Apple and Google announced a multi-year partnership on January 12, 2026 that places Google's Gemini models at the core of Siri and Apple Intelligence, a deal worth roughly **$1 billion** per year that reshapes how AI reaches over two billion active Apple devices. After years of falling behind in the generative AI race, and a string of embarrassing Siri delays, Apple has made a decisive bet: rather than building frontier-scale models in-house, it will license the technology from the company that already dominates AI on the other side of the smartphone market.
Apple chose Google over OpenAI and Anthropic
In a joint statement, the two companies were direct about the scope of the collaboration:
"Apple and Google have entered into a multi-year collaboration under which the next generation of Apple Foundation Models will be based on Google's Gemini models and cloud technology. These models will help power future Apple Intelligence features, including a more personalized Siri coming this year."
Apple told CNBC separately: "After careful evaluation, we determined that Google's technology provides the most capable foundation for Apple Foundation Models."
That "careful evaluation" included pitches from OpenAI and Anthropic. According to Bloomberg's Mark Gurman, who first reported the deal's parameters in November 2025, what tipped the scales toward Google was not just model performance but also price. Apple evaluated all three providers on capability, infrastructure fit, and cost, and Google won on the combination.
The partnership is non-exclusive, meaning Apple retains the option to bring in additional AI providers. And its existing arrangement with OpenAI, which integrated ChatGPT into Siri in late 2024 as an optional fallback for "world knowledge" queries, remains intact. Apple confirmed to CNBC it "isn't making any changes" to the OpenAI agreement. But the structural message is clear: Gemini is the new foundation. ChatGPT becomes the side door.
A 1.2 trillion parameter model replaces Apple's 150 billion parameter system
The technical leap is dramatic. According to Gurman's November 2025 Bloomberg report, Google is building a custom Gemini model for Apple with approximately 1.2 trillion parameters, roughly eight times the scale of the 150 billion parameter models currently powering Apple Intelligence in the cloud.
This model will handle Siri's summarizer and planner functions, the reasoning components that synthesize information across apps and decide how to execute multi-step tasks. Apple demonstrated examples during the announcement: asking Siri about "my mother's flight and lunch reservation plans" and having it pull information from Mail and Messages automatically, or asking "Can I make my 5 PM flight if I leave now?" and getting a reasoned answer that checks traffic, flight status, and TSA wait times.
The integration is white-labeled. Users will not see Google branding anywhere. From the outside, this is still Siri. Underneath, it runs on Gemini.
Some capabilities will continue using Apple's own models. The arrangement creates a tiered architecture: simple tasks stay on-device, mid-range requests route to Apple's Private Cloud Compute servers, and the most demanding reasoning tasks use the Gemini-powered models.
Apple's privacy architecture stays intact, but the infrastructure story is shifting
Privacy was always going to be the central tension in this deal. Apple has built a brand on the promise that user data stays on Apple infrastructure, and the company is emphatic that this principle holds.
CEO Tim Cook told CNBC on January 29: "We're not changing our privacy rules. We still have the same architecture that we announced before, which is on device plus Private Cloud Compute."
Under the original plan, Google's custom model would be trained using Google Cloud infrastructure and TPU chips, but inference, the actual processing of user requests, would run on Apple's own Private Cloud Compute servers. No user data would be shared with Google.
However, reporting from March 2026 introduced a wrinkle: Apple has asked Google to explore hosting dedicated Gemini-powered Siri servers on Google's own cloud infrastructure, potentially with custom security configurations to meet Apple's privacy standards. If this materializes, it would mark a significant departure from Apple's self-reliant infrastructure approach and could draw regulatory scrutiny, particularly under the EU Digital Services Act and ongoing US antitrust proceedings involving both companies.
Google confirmed publicly that it "will not receive Apple user data" through the deal. But partnering with a company whose core business model is data collection and targeted advertising creates a trust tension that Apple will need to continually manage.
Two years of Siri failures forced Apple's hand
This partnership did not emerge from a position of strength. It emerged from a crisis.
Apple first teased the "smarter Siri" at WWDC in June 2024, promising conversational awareness, on-screen understanding, and deep app integration. By March 2025, the company was forced to publicly acknowledge that the features would take longer than expected. Software chief Craig Federighi and other executives found in internal testing that "features didn't work properly, or as advertised."
Behind the delays was a deeper structural problem: Apple was hemorrhaging AI talent. Over the course of 2025, the company lost over a dozen senior AI researchers to competitors. Ruoming Pang, who led the Foundation Models team, departed for Meta after being offered a reported **$200 million** compensation package by Mark Zuckerberg. Other key researchers left for OpenAI and Anthropic. The Foundation Models team reportedly comprised only 50 to 60 people, making each departure acutely damaging.
Against this backdrop, Apple's decision to license Gemini was less a strategic pivot and more a pragmatic rescue operation. The company needed a frontier-capable model, and it needed one fast.
The Gemini-powered Siri is already running into trouble
The deal was announced in January. By February, the rollout was already slipping.
Apple had planned to ship the first Gemini-powered Siri features in iOS 26.4, targeted for March 2026. But on February 11, Bloomberg reported that internal testing was going poorly. Engineers found that the new Siri sometimes failed to process queries correctly, took too long to respond, and in some cases defaulted to using ChatGPT rather than Apple's own technology.
