The big question about which artificial intelligence engine will power the next Siri on the iPhone has finally been answered. Apple has opted to rely on Gemini, Google's family of models, to shape the new generation of its assistant and incorporate many of the functions of Apple Intelligence, in what represents one of the most striking technological agreements in recent years.
The collaboration, which has been confirmed by both Cupertino and Mountain View, paints a picture of Siri evolving from a limited and somewhat outdated assistant into a much more advanced conversational system , with greater contextual memory, the ability to handle complex tasks, and deeper integration across iOS, iPadOS, and macOS. However, this also raises reasonable concerns about how privacy will be managed, given its reliance on the cloud infrastructure of one of its biggest competitors.
An announcement coming from the top of Google Cloud
The most direct confirmation of this move came during Google Cloud Next, Google Cloud's major annual event held in Las Vegas . There, Thomas Kurian, head of this division, officially confirmed an alliance that had been circulating in leaks and industry rumors for months.
Kurian spoke of a “monumental partnership with one of the world’s most iconic brands” to bring Gemini technology to millions of users, clearly referring to Apple. He explained that Google will act as the cloud provider for the development of the next generation of Apple Foundation Models, foundational models that will be built on Gemini and form the backbone of Apple Intelligence.
In his remarks, the executive emphasized that these models will drive new intelligent features across the entire Apple ecosystem , including a more personalized Siri that will reach users throughout this year. His words align with what Apple has been hinting at in recent months, but they provide further support for the timeline and scope of the collaboration.
The significance of the announcement lies not so much in revealing something radically new, but in Google's public acknowledgment of Gemini's role as a key component in the future Siri . Until now, most details came from leaks, analysts, and sources close to the project; hearing a senior Google executive confirm it in an official setting reinforces the roadmap's credibility.
Gemini as the engine of the new Siri and Apple Intelligence
The agreement between the two companies means that Gemini will become the engine of the new Siri , leaving behind the previous, more limited approach with fewer generative capabilities. The idea is not for Google to replace Apple in the assistant's design, but rather for its models to serve as a technological foundation upon which Apple will build its own experiences and customization layers.
In practice, the alliance will result in two major levels of integration . On the one hand, a lighter version of Gemini will support the first wave of Siri's enhanced features, focused on more accurate responses, greater context, and basic personalization. On the other hand, more advanced models will be reserved for the later phase, when the assistant will more closely resemble a conversational chatbot integrated into the system.
This strategy aligns with Apple Intelligence's vision: combining proprietary models with third-party technology when necessary , while continuing to develop its own AI infrastructure in the medium term. This allows Apple to gain time and ensure an immediate leap in quality, while simultaneously working on internal engines to reduce its future reliance on Google's cloud.
According to the information that has come to light, the collaboration is not limited to a static model. The foundational models based on Gemini will evolve over time , with new versions and adjustments, as Apple advances in its own developments. In other words, it's not a simple "plug and play" solution, but a living platform on which to build more features.
Calendar: From the first push in iOS 26.4 to the big leap in iOS 27
The rollout of this new generation of Siri is planned in two clearly differentiated phases , linked to specific versions of Apple's operating systems for mobile phones, tablets, and computers.
In the first stage, the company plans to introduce an improved Siri with a lighter Gemini model through iOS 26.4. This update, which Apple initially scheduled for spring, is designed to offer more accurate responses, greater understanding of natural language, and a somewhat more personalized experience than the current one, without yet reaching a complete overhaul of the assistant.
The second phase will occur with the release of iOS 27 and its equivalents on iPadOS and macOS . At that point, the new Siri will rely on more advanced Gemini models and become much closer to an integrated chatbot, with expanded capabilities and smoother interaction. Internally, this development is codenamed “Campos.”
Google has reinforced the idea that a more personalized Siri will arrive toward the end of the year , which aligns with the usual fall release window for major iOS updates. Furthermore, Kurian reiterated that Google Cloud will be Apple's preferred provider for hosting these models, a key element for deploying the new features globally without overloading the infrastructure.
Although there have been delays and internal adjustments — mainly due to accuracy and performance issues detected in testing — Apple is still within the timeframe it publicly set : a thoroughly revamped Siri during 2026, with the big showcase planned for WWDC and the massive rollout linked to the arrival of iOS 27.
This is how Siri will change: from limited assistant to system chatbot
Beyond the dates, what's truly relevant for users is how the assistant's user experience will be transformed . The new Siri won't simply be "the same old Siri, but a little smarter," but will adopt an approach much closer to that of current generative AI chatbots.
In the first phase, the assistant is expected to gain better comprehension of chained requests, improved short-term memory within a single conversation, and a greater ability to interpret the context of what is being discussed. The classic colored ball at the bottom of the screen will be replaced by its own application, with its own interface, allowing for more natural conversations.
