Apple just lost another four AI researchers and a senior Siri executive to Meta and Google DeepMind. This isn't a coincidence. It's a pattern that should worry anyone tracking the tech giant's future.
The departures include Yinfei Yang, who's launching his own startup, plus Haoxuan You and Bailin Wang heading to Meta's Superintelligence division. Zirui Wang jumped ship to Google DeepMind. But here's the kicker: Stuart Bowers, a senior Siri executive who reported directly to chief Mike Rockwell, also left for Google DeepMind to work on Gemini models. That's the exact same technology Apple now licenses to rebuild Siri.
You can't make this stuff up.
Last July, Apple hemorrhaged Ruoming Pang, who led its foundation models team, to Meta. The price tag? North of two hundred million dollars. Then in December, John Giannandrea stepped down after seven years as SVP of Machine Learning and AI Strategy. Throughout 2025, dozens more fled to Meta, OpenAI, and various startups. Analysts call it a knowledge vacuum forming at precisely the wrong time.
The Trade-Off Nobody's Talking About
Apple made a choice. Rather than build frontier AI models internally, they'd partner with external providers and ship competitive experiences faster. In January 2026, they formalized a multi-year deal with Google worth roughly one billion dollars annually. Google's Gemini now powers Apple Intelligence and Siri, with a major overhaul planned for spring 2026—about eighteen months behind schedule.
There's logic here. Outsource expensive, uncertain R&D to partners while focusing on what Apple does best: integration, privacy architecture, and hardware differentiation through Apple Silicon. Software chief Craig Federighi, now working under new intelligence president Amar Subramanya, has always been cautious about generative AI. He sees it as unpredictable and hard to control. Apple even rejected using AI to reorganize iPhone home screens because they feared confusing users.
Three Risks Investors Are Overlooking
The market hasn't priced in the real dangers lurking in this approach.
First, Apple's AI moat shifted from owning the models to controlling distribution, user experience, and silicon. That's still valuable, but only if Apple captures meaningful economics from AI services. With Gemini running the show, who gets the take rate? Someone's paying per-query costs. Either Apple's services margin compresses, or they'll need usage caps that hurt competitiveness. Investors assuming AI will magically boost services revenue need reality checks.
Second, losing talent kills execution speed and integration quality. Frontier AI work depends heavily on tacit knowledge. The engineers who understood why Siri failed, built evaluation frameworks, and designed system integration layers just walked out the door. Even with a superior base model, Apple could fumble the execution. The gap between credible AI assistants and broken ones lives in debugging, context management, and reliability engineering—expertise that left with those employees.
Third, consider the irony. Stuart Bowers now works on Gemini at DeepMind while Apple builds Siri using Gemini. That's not just awkward. It means knowledge flows against Apple in future negotiations around pricing, priority access, and customization. Apple might eventually need an acquisition, internal rebuild, or multi-model strategy to regain leverage.
The Make-or-Break Moment
The Siri upgrade arriving this spring represents everything. If it's incremental rather than transformative, Apple has fallen decisively behind while users cement habits with ChatGPT, Gemini, and Copilot. Success demands Siri become an OS-native agent handling multi-step actions across apps without hallucinating constantly.
For investors, the base case looks like this: Apple ships improved Siri without a clean monetization story, making the dependency discount more visible. Google gains cloud revenue and validation as a premium consumer platform supplier. Meta's aggressive talent grab—offering signing bonuses reportedly hitting one hundred million dollars to OpenAI employees—might work if it translates into model leadership. Then again, those compensation levels could signal peak talent-cost inflation.
At current prices—258 dollars for Apple, 338 dollars for Alphabet, and 716 dollars for Meta—the market hasn't absorbed this strategic realignment. Not even close.
NOT INVESTMENT ADVICE