After a year of talking up “personal intelligence,” Apple is swapping the person in charge of it. On Monday, Apple announced that AI chief John Giannandrea is stepping aside and that Amar Subramanya, fresh from Google and Microsoft, is coming in to turn an “invisible” AI plan into something investors can actually see.
This is CEO Tim Cook quietly rewriting Apple’s AI playbook. Cook had reportedly lost confidence in Giannandrea, who spent seven years steering machine learning and AI strategy; he’s moving into an advisory role ahead of a 2026 retirement. Subramanya arrives from a corporate VP role in AI at Microsoft and a 16-year run at Google, where he headed engineering for the Gemini assistant. He now has Apple Foundation Models, ML research, and AI safety on his desk and a delayed Siri relaunch in mid-2026 circled in red.
Apple’s AI push has been AWOL — just as rivals build their brands around agents and chatbots. The company has had a slow start in generative AI in particular, while Microsoft stuffs Copilot into everything, Google leans on Gemini, Meta trades on AI recommendation engines, and Samsung plasters “Galaxy AI” across its hardware. Meanwhile, Apple is still trying to convince people that its AI exists at all.
“AI has long been central to Apple’s strategy,” Cook said in Monday’s press release. But Apple’s keep-it-on-device, keep-it-quiet era has stalled out, and an AI culture reset now sits between Cupertino and a real growth story.
People haven’t exactly been delicate about why this switcheroo is happening. Investor’s Business Daily has been running with headlines such as “Apple Stock Has A Problem. It’s Apple AI.” Refrains that “Apple’s AI strategy is off to a poor start,” that these are “underwhelming models,” and that there are “troubling cracks” have echoed.
In a Tuesday analyst note, Wedbush’s Dan Ives said Apple’s “invisible AI strategy” sits on top of 2.4 billion iOS devices and 1.5 billion iPhones. That means that an “AI monetization piece could add $75 to $100 per share to the Apple story over the coming few years” and that “no ‘AI premium’ is factored into Apple’s stock at current prices” — even as Wedbush keeps an Outperform rating, a $320 price target, and Apple on both its Best Ideas and AI 30 lists.
Ives wrote that “the innovation coming out of Apple Park has been very disappointing so far” and called this week’s move “a major reset,” the first step in a push to “go on the offensive” after a year where Meta, OpenAI, and Anthropic have happily poached Apple AI talent. Ives said “outside hires [are] a necessary move to improve the AI strategy” and predicts more of them — even if Cook has to “pay up to bring in technology executives from the outside that will change the AI innovation culture.” Subramanya is Exhibit A.
Ives’ note is a sharply worded reminder that a $4 trillion company has been coasting through the AI moment. In an earlier note, he upgraded the language to “disaster,” saying Apple is “significantly lagging behind competitors like Microsoft, Google, Meta, and OpenAI” and that “no one on the Street believes” the company is delivering meaningful AI innovation internally. Apple keeps insisting that its approach is more thoughtful and private; the criticism is that it has also been slower and smaller.
This leadership shuffle lands after an awkward first year for Apple Intelligence, the “personal intelligence” layer Apple unveiled in 2024 to wrap generative features around iPhones, iPads, and Macs. The company shipped useful but modest tools — writing helpers, notification summaries, image clean-up — while the marquee act, a rebuilt Siri, slipped into 2026 amid repeated delays.
For a company that sells magic as much as hardware, a slow, mostly invisible AI rollout has become harder to wave away. Earlier this year, Apple’s head of Siri at the time, Robby Walker, reportedly told employees that the AI-driven overhaul had suffered “ugly” and embarrassing delays, with early versions giving wrong answers nearly a third of the time. Externally, Siri users have spent the year airing out their frustration on Reddit — while Cook tells interviewers Apple is “making good progress” on a more personalized assistant powered by Apple Intelligence. According to recent reports, the revamped assistant will run on a custom version of Google’s Gemini, with Apple paying roughly $1 billion a year for a 1.2 trillion–parameter model that “dwarfs” its current cloud-scale system.
Some of Giannandrea’s old organization is being redistributed to operations chief Sabih Khan and services boss Eddy Cue, and, earlier this year, Cook moved Vision Pro chief Mike Rockwell over to take charge of Siri after losing confidence in Giannandrea’s ability to get the assistant’s overhaul out the door. Subramanya will report to Federighi, whose software remit keeps expanding and who is now showing up in shortlists of potential Cook successors; his expanded oversight is, according to the press release, supposed to help deliver “intelligent, trusted, and profoundly personal experiences.”
Giannandrea’s era was defined by the idea that AI should feel like infrastructure. His team helped design Apple silicon, pushed on-device models, and built the plumbing that eventually carried the Apple Intelligence brand: writing tools, notification triage, photo tweaks, all tucked into iOS and macOS rather than staged as a chatbot. The Siri reboot was supposed to be the proof that this quiet work could add up to something impressive. Instead, that reboot became the focal point for investor grumbling, analyst notes, and — now — an executive shuffle.
Now, Subramanya has to persuade a famously cautious company that “personal intelligence” can be louder, faster, and more visible without blowing up Apple’s privacy religion.
By the time Siri gets its next big reveal, this AI handoff will either look prescient or desperate. A stronger assistant would suggest that Apple has finally learned how to turn its quiet AI work into a product worth bragging about. A shaky one — or an incremental one — could confirm what the skeptics have been saying all year: that the world’s most valuable hardware company still hasn’t figured out what it wants intelligence on that hardware to actually do.
Leave a Comment
Your email address will not be published. Required fields are marked *