Siri has a new brain, and it did not come from Cupertino. On Monday, iOS 27 rolled out worldwide as a free update, and the update everyone actually cared about was buried inside it: a rebuilt Siri running on generative AI models built with Google’s Gemini and, for parts of the on-device pipeline, Nvidia hardware. For a company that has spent a decade insisting it builds its own silicon, its own chips, and its own software stack from the ground up, handing the brain of its most personal product to a rival is not a small decision.
It is also not the whole story. Siri AI landed in English only. It is gated behind an opt-in waitlist rather than switched on for everyone. Europe does not have it yet, and neither does China. If you have been reading about this rollout for months, none of that should surprise you. What changed this week is that the promises became a product people can actually open and talk to.
Quick facts
- iOS 27 shipped worldwide on September 14, 2026, as a free over-the-air update
- Its centerpiece is a rebuilt Siri, developed under a multi-year Apple-Google partnership using Gemini models
- Siri AI is English-only at launch, opt-in, and reportedly gated behind a waitlist for early users
- Five more languages, French, Japanese, Korean, Portuguese, and Spanish, are scheduled for October
- The feature is unavailable in the European Union and China at launch, for regulatory and market reasons
- It requires an iPhone 15 Pro, iPhone 15 Pro Max, or iPhone 16 or later; the iPhone 11 and similar older models get iOS 27 without it
- New capabilities include cross-app actions, on-screen awareness, and use of personal context stored on the device
Why Apple picked a rival’s model
Apple confirmed the Gemini partnership back in January, and Google followed up in April with its own confirmation that the deal was moving forward. The arrangement lets Apple base a new generation of its Apple Foundation Models on Google’s Gemini technology and cloud infrastructure, while Apple keeps control of the parts it has always cared about most: on-device processing, personal context, and the privacy story it tells customers. Reporting on the deal has put its value at roughly a billion dollars a year, which is a strange number to sit with. Apple, a company that made its own chip business a competitive weapon, is now paying a search company for the model powering its assistant.
The honest reading is that Apple tried to build this in-house and ran out of runway. The original Siri overhaul was promised for 2025, slipped repeatedly, and by the time reporting confirmed which iPhones would even qualify for the new assistant, the pattern was already familiar: big promise, real delay, narrowing scope. Borrowing Gemini was the fastest way to close the gap with what ChatGPT and Google’s own Assistant could already do, even if it meant an uncomfortable admission about where Apple’s AI research actually stood.
What the new Siri can do that the old one could not
The feature list reads less like a chatbot upgrade and more like an attempt to make Siri a coordinator across the whole phone. It can pull a photo out of your library and drop it into an email without you naming the file. It can read your notes and turn loose plans into an actual schedule. It is meant to understand what is on your screen right now, not just what you say out loud, and to carry context from one app into the next instead of treating every request as a blank slate.
| Capability | Old Siri | Siri AI (iOS 27) |
|---|---|---|
| Cross-app actions | Limited, mostly single-app shortcuts | Can move data between apps, e.g. photos into email |
| On-screen awareness | None | Understands what is currently on screen |
| Personal context | Basic contacts and calendar lookups | Draws on notes, messages, and files to build tasks |
| Underlying model | Apple’s own smaller models | Google Gemini, with on-device processing by Apple |
| Language support at launch | Dozens of languages | English only, five more arriving in October |
The rollout, mapped
Why Europe and China are still waiting
Apple had already signaled that Europe would sit this launch out, and that prediction held. The European Union’s Digital Markets Act imposes interoperability requirements that Apple has repeatedly said it cannot meet on the timeline it wants, particularly around how a third-party model gets access to on-device data and how competitors get equivalent access to the same hooks. Interestingly, macOS falls under different DMA obligations than iOS and iPadOS, so Mac users in the EU may see Siri AI before iPhone owners in the same country do. China’s absence is simpler to explain: Google’s services are broadly unavailable there, and Apple’s China-specific AI partnerships route through different local providers entirely.
Why the waitlist matters
Gating Siri AI behind an opt-in waitlist, rather than flipping it on for every eligible device, suggests Apple is managing either server load, model cost, or both. Gemini queries route through cloud infrastructure that Apple does not fully control, and a hard cap on early users is a cheap way to avoid a rocky first impression.
The hardware line nobody can upgrade around
The device requirement is the part users have the least patience for. iPhone 15 Pro, 15 Pro Max, and iPhone 16 or later qualify; an iPhone 11 gets the rest of iOS 27 but not the new Siri. Apple did something similar with watchOS 27, rebuilding Siri there too while quietly limiting which watches get the full experience. The pattern across both platforms is the same: the newest silicon runs the heaviest on-device processing, and everything older is left with a lighter, less capable assistant, regardless of how recently it was purchased.
What to watch next
- How long the waitlist stays closed. A fast expansion to all eligible users would suggest server capacity is holding up better than expected.
- Whether the EU gets a firm date. Apple has resolved DMA disputes before by building region-specific versions of features. Watch for a scaled-down EU release rather than an indefinite wait.
- Real-world reliability of cross-app actions. Pulling photos into emails and building schedules from notes sounds simple in a keynote demo and tends to be the first thing that breaks in daily use.
- Whether Google’s involvement becomes more visible. Apple has been careful to keep the Gemini branding out of the user-facing experience. That may not hold if the model underperforms and users start asking why.
For now, the headline is simple enough. Apple’s most personal piece of software runs on someone else’s model, most of the world cannot use it yet, and the company that spent years promising to build everything itself just admitted, in the most public way possible, that it needed help.
Sources and further reading
- Apple Newsroom: Siri AI arrives with iOS 27
- Engadget: Siri AI is here as Apple releases iOS 27
- 9to5Mac: Requirements to get Siri AI in iOS 27
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.