With the arrival of iOS 27, things will go a step further: Siri will begin to behave like a chatbot integrated into the system , capable of interacting with applications, understanding what's on the screen, and accessing information stored on the device to tailor its responses. Without changing how it's invoked (by voice or with the power button), the assistant will become a central part of the experience on iPhone, iPad, and Mac.
Among the features being considered for this advanced phase are web-based information retrieval with generated summaries, on-demand image creation , writing and summarizing long texts, document review, and even generating small code snippets. All of this would be integrated into a conversational interface that eliminates the need to constantly switch between apps.
The difference with respect to the current Siri will not only be quantitative, but qualitative: the assistant will cease to be an accessory utility and will become an intelligent layer on top of the operating system , capable of coordinating several tasks at once, remembering the recent context and acting as a kind of "brain" that orchestrates the information and actions of the device.
Conversational memory and privacy limits
One of the most significant changes will be the management of conversational memory . Like other advanced AI systems, the new Siri will be able to remember past interactions for a certain period of time, allowing it to continue topics, resume tasks, or refine suggestions based on what has been discussed previously.
This storage will not be unlimited. Apple plans to set a specific limit to prevent the accumulation of excessively long chat histories , something that could conflict with its privacy commitments. The exact limit and how these conversations will be deleted or anonymized have not yet been made public, raising questions among users and experts.
The company insists that separating user identity from processed data will be a key design element . Information that the iPhone cannot handle locally will be sent to the cloud without being associated with a recognizable personal profile, so that the models can learn from usage patterns without building a detailed record of each person.
In practice, this will mean that much of the data feeding Siri's AI will travel "de-identified." Apple promises that neither Google nor third parties will be able to link these requests to a specific user , and that the system will not use photos, messages, or other sensitive content to train models in an identifiable way.
It remains to be seen how all this will translate into concrete options in the system settings, what degree of control the user will have over memory erasure, and to what extent there will be transparency regarding the amount of information that reaches the cloud and for how long it is kept.
Google's cloud, the Achilles' heel for privacy
The most delicate aspect of this agreement is precisely that Apple will have to rely on servers belonging to its main rival in the online advertising business . Although the Cupertino company has for years championed privacy as one of its defining characteristics, it is now forced to cede part of Siri's processing to Google's infrastructure.
According to reports, Apple will pay a very large sum—rumored to be around one billion dollars annually —to use Google's data centers almost exclusively for these functions. The goal is to ensure the service doesn't become overloaded, while also establishing a clear separation from the rest of Google Cloud's workloads.
Apple insists that its private cloud architecture and anonymization mechanisms will secure the process , preventing Google or other parties from snooping on the data passing through those servers. The idea is that the phone will handle everything it can with its own processor and only resort to the cloud when it needs additional power.
Despite these assurances, it is still striking that a company so protective of its ecosystem would turn to Google to support a critical part of its AI strategy . The company has had to accept that, at least in the short term, it lacks the necessary infrastructure to single-handedly handle the volume of requests that the new Siri will generate.
Meanwhile, Apple continues to explore ways to reduce this dependence in the future. The medium-term plan involves developing more of its own models and strengthening its local processing capabilities on devices, so that the cloud is increasingly limited to highly complex tasks or specific high-load scenarios.
A shift in Apple's strategy and a triumph for Google
The decision to build the new Siri on Gemini represents a significant shift from Apple's traditional self-sufficiency . For years, the company has sought to develop its own key technologies in-house, avoiding over-reliance on third parties in strategic areas.
In the field of generative artificial intelligence, however, competitive pressure and accumulated delays have forced Apple to make a move. After analyzing alternatives from players like OpenAI and Anthropic , the company has finally opted for Google Gemini's proposal, both in terms of models and cloud infrastructure.
For Google, the deal represents a significant victory in the race to become the leading AI provider . The integration of Gemini into millions of iPhones, iPads, and Macs returns Google to a central role within the Apple ecosystem, this time extending beyond the search engine and pre-installed apps.
At the same time, Apple is maintaining room for maneuver: the company makes it clear that Gemini is an important pillar in this phase, but not the final destination . The long-term goal is to have its own more powerful engines and an infrastructure that allows it to regain complete control of the process, without giving up on leveraging what Google has already built.
All this movement comes at a time when Siri has clearly fallen behind other assistants and chatbots , both in accuracy and capabilities. The new phase that begins with Gemini is, for Apple, an opportunity to close the gap and, if successful, offer a differentiated experience thanks to deep integration with its systems.
The result of this alliance will be a much more ambitious assistant, supported by the power of Google's models but filtered through Apple's obsession with product control and privacy. It will reach European and Spanish users in several waves throughout 2026, coinciding with the new versions of iOS, iPadOS, and macOS, and with a special focus on WWDC, where the company is expected to finally show what the new Gemini-based Siri is capable of.