Why is Apple building its next AI models on Google's Gemini?

Yesterday, Monday, January 12, 2026, Apple and Google confirmed a partnership that had been rumored for months. In a short joint statement, the two companies said they have entered a multi-year collaboration: the next generation of Apple Foundation Models, the AI models behind Apple Intelligence, will be built on top of Google's Gemini models, using Google's cloud technology as well.

Apple's reason, in its own words, is that after careful evaluation it found Google's AI technology to be the most capable foundation for its models. The first visible result should be the more personal version of Siri that Apple has promised, now expected this year.

What was announced

The statement itself is brief. Here is what is confirmed, what has been reported, and what remains open:

Apple declined to comment on the terms, and Google referred reporters back to the joint statement.

The background

Apple introduced Apple Intelligence in 2024, a set of AI features built into its operating systems, such as searching your photos by description and summarizing notifications. Much of that work runs directly on the device, and harder requests go to Private Cloud Compute, Apple's own tightly controlled servers, so that personal data stays within Apple's systems.

Those features run on Apple's own models, the Apple Foundation Models. TechCrunch describes the result as quiet, sometimes nearly invisible AI that lacks the wow factor of ChatGPT or Gemini. The biggest disappointment was Siri: Apple advertised a smarter, more personal assistant and then delayed it last year, saying the features would take longer than planned and arrive in the coming year.

Apple already works with OpenAI: Siri and Apple Intelligence can pass complicated questions to ChatGPT. TechCrunch reports that Apple also tested technology from OpenAI and Anthropic before settling on Google.

How "based on Gemini" could work

A foundation model is a large, general AI model trained on huge amounts of text and other data, which can then be adapted to many tasks. Building one from scratch takes years of research and enormous computing power. Saying Apple's next models will be "based on" Gemini means Google's models and know-how become the starting point, rather than Apple training everything alone.

Neither company has explained the technical arrangement. Apple's statement that Apple Intelligence will keep running on its devices and Private Cloud Compute suggests that the resulting models will run inside Apple's own systems, not as a simple pass-through to Google's Gemini app.

A rough analogy: a carmaker that buys an engine design from a rival but builds the car, the dashboard and the safety systems itself. The driver sees the carmaker's brand and controls; what is under the hood comes partly from someone else.

Why it matters

For Apple, the deal is a pragmatic admission that it is behind on the largest AI models and would rather license the best available technology than wait. Apple has long preferred to build its own hardware and software, so relying on a rival for a core technology is a notable shift.

For Google, it is a major vote of confidence. CNBC notes that Google's shares rose on the news and that Alphabet briefly passed $4 trillion in market value. Gemini will now help shape the assistant on a vast number of iPhones, iPads and Macs, on top of Gemini's own apps.

The deal also links two companies already tied together. Google has long paid Apple to be the default search engine on its devices, a practice a US judge ruled illegal in 2024. In December 2025 the court's final remedies barred Google from exclusive default deals lasting more than a year.

For users, the hopeful reading is a more capable Siri that can understand context and act across apps, with Apple's privacy rules still in place. The open questions are when, how well it works, and how clearly Apple explains where each request is processed.

A worked example: requests to try when the new Siri arrives

Apple has not listed specific features. Requests that connect personal information across several apps are a fair test of a "more personalized" assistant:

"When is my sister's flight landing, and how long will it take me to get
to the airport from here?"

"Add the address in this message to her contact card."

"Find the photo of the whiteboard from last Tuesday's meeting and send
it to Marc."

Each one needs Siri to connect information from different apps, exactly where the current assistant struggles.

What will change on your iPhone, and what stays the same?

For now, nothing changes. The statement describes future models and future features; no update was released with the announcement. Apple says the more personalized Siri is coming this year, and earlier reports point to spring.

What Apple says stays the same is the privacy approach. Apple Intelligence will continue to run on your device and on Private Cloud Compute, under Apple's standards. Apple also told CNBC it is making no changes to its ChatGPT agreement, so that optional integration should remain.

When the new features arrive, read Apple's description of where requests are handled, check the privacy settings for Apple Intelligence and ChatGPT, and test the assistant on your own everyday tasks before relying on it for anything important.