How Apple’s late AI adoption could transform from a weakness to a strength

Apple will turn its Siri assistant into a full-fledged chatbot by next year. The company is working on a personal AI device to compete with the one OpenAI is building with Jony Ive. And Apple is putting control over its AI strategy into new hands within the company. So say a flurry of new reports, all advancing the larger story that Apple is doing what it can to get itself back in the AI race. 

And it’s doing it in a way that may allow it, in classic Apple fashion, to lead from behind. That is, it may hang back and benefit from the hard lessons learned by others marketing a new technology, then arrive fashionably late with a more polished product. 

Apple and Google announced on January 12 that the (notoriously slow) Siri assistant will be powered, at least in part, by Gemini models developed by Google’s DeepMind division. Apple has in the past voiced concern about the privacy implications of sending user data to AI models outside its own infrastructure. Apple has said it plans to run its AI models either within a secure Apple cloud, or, even better, on chips inside Apple devices. 

Bringing in Gemini

But that may be changing. Bloomberg’s Mark Gurman reports that Apple is now in talks with Google to run the Gemini models powering Siri and Apple Intelligence features within the Google Cloud. Previous reports said Apple could be paying Google as much as $1 billion per year for access to the Gemini models. 

The “new” Siri is expected to show up with iOS 26.4 in March or April, the report states. The assistant will reportedly gain a better contextual understanding of the user by accessing some types of personal data stored on the user’s device. It may also have an awareness of what the user is viewing or working on on their screen, as well as better internet search. These are the same “Apple Intelligence” features the company promised to deliver in 2024, but later postponed, explaining that it wasn’t happy with the performance and reliability of the AI.

Then, in 2027, another upgrade will make Siri feel more like a real chatbot, meaning that users will be able to have extended back-and-forth exchanges with the assistant (including via their voice), as is common with OpenAI’s ChatGPT and Google’s Gemini chatbots. Apple also plans to integrate the smarter Siri deeper into the operating system, which could make it a more functional intermediary between the user and the capabilities of the device. 

The Information reports that Apple is also in the early stages of developing a small personal AI device–about the size of an AirTag–that can clip to a lapel and contains two cameras, three microphones, a little speaker, a battery, and inductive charging tech. (A patent search yielded no Apple designs fitting this description.) OpenAI attracted a lot of attention last year after announcing that it was developing a personal AI device in collaboration with Apple’s former design guru Jony Ive. It’s worth noting, however, that other companies have tried selling such a product, notably Humane (founded by some ex-Applers), and none have found much success.

A new AI leader inside Apple

The Google Gemini deal and the AI device reports come in the wake of a pretty major power shift at Apple regarding the company’s AI strategy. Apple struggled for years to build its own AI models under the leadership of ex-Google AI chief John Giannandrea (while balancing its historical concern for data privacy), and failed to deliver models that performed like those from OpenAI, Anthropic, and Google. 

Now Apple has reportedly put its consumer AI problems in the lap of software chief Craig Fedherighi, who is known for his tall hair and utilitarian (and somewhat skeptical) views on the new technology. Fedherighi has viewed AI as an enabling technology that should work behind the scenes to make phone features work better. He’s also expressed concern over the predictability and reliability of the technology.  

Fedherighi is taking the reins at a pivotal moment. Apple is in a tough spot with AI. It fears the appearance of falling further behind OpenAI (and the punishment it might take from Wall Street), but it’s also traditionally hesitant to rush into an emerging technology that isn’t yet totally proven and reliable (AI chatbots still make mistakes and consumers don’t fully trust them). Apple is most comfortable taking a mature technology (like cell phones) and reinventing it for the mainstream with simplicity, utility, and artful design. 

So partnering with Google, talking about plans for new Gemini-powered features in the future, and preparing a personal AI device might be just the right moves for Apple right now. The deal with Google buys Apple time while its own researchers find ways to balance the twin needs of data privacy and high-performing models. (It might also shift some of the responsibility onto Google if the new Siri doesn’t work as promised.) 

Being late has worked out before

There’s precedent for this. Apple relied on Intel processors in its computers while it built up the expertise and experience to make its own. Apple used Intel processors in its Mac computers for 15 years, before switching its lineup to Apple-designed chips. But planning the features of future Macs was difficult because it all depended on Intel’s roadmap for releasing new chips. Apple became acutely aware of the speed and efficiency gains to be had from designing custom chips that could be deeply integrated with its Mac operating system. Those improvements were first realised in the company’s first M-series Macs in late 2020. 

Apple also relied entirely on Qualcomm cellular modem chips for the iPhone before it was able to build its own. The first Apple-designed modem, the C1, shipped inside the iPhone 16e in 2025. While Qualcomm says it will continue to supply modems for iPhones in 2026, Apple’s intent is to build its own modem into the integrated silicon “system-on-a-chip” processor that powers iPhones, which could yield faster and more reliable cellular connections.

However, in other cases, like internet search, Apple has been content to rely on Google as a partner. Who knows, Apple may believe generative AI will become a commodity in the future (models are indeed getting more efficient and cheaper to access), something it would rather buy than build. 

The news of an Apple AI device doesn’t hurt either. It shows an Apple that still has the appetite to mold emergent technologies into its own image and bring them to the mainstream, even without Ives.

Apple’s relationship with generative AI is seen as rocky and mainly unsuccessful so far. That narrative may have seeped through the shell of Apple’s spaceship in Cupertino and caused a rush to ship a technology that wasn’t quite baked. The hype around generative AI is so thick now that we’re quick to judge any technology company that isn’t betting the farm on it.

But it’s part of Apple’s culture to take the long view on new technology waves (notice that it didn’t change the company name when the “metaverse” was having its moment). Its hesitation, intentional or not, may give it more time to judge the real scale of the AI revolution, and more time to understand what it all should mean to Apple and its customers.