MacRumors reported that Apple is now expected to spread the rollout across multiple releases: some features may arrive in iOS 26.5 (targeted for May), while others could slip to iOS 27 in September. Apple pushed back on the characterization of a "delay," telling outlets the revamped Siri is "still on track" for 2026, but the situation remains fluid.
This is the third time Apple has failed to deliver on its Siri overhaul timeline: first at WWDC 2024, then in the March 2025 postponement, and now the iOS 26.4 slip. The pattern raises real questions about whether the challenges are model-related or systemic to how Apple approaches AI integration.
Google wins control of both mobile AI ecosystems
For Google, the deal is arguably more consequential than it is for Apple. Google's Gemini models now power AI features on both major mobile operating systems: directly on Android, and through this partnership on iOS. Combined, that covers virtually every smartphone user on the planet.
Wall Street responded accordingly. On the day of the announcement, Alphabet's market capitalization crossed **$4 trillion**, making it only the fourth company to reach that milestone after Nvidia, Microsoft, and Apple. Alphabet shares rose 1% on the news, with the Apple deal serving as the final catalyst in a broader AI-driven rally.
On Alphabet's Q4 2025 earnings call in February, the company reported revenue of *$113.83 billion** (beating estimates of *$111.43 billion**) and noted that the Gemini app had reached 750 million monthly active users, up from 650 million the prior quarter. Alphabet also disclosed that Gemini serving costs had been reduced by 78% over the course of 2025 through model optimizations.
Dan Ives of Wedbush Securities called the deal "a major validation moment for Google" and "a stepping stone" for Apple to get its AI strategy on track. He reaffirmed a Buy rating on Apple with a $350 price target.
Not everyone was impressed. Elon Musk, whose xAI company builds the competing Grok model, posted on X calling the partnership "an unreasonable concentration of power" given Google's existing dominance through Android and Chrome. Musk's xAI is separately suing Apple and OpenAI over their ChatGPT integration, alleging it preempts competitors like Grok from fair access to the App Store.
Apple still wants to replace Gemini with its own models
For all the scale of this deal, Apple has been clear, at least internally, that Gemini is a bridge, not a destination.
According to Bloomberg, Apple's models team is actively developing a 1 trillion parameter cloud-based model intended to eventually replace Google's technology. The target for consumer readiness is as early as 2027, though some reports suggest late 2026 is possible.
The challenge is execution. With a diminished AI research team and a track record of missed timelines, Apple's ability to build a competitive frontier model on that schedule is far from certain. The company has announced aggressive hiring plans to rebuild, but attracting top AI talent to a company that just publicly acknowledged it cannot build its own models is a hard sell.
Gurman noted that Apple "still doesn't want to use Gemini as a long-term solution." That reluctance makes strategic sense: depending on a rival for your core product intelligence is an inherently vulnerable position. But wanting to replace Gemini and actually doing so are very different things.
Developers will benefit from a smarter Siri, eventually
For iOS developers, the Gemini integration should expand what Siri can do with third-party apps. Apple's SiriKit and App Intents frameworks will allow developers to expose their app functionality to a substantially more capable voice assistant, enabling richer hands-free workflows and more complex multi-app interactions.
The practical impact depends on the rollout timeline. If the Gemini-powered Siri ships incrementally across iOS 26.5 and iOS 27, developers may need to wait until WWDC in June 2026 for concrete APIs and integration guidance. Apple has not yet published updated developer documentation specific to the Gemini partnership.
The $1 billion annual fee masks a deeper strategic cost
Gurman's reporting pegged the deal at approximately **$1 billion** per year. Bloomberg noted this "seems incredibly cheap, considering all the other big tech giants are spending hundreds of billions of dollars a year on their AI efforts."
For context, Alphabet itself plans capital expenditures of *$175 billion** to *$185 billion** in 2026, much of it on AI infrastructure. Apple's **$1 billion** annual licensing fee looks modest next to building and maintaining frontier models from scratch.
But the real cost is strategic. Apple is paying Google for the engine that makes its flagship product intelligent. That dependency gives Google insight into Apple's product roadmap, creates antitrust entanglements, and makes the iPhone's core differentiator (Siri) contingent on a rival's technology. The financial price may be low. The strategic price is yet to be determined.
The Bottom Line
The Apple-Google Gemini deal is the most significant partnership in consumer AI since Microsoft's investment in OpenAI. It acknowledges a reality that has been clear to outside observers for over a year: training frontier-scale models is now so capital-intensive and talent-dependent that even the world's most valuable company prefers to buy rather than build.
For users, the promise is a Siri that finally works the way Apple described it at WWDC 2024, though the repeated delays suggest patience is still required. For the industry, the deal signals that the AI market is consolidating around a small number of model providers, with Google now positioned as the foundational layer for both mobile ecosystems.
The deeper question is whether Apple can reclaim its AI independence. The company says it is building a 1 trillion parameter model to replace Gemini. If it succeeds, the Google deal becomes a well-executed stopgap. If it does not, Apple becomes permanently dependent on Google for the intelligence layer of every product it ships.
To understand the technology underpinning this shift, read our explainer on How Large Language Models Actually Work. For more on Google's latest model capabilities, see our coverage of Google Just Dropped Gemini 3.1 Pro.