- → At WWDC on June 8, 2026, Apple unveiled a Siri completely rebuiltbilled as Apple Intelligence’s biggest leap.
- → The new in-house models are co-developed with Google Gemini ; the server model runs on GPUGraphics Processing Unit, a highly parallel chip that has become the standard for training AI models. Nvidia in Google Cloud, wrapped in Private Cloud Compute.
- → Apple-Google multi-year agreementabout $1 billion per yearconfirmed by a rare joint press release.
- → Via the Extensions, the user chooses his engine: ChatGPT, Gemini or Claude (Claude’s first native integration at Apple).
We’ve been waiting for him for two years. At WWDC on June 8, 2026, Apple finally presented the Siri that the brand promised: a completely rebuilt assistant, capable of understanding contextThe context window: the maximum number of tokens that a model can process in a single request (eg: 200k, 1M).act in applications and generate images. But the real surprise isn’t in the demo. It is in the engine: to achieve this, Apple has joined forces with its great rival, Google.
It is a strategic admission as much as a technical feat. Apple, which swore to build its intelligence alone, ended up relying on Gemini. Deciphering a shift that affects nearly two billion devices.
June 8 · announcement at WWDC
~$1 billion/year · the agreement with Google
3 engines · ChatGPT, Gemini, Claude
iOS 27 · deployment
01, The new Siri
An assistant redesigned from scratch
The new Siri is no longer a simple trigger for commands. It is now a full-fledged application on iPhone, iPad and Mac, capable of understanding complex requests in natural language, keeping the context of a conversation, taking action across applications and generating images. Apple presents it as the biggest upgrade to its Apple Intelligence offering since its launch.
The objective is clear: to immediately catch up with ChatGPT, Gemini and Claude, who have accustomed the general public to much more capable assistants than the old Siri.
02, The deal with Google
A billion dollars a year for Gemini
To power this Siri, Apple signed a multi-year agreement with Google, estimated at around a billion dollars per year, covering Gemini models and cloud infrastructure. An extremely rare occurrence between two giants waging a head-on war, the agreement was confirmed by a joint press release.
Concretely, Apple’s most powerful server model, called AFM Cloud Pro, runs on Nvidia Blackwell B200 GPUs hosted in Google Cloud. Everything remains wrapped in Private Cloud Compute, Apple’s privacy architecture, to preserve the in-house argument of data protection.
03, The nuance
“Not a drop of Gemini”: really?
Apple insists on one point: its new Apple Foundation Models are not a simple Gemini model filed as is. They are co-developed with Google, using technology distillationThe transfer of knowledge from a large model to a smaller one, to save cost and speed. and training, but remain, according to Apple, its own models.
The nuance matters for the brand image, but the substance is there: Apple is no longer building its foundation model alone. The champion of vertical integration, who mastered everything from silicon to software, had to open the door to a competitor’s AI to stay in the race.
04, The choice of engine
ChatGPT, Gemini or Claude: the choice is yours
This is the other big news. Via an Extensions system, the user can choose the AI engine that responds to the most complex requests: ChatGPT, Gemini or, for the first time in the Apple ecosystem, Claude from Anthropic.
ChatGPT historic integration, already present since Apple Intelligence; the default choice for many users.
Gemini now at the heart of the device, both the engine of the in-house models and a selectable option.
Claude great first: Anthropic enters the Apple ecosystem natively, a huge showcase for the model.
This open choice is good news for the user, and recognition that the battle of assistants is no longer won with a single model.
05, And in Europe
A more constrained deployment on this side of the Atlantic
As is often the case, Europe will move forward at a different pace. The European deployment of these functionalities will be more supervised, in particular by theAI ActThe 2024 European regulation which regulates the development and use of AI according to risk levels. the broadest phase of application of which arrives on August 2, 2026, and by the rules on competition and data protection. Apple has a habit of shifting or adapting certain Apple Intelligence functions for the European market.
For French users, the promise remains strong, but the exact date of full availability will depend as much on Brussels as on Cupertino.
To go further: The Gemini model sheet · The Claude model sheet · Compare AI models · ChatGPT vs Claude vs Gemini vs Mistral
Frequently asked questions
- What did Apple announce for Siri at WWDC 2026?
- Apple unveiled a completely rebuilt Siri on June 8, 2026: a full-fledged application on iPhone, iPad and Mac, capable of understanding complex requests, taking action in applications and generating images. This is the biggest upgrade to Apple Intelligence yet.
- Does Apple use Google Gemini for Siri?
- Yes, in part. The new Apple Foundation Models are co-developed with Google Gemini, and the most powerful server model (AFM Cloud Pro) runs on Nvidia GPUs in Google Cloud, wrapped in Private Cloud Compute. Apple specifies that these are not simple Gemini models, but its own models.
- How much does the Apple-Google deal cost?
- The multi-year deal is estimated to be worth around $1 billion per year, covering Gemini models and cloud infrastructure. Rarely, it was confirmed by a joint press release from the two companies.
- Can we choose another AI than Gemini on the new Siri?
- Yes. Via an Extensions system, the user can choose the engine that responds to complex requests: ChatGPT, Gemini or Claude. This is the first native integration of Claude (Anthropic) into the Apple ecosystem.
- When will the new Siri be available?
- The new Siri arrives with iOS 27 and future Apple systems. The European deployment will probably be later and more regulated, in particular because of the AI Act and the rules on competition and data.
- Why is Apple partnering with a rival like Google?
- Because Apple has fallen behind on generative AI compared to ChatGPT, Gemini and Claude. Co-developing its models with Gemini allows it to catch up quickly, even if it means giving up its habit of building everything alone.
- Will the deployment be different in France?
- Yes, probably more supervised and later. The European Union imposes a stricter framework (AI Act, competition, data protection), and Apple regularly shifts certain Apple Intelligence functions for the European market.

