"Shelling" Google Gemini, but Apple hasn't given up on developing their own model yet

Writing by: Xiao Jinya, Alphabet AI

Apple “bowed its head.”

On January 12th, local time, Apple and Google issued a joint statement:

“After careful evaluation, Apple believes that Google’s artificial intelligence technology provides the most powerful foundation for Apple’s foundational models and is excited about the innovative experiences it will unlock for Apple users.”

Both companies stated that these models will help support future Apple Intelligence features, including the more personalized Siri set to launch later this year.

Once the news broke, Elon Musk was the first to express displeasure, commenting on X that this kind of “concentration of power” is unreasonable, especially considering Google still owns Android and Chrome browsers.

This is essentially saying that Google now controls even Apple AI. But Apple would definitely not agree with that.

For Apple, this is a temporary bowing, but not surrender or giving up.

The introduction of Google Gemini is more like a helper Apple is seeking during the transition period; it will not completely replace Apple’s self-developed models. Moreover, Apple is still working on models with trillions of parameters.

Gemini Does Not Replace Apple’s Self-Developed Models

Apple and Google’s public statements are limited and deliberately vague, but one thing is clear: Gemini will participate in building the new version of Apple’s foundational models.

This was reported by Bloomberg’s Mark Gurman as early as November last year. He cited sources saying that Apple pays Google about $1 billion annually.

However, this is not a “one-to-one” deal where Gemini directly replaces Apple’s foundational models. Instead, it is a system architecture that allows Apple to continue using its own models without sharing any data with Google.

As the initial rumors indicated, Gemini will not be directly embedded into Apple’s operating system.

Instead, everything will still be centered around Apple’s foundational models, but at the underlying architecture level, Gemini will serve as its “foundation.” In other words, Apple’s use of Gemini is in a “shell” and “distillation” middle ground. Of course, under the premise of cooperation between both parties, this is certainly legal.

This point is crucial; interpreting this news as “Apple’s AI will be driven by Gemini” is inaccurate.

When you buy an iPhone, it internally uses chips manufactured by TSMC, displays from Samsung, and various components from around the world. But no one would claim that Samsung is “driving” the iPhone.

The Apple Intelligence on the device side will still be supported by Apple’s foundational models.

In other words, the Gemini model will serve as a tool to help train Apple’s foundational models, enabling them to better perform the various tasks set by Apple.

Gurman’s report last year mentioned that, according to agreements between Google and Apple, Google’s Gemini model will be responsible for Siri’s summarization and planning functions—components that help the voice assistant synthesize information and decide how to execute complex tasks. Some functions of Siri will continue to use Apple’s self-developed models.

The report also stated that this model will run on Apple’s private cloud servers, ensuring user data is completely isolated from Google’s infrastructure. Apple has allocated AI server hardware to support the operation of this model.

Since both parties’ public statements are currently vague, the above content has not been refuted.

This deal is fundamentally different from collaborations like “Google Search,” as it is completely invisible to end users. When interacting with Siri, there will be no indication of Gemini, nor should there be.

Apple Has Not Given Up

Further cooperation with Google is a “transition” choice for Apple.

Apple users have been looking forward to the new Siri since 2024, when Apple showcased it at WWDC and said it would be gradually rolled out. But since then, it has been repeatedly delayed, becoming a major embarrassment for Apple.

Currently, multiple media reports suggest that the new Siri will be released in March along with iOS 26.4.

In addition, Apple Intelligence itself has not been progressing smoothly; the initially launched features like “news summaries” have been questioned due to errors.

Worse still, when Meta suddenly launched an “AI talent war” last year, Apple also suffered heavy losses. The most well-known is that Apple’s AI foundational model team leader, Pang Ruoming, was poached. It is said that Meta offered him a sky-high salary of $200 million.

In this situation, continued delays would not only mean falling behind in the AI race but also greatly damage market trust.

At this point, seeking help from third parties to push forward with the most urgent tasks (such as the new Siri) according to the schedule and delivering results to users is Apple’s top priority.

And choosing Google is not surprising either.

Apple and Google have a long-standing partnership, such as Google paying Apple over $20 billion annually for years to be the default search engine on Safari.

Their cooperation has previously faced antitrust lawsuits, but the latest rulings only made some adjustments to specific practices. The $20 billion partnership remains permitted.

Moreover, Gemini is currently one of the most powerful models. According to previous reports, Apple tested several mainstream models including Gemini, GPT, Claude, and others, ultimately selecting Gemini.

By the way, in the current cooperation involving Google Gemini, Apple emphasizes “user privacy,” stressing that everything runs on “Apple’s private cloud” and no data is shared with Google. This is a lesson learned from past collaborations. In 2012, there was the “Safari tracking cookie scandal,” where Google was accused of exploiting Safari browser privacy loopholes to place tracking cookies, bypassing default policies to block third-party cookies. Google was fined $22.5 million by the FTC and required to make rectifications. Additionally, Google reached a $17 million settlement with 37 US states and the District of Columbia to resolve a class-action lawsuit.

In summary, this collaboration between Apple and Google can be seen as a temporary “bow,” but not “surrender” or “abandonment.”

Apple is still developing its own trillion-parameter models, expected to be launched around 2027.

Even after the model is released, Apple is likely to retain Gemini for reinforcement training, comparison, and further optimization. Using it as a “tool” to strengthen itself is what Apple expects from Google Gemini.

The happiest is still Google

Although we just spent a long time trying to clarify what this deep cooperation between Apple and Google is and isn’t,

it is foreseeable that the perception of “Apple AI relies on Gemini” will be hard to dispel.

Some can’t help but imagine: what if, by March or April, Siri suddenly becomes super useful…

It’s not hard to imagine. The current awkward situation Apple faces is that if introducing Gemini truly has an immediate and significant effect, accelerating the progress of Apple Intelligence and Siri, then the reputation of “Apple is not capable” will be confirmed.

In this case, even if Apple successfully develops that trillion-parameter self-developed model, transitioning from a phase of reliance to independence, Gemini will still be regarded as the hero, not Apple itself.

As for Google, the market has already given positive feedback.

After the joint statement from Apple and Google was released, Alphabet’s stock price rose by 1.7% at one point, and by the end of the day, its market value broke through $4 trillion for the first time, setting a company record, while Apple’s stock price increased by less than 1%.

This collaboration also put other competitors under pressure.

Elon Musk, who publicly expressed dissatisfaction, is one of them. He criticized Google, which controls Android and Chrome, for now invading Apple, calling it a “concentration of power” and implying potential monopoly.

Another less fortunate person is Sam Altman of OpenAI. Previously, OpenAI had a partnership with Apple, where ChatGPT would serve as a “supplementary option” when answering user questions via Siri. That is, just as Siri sometimes lists web search results, in the future, it might give ChatGPT’s answers when necessary.

This partnership once brought a lot of glory to OpenAI, but now, in the face of Apple’s and Google’s new cooperation, it looks pale in comparison.

Clearly, this “supplementary option” cooperation on the front end is far less significant than Google Gemini’s role as a “foundation helper” behind the scenes. It’s like Apple hosting a performance, with OpenAI performing on stage, but Google behind the scenes as the lighting technician. Altman has not yet commented on this.

And in the Chinese market, whether Apple is also seeking a “transition helper” before developing its own trillion-parameter self-developed model remains to be seen.