on where they’re going and what those teams are doing. Meta has been aggressively staffing up its Superintelligence research arm and recommendation systems—both areas that directly intersect with large-scale foundation models and personalization, exactly the kinds of capabilities Apple is now saying it wants to deliver on-device. Google DeepMind, for its part, is not just a competitor; it’s the partner Apple now relies on to supply and help train the Gemini models that will sit behind Siri’s upcoming “AI upgrade.” When one of your researchers and one of your senior Siri executives leave to join the very lab that’s selling you the core tech, it raises questions about who really owns the AI crown jewels in the long run.

Then there’s Stuart Bowers, the senior Siri executive whose exit might be the most uncomfortable of them all. Bowers, who previously helped lead Apple’s now-shelved self‑driving car project, had his responsibilities around Siri expanded last year and reported directly to Siri chief Mike Rockwell. Now he’s off to Google DeepMind to work on Gemini—the same family of models Apple is paying to use for its own Siri overhaul. The move underlines a strange dynamic: Apple is trying to pitch a future where Siri becomes smarter, more conversational, and more useful, but some of the people who were supposed to make that happen internally are now helping the supplier instead.

All of this is landing on top of a major reshuffle in Apple’s AI leadership. In late 2025, Apple announced that John Giannandrea, its high‑profile senior vice president for Machine Learning and AI Strategy, would step down from his role and move into an advisory position before retiring in spring 2026. His organization is being split, with new AI leadership—like Amar Subramanya, now vice president of AI under Craig Federighi—taking over responsibilities for Apple Foundation Models, ML research, and AI safety and evaluation. On paper, Apple frames this as a maturation of its AI efforts, but when you pair that with a steady drip of departures from the very teams building those models, it feels less like a clean handoff and more like a moving target.

Meanwhile, the strategic bet Apple is making to catch up in AI is unusually external for a company that has spent two decades insisting that the most important technologies have to be built in-house. Under a multi‑year deal announced in January 2026, Apple will base the “next generation of Apple Foundation Models” on Google’s Gemini models and cloud infrastructure, using them to power a much more personalized Siri and other “Apple Intelligence” features rolling out later this year. The companies haven’t publicly detailed which Gemini variants will ship or how this stack will coexist with Apple’s own models and any ChatGPT integration, but the direction of travel is clear: Gemini is becoming the mental model behind Siri’s next era.

From the outside, that creates a kind of identity crisis for Siri. For years, Apple’s pitch was that Siri might be less flashy than competitors but more private, more integrated, more “Apple.” Now, the company is promising a conversational Siri chatbot experience in upcoming iOS releases, driven in part by models that live on Google’s cloud and have been trained and tuned by DeepMind researchers—including, now, some former Apple staffers. If it works, users may not care who wrote which line of code, but inside the industry, it blurs the line between Apple’s own AI roadmap and Google’s, especially as regulators watch distribution and default‑placement deals for AI with the same suspicion they applied to web search.

The talent story is what ties all of this together. AI researchers today have more leverage than almost any other group in tech, and they’re voting with their feet. Meta, which has been aggressively open‑sourcing Llama and talking up its “open” AI posture, offers the chance to publish, influence the broader ecosystem, and work on large‑scale research problems with immediate consumer impact in Instagram, Facebook, and WhatsApp recommendations. Google DeepMind offers pure research prestige and direct control over frontier foundation models like Gemini, deployed across search, Android, and now, via Apple, the iPhone itself. Compared with that, Apple’s more secretive culture and its late, somewhat reactive pivot toward big, branded “AI moments” may feel less attractive to researchers looking to leave a visible mark on the field.

None of this means Apple is out of the AI race. The company still has massive advantages: billions of active devices, custom silicon designed for on‑device machine learning, and a user base that tends to adopt new system‑level features at scale once they ship. It has also been quietly baking machine learning into camera pipelines, accessibility features, search, and on‑device intelligence for years, long before “AI” became the catch‑all marketing term it is now. But the current wave of departures, especially to Meta and Google DeepMind, exposes the tension between that incremental, behind‑the‑scenes AI and the new era of giant, branded foundation models that demand both talent density and a loud story about where you’re headed.

For users, the near‑term impact may simply look like this: a much smarter Siri shows up on your iPhone later this year, powered in part by Google’s Gemini and tuned by teams that no longer all sit inside Apple Park. For Apple, though, the bigger question is whether it can stabilize its AI organization, convince world‑class researchers that Cupertino is still where the most interesting problems get solved, and eventually bring more of this capability back under its own roof. If it can’t, the company that once prided itself on owning the full stack—from chips to software to services—may find that in the AI era, the most important layer of that stack is increasingly being built somewhere else.