Both iPhone and Android Now Run on Gemini

Two things happened over the summer of 2026.

• 8 June, Apple unveiled its rebuilt Siri at WWDC. The hardest reasoning is handed off to Google’s Gemini models
• 12 August, Google announced the Pixel 11 series and Gemini Intelligence at Made by Google 2026

Taken separately they are each a product announcement. Put together they make a different picture. The two major mobile platforms now run on the same model family.

What Google shipped: agents became the default

The important part of the Pixel 11 announcement was not the hardware but Gemini Intelligence. This is not a chatbot but an agent framework that carries out tasks on the user’s behalf.

The hardware changes are aligned to that purpose.

• On-device AI processing up to 3.5× faster. It means the model running inside the device got bigger
• Real-time sign language translation, and a Magic Capture camera that analyses 400 frames
• A Pixel 11 Pro Fold about 1mm thinner and 10% lighter than its predecessor
• “Pixel Glow” LEDs that use light and colour around the camera bar to signal notifications

The key number here is 3.5×. For an agent to read the screen and drive apps, the model has to run constantly, and locally. Sending everything to a server is slow, expensive and a privacy problem. On-device performance is not a feature but a precondition.

What Apple chose: borrow rather than build

Apple went the opposite way. Instead of pushing ahead with its own model, it signed a multi-year deal with Google and put Gemini-based models into Siri’s foundation layer. The deal is reported at around $1 billion a year.

The capabilities announced for the new Siri are these.

• Search that uses personal context across messages, mail and photos
• Executing tasks that span multiple apps
• Answering questions about what is on screen
• Looking up information on the web

The architecture combines on-device processing with Private Cloud Compute. Rollout is planned alongside the iOS 27 family.

Seen coldly, there is one reason Apple made this choice. It was late. The Siri overhaul was postponed repeatedly, and in the meantime the gap against competing products became visible to users. It judged the cost of borrowing and buying time to be lower than the cost of continuing to build alone.

Three consequences of this arrangement

1. The axis of differentiation moves from model to integration

When everyone uses the same model, answer quality is a hard place to differentiate. So what separates them? How deeply they reach the data on the device, and how safely they can drive apps. That is a fight the owner of the operating system wins, and a fight third-party AI apps lose.

2. Google’s position has become strange

Google competes with Apple on Android while simultaneously selling Apple its models. The structure resembles the default-search-engine deal. The revenue is stable, but it is not a picture competition authorities enjoy.

3. Users get fewer choices

When the default assistant is embedded in the operating system, most users simply use it. People who install a separate AI app remain a minority. That is not a question of performance but of position.

What actual users should check now

As platform assistants shift toward reaching into personal data, the number of settings worth checking goes up.

1. Scope of personal context access. Check whether you can switch off individual categories — messages, mail, photos — rather than all or nothing
2. The line between on-device and server processing. What finishes inside the device and what leaves it is written in the policy documents. You cannot tell from the screen
3. The agent’s app-driving permissions. A feature that presses buttons for you also makes mistakes for you. For apps involving payments or transfers, leaving a confirmation step in place is wiser
4. Wait before changing your existing habits. New assistants change features frequently in their first few months

In short

1. Google shipped Gemini Intelligence with the Pixel 11. Raising on-device AI processing speed up to 3.5× is the crux
2. Apple chose a multi-year deal putting Gemini-based models into Siri instead of building its own. It ships alongside the iOS 27 family
3. As a result both major mobile platforms run on the same model family
4. The competitive axis moves from model quality to device data access and app integration. Whoever owns the operating system has the advantage
5. Users need to check personal context access scope and agent permissions category by category

Frequently asked questions

With a Gemini-based Siri, is Google seeing my data rather than Apple?

Apple has described an architecture combining on-device processing with Private Cloud Compute to limit how far personal data travels outside. That said, which queries take which path is not shown on screen, so the exact data flow has to be checked in the published policy documents.

Do I have to buy a Pixel to use Gemini Intelligence?

Gemini itself is available on other Android devices and through apps. Features that lean on on-device processing depend on hardware performance, though, and the pattern has repeatedly been that they arrive first on the newest Pixel and expand later.

Do I still need the AI apps I already use?

What a default assistant covers is everyday tasks. For long-form document work, coding and field-specific uses, dedicated tools are still better. It is more accurate to see the default assistant as a default rather than a replacement.

Isn’t it risky to let an agent drive apps for me?

The convenience and the risk come from the same place. For hard-to-reverse actions such as payments, transfers and sending messages, keeping a user confirmation step is safer, and most platforms let you adjust that step in settings.