Mark Gurman, writing in Bloomberg:
Apple shakes up its AI efforts with a Google partnership, management changes and two new versions of Siri.
The framing of this story—"AI strategy reset," "management changes," "two new versions of Siri"—obscures what is actually happening. Apple, the company that built its modern identity on vertical integration and controlling the entire stack, is outsourcing its most important AI capability to Google.
This is not a partnership. It is a capitulation.
And the timing could not be more revealing: Apple is deepening its dependence on Google for AI at the precise moment that a federal judge has barred Google from paying to be the exclusive search engine on Apple devices—a deal worth $20 billion per year to Apple.
Apple is not pivoting. Apple is trapped.
The Whole Widget, Broken
To understand why this matters, you have to understand what Apple is supposed to be.
Steve Jobs famously said that Apple's advantage was that it made "the whole widget"—hardware, software, and services, all designed together. This wasn't just a business strategy; it was a philosophy. Apple believed that the best products came from controlling every layer of the stack, from silicon to user interface.
The transition to Apple Silicon was the ultimate expression of this philosophy. By designing its own chips, Apple could optimize every aspect of the Mac experience in ways that Intel never could. The M-series chips were not just faster; they were integrated. The company's quest to build key components in-house—displays, modems, batteries—reflected the same conviction.
AI was supposed to follow the same pattern. When Apple unveiled Apple Intelligence at WWDC 2024, the emphasis was on on-device processing, Private Cloud Compute, and user privacy. Apple was going to do AI the Apple way: locally, privately, integrated.
That vision is now dead.
A Decade of Siri Failure
The Gemini deal is not a sudden pivot; it is the culmination of a decade of Siri failure.
When Apple launched Siri in 2011, it was revolutionary—the first mainstream voice assistant. But Apple squandered its lead. While Google and Amazon poured resources into their assistants, Apple let Siri languish. By 2016, former Apple employees were describing how Siri's progress was crippled by lack of ambition. By 2017, Recode was writing that Apple wasted its lead with Siri.
The problem was structural. As the Wall Street Journal detailed last June, Apple's AI team failed to keep pace with competitors. Internal chaos—political infighting, unclear ownership, a culture that prioritized polish over speed—meant that Apple was always two years behind.
Even after Apple Intelligence launched, the problems continued. Former employees told the Financial Times that integrating LLMs with Siri led to bugs and an "incredibly frustrating" experience. Craig Federighi and other executives voiced concerns that the new Siri wasn't ready.
The result: eighteen months after Apple Intelligence was announced, Siri is still a joke. And Apple's solution is to hand the keys to Google.
The Google Trap
The strategic implications of the Gemini deal are profound—and almost entirely negative for Apple.
Apple has been here before. For over a decade, Google has paid Apple approximately $20 billion per year to be the default search engine in Safari. This deal represents 14-16% of Apple's Services revenue—nearly pure profit flowing directly to the bottom line.
But this dependency has always been precarious. In August 2024, a federal judge ruled that Google illegally monopolized the search market, in part through these default agreements. The remedy phase has been brutal: Google is now barred from paying to be the exclusive search engine on Apple devices. Apple and Mozilla face a massive revenue hit.
Apple's response to losing its Google search revenue? Sign another billion-dollar deal with Google, this time for AI.
The Gemini deal is reportedly worth ~$1 billion per year. That's a fraction of the search deal, but the strategic cost is far higher. With search, Apple could at least claim it was neutral—Google paid to be the default, but users could switch. With Siri, there is no switching. If Siri is powered by Gemini, Apple's core assistant runs on Google's intelligence.
The Arc of the Deal
The Gemini partnership didn't happen overnight. The TEXXR corpus traces the full arc:
- March 2024: Apple and Google enter active talks about using Gemini to power some iPhone features.
- May 2024: Apple finalizes a deal with OpenAI for ChatGPT integration—a hedge.
- June 2024: Apple Intelligence announced at WWDC, emphasizing on-device processing.
- August 2025: Talks with Google intensify for a "revamped Siri."
- September 2025: Apple plans AI web search for spring 2026—an attempt at independence.
- November 2025: Apple finalizes the Gemini deal for ~$1B per year.
- January 2026: The deal is signed. Details emerge: Gemini will power new Siri features.
- January 25, 2026: Gurman reports the full scope: two new Siris, management changes, a "strategy reset."
The trajectory is clear: Apple tried to do AI itself, failed, and ultimately surrendered to Google.
The Succession Question
The Gemini deal also intersects with another slow-moving story: Tim Cook's succession.
Cook, now 65, has been CEO for nearly fifteen years. Potential successors have been identified—hardware chief John Ternus, services head Eddy Cue, software chief Craig Federighi. Jeff Williams, the presumed heir apparent, retired last July.
The AI failure is, in part, a Tim Cook failure. The Cook era has been defined by operational excellence, not product vision. Apple under Cook has been brilliant at extracting value from existing product lines—the iPhone, Services, wearables—but has struggled to create new categories. The Vision Pro stumbled. The car project was cancelled. And now AI.
The Gemini deal is Cook's implicit admission that Apple cannot solve AI on its own. It raises the question: can Apple's next CEO do better? Or is the "whole widget" philosophy simply incompatible with the AI era?
The Counter-Argument
There is a charitable interpretation of the Gemini deal.
Apple has always been willing to use best-in-class components when it made sense. Apple doesn't make its own NAND flash; it buys from Samsung and SK Hynix. Apple doesn't make its own OLED displays; it buys from Samsung and LG. Integration matters at the points of differentiation; commodity components can be outsourced.
Perhaps Apple views LLMs as a commodity. If Gemini, Claude, and GPT-4o are all roughly equivalent, then it doesn't matter which one powers Siri—what matters is the integration layer, the user experience, the privacy framework.
This is the argument Ben Thompson made about AI modularization: value accrues to whoever owns the integration points, not necessarily whoever owns the models. Apple could, in theory, use Google's models today and switch to its own models tomorrow, with users none the wiser.
But this argument has a flaw: AI is not a commodity. The model is the product. When you ask Siri a question and Gemini answers, the quality of that answer—its accuracy, its tone, its helpfulness—is determined by Google, not Apple. Apple can wrap Gemini in a pretty interface, but it cannot control what Gemini says.
For a company that built its brand on control, this is an existential compromise.
What Comes Next
According to Gurman, Apple plans to unveil the Gemini-powered Siri next month in iOS 26.4, with a "reimagined chatbot-like Siri" coming at WWDC in June.
The market will likely cheer. Finally, Siri that works! Apple stock may rise. The narrative will be one of pragmatism—Apple doing what it takes to compete.
But the structural damage is done. Apple is no longer the company that makes "the whole widget." It is a company that makes beautiful hardware, runs someone else's AI, and collects a toll on the ecosystem.
That may be a fine business. It is not the Apple that Steve Jobs built.
This analysis draws from the TEXXR coverage of the Apple AI reset story, including 26 related articles. Explore the full Apple AI coverage, trace the evolution of Siri as a topic, or see Apple-Google-Gemini relationships in Nexus.