The alliance between Apple and Google in artificial intelligence It's now official, and it goes far beyond a simple one-off agreement. The Cupertino company has decided to rely on GeminiGoogle's generative AI, to make the leap it had been postponing for years with Crab and with its platform Apple IntelligenceThe move marks a turning point in Apple's strategy and reshapes the AI landscape globally.
After months of rumors, leaks, and delays, Apple has confirmed that the next generation of Siri and its Apple Foundation Models It will be built on Gemini's models and Google's cloud technology.However, it will do so under its own rules: the models will run on Apple's infrastructure, will be integrated invisibly to the user, and will maintain the company's traditional focus on privacy and ecosystem control.
A multi-year agreement that changes Apple's strategy
Apple and Google have announced a collaboration of several years through which the new batch of Apple Foundation Models will be based on Gemini and the Google's cloud infrastructureIn the words of both companies, after a detailed evaluation of different options, Google's AI technology provides "the most capable foundation" and "the strongest foundation" for the models that Apple needs for the next phase of Apple Intelligence.
According to various industry sources, including previous leaks from Bloomberg and CNBCThe agreement would imply that Apple could end up paying around 1.000 million dollars per year by licensing and adapting Gemini's models. There's even talk of access to a model of around 1,2 billion parameters, far exceeding the approximately 150.000 billion parameters of the most powerful internal models that Apple had on the table until now.
This shift comes after the company acknowledged, more or less explicitly, that its Internal development of foundational models was not progressing at the expected paceThe Siri update and the more ambitious features of Apple Intelligence, which were initially slated for 2025, have been postponedThis left the company in an awkward position given the speed with which competitors like OpenAI, Microsoft, and Google were deploying generative products.
The alliance with Google also fits with other previous ties between the two tech companies, such as the agreement that makes Make Google Search the default search engine in Safari in exchange for several billion dollars annually. Now, the agreement extends to an even more sensitive and strategic area: the artificial intelligence that will shape the experience of using an iPhone, an iPad, or a Mac.
How Apple will use Gemini: a background brain, no Google logo
One of the key aspects of the agreement is that Apple will not "integrate Gemini" as Android users know it.Instead, it will license the model to build its own Apple Foundation Models on top of it. Technically, Google provides the base model—the "brain"—and Apple handles the rest. post-training: adjust its behavior, its response tone, and its deep integration with iOS, iPadOS, macOS, and the rest of the ecosystem.
Several experts who have analyzed the architecture, such as the developer Max WeinbachThey explain that Apple applies a kind of harness Regarding Gemini: the model retains its reasoning and language comprehension abilities, but learns Apple's "house rules." This includes everything from the style of responses to the limits of what it can do with personal data and how it interacts with apps like Messages, Mail, Reminders, and Shortcuts.
Therefore, The user will never see any visible Gemini branding on their iPhone.There will be no "Powered by Google" labels or a name change for the assistant: Siri will still be SiriAlthough internally it uses a customized version of Gemini as the engine for certain tasks. Apple insists that what it has done is license a Gemini model to make it part of their Apple Foundation Models, do not delegate the user experience to an external service.
In addition, the company plans distill lighter versions of that technology to run them directly on the device, similar to what Google does with Gemini Nano on Android. This will allow some of the intelligence to run locally, offline, maintaining performance and reducing constant reliance on the cloud for less demanding tasks.
A much smarter Siri: what will change with Gemini
The immediate focus of the agreement is on relaunch Siri as a truly capable and conversational assistantApple has been promising a "smarter" Siri for years, but progress has been gradual, and the general perception is that the assistant has fallen short compared to the experiences offered by ChatGPT or Gemini himself on other devices.
With the new architecture, Siri will combine Apple's internal models with Gemini's powerApple's AI models will handle tasks closely tied to the device and personal context (for example, switching between apps, running shortcuts, managing notifications, or performing very specific actions on local data), while Google's AI will take over functions such as the content summary, complex planning, or multi-stage reasoning.
In practice, this will allow the assistant Understand natural language better, handle longer and more complex requests, and maintain smoother conversationsThe idea is that Siri can understand context across multiple interactions, retrieve relevant information from apps and documents, and suggest next steps without needing to repeat commands all the time.
Sources cited by media outlets such as CNBC indicate that Apple will have access to a model of about 1,2 billion parametersThis represents a significant leap forward compared to their previous models. This difference in scale should translate into More accurate answers, shorter waiting times, and a better ability to "connect the dots" between different queries, both in Spanish and in other European languages.
The Siri version codenamed "Linwood" It is being developed as part of this transition. Internally, the idea is that this new Siri can act almost like an agent that orchestrates different actions on the device: from organizing an entire trip by combining emails, maps, and calendar, to preparing personalized daily summaries with the information that interests the user most.
Calendar: iOS 26.4, iOS 27, and a two-stage transition
Apple has not yet given a closed date for the final rollout of the new Siri, but the pieces of the timeline are starting to fall into place. Various analysts and leaks suggest that The first major wave of changes will arrive with iOS 26.4, an update that would be located in the first quarter of 2026, probably between March and April, after a beta testing period that could start at the end of January or the beginning of February.
This first phase would have a clear objective: "rescue" SiriThat is, to bring the voice assistant to a competitive level without dismantling the rest of Apple Intelligence. Current Apple models would continue to support many existing features, while The advanced reasoning and natural conversational aspects would begin to rely on Gemini..
In the medium term, the company is preparing a second stage with iOS 27where the use of Gemini would extend beyond Siri to encompass New Apple Intelligence FeaturesThere is talk of profound improvements in text and image generation, more sophisticated writing assistance tools, advanced automation through Shortcuts, and even tighter integration with productivity applications.
This phased transition follows a rather pragmatic logic: It doesn't make sense to redesign the entire AI platform overnight. when millions of iPhones, iPads, and Macs depend on it. First, the most visible part—Siri—is reinforced, and then the new "brain" is rolled out to the rest of the system as performance, stability, and acceptance of the change are validated.
For users in Spain and the rest of Europe, the rollout of these new features It will also be conditioned by the European regulatory framework and by the requirements set by regulations such as the AI Regulation or the Digital Markets Act. It wouldn't be surprising to see a phased rollout by region or certain features temporarily limited until approval is received from the relevant authorities.
Privacy and technical architecture: three levels for processing your data
The big question, especially in Europe, is What about privacy? when the AI that powers Siri comes from a third party like Google. Apple and Google have tried to dispel doubts from the outset, insisting that Apple Intelligence will continue to run on devices and in Private Cloud ComputeApple's private cloud platform designed specifically to securely process sensitive data.
The architecture is based on three levels of processingIn the first one, the one closest to the user, we find the models that run directly on the device (on-device). These are lighter versions, derived or distilled from larger models, that run on the Neural Engine Apple's chips handle everyday tasks like rewriting messages, summarizing notifications, and making quick suggestions, all without the information leaving the iPhone or iPad.
The second level is Private Cloud ComputeApple's private cloud. Only when the device detects that a request requires more computing power—for example, due to the length of the text or the complexity of the reasoning—is the necessary data sent to this infrastructure. Beforehand, the system performs a cryptographic validation of the device to ensure it's legitimate. The servers, according to Apple, They do not retain a persistent memory of these requests.They process, respond, and discard.
The third level comes into play GeminiThe novelty is that Google's large-scale models will run within Apple's own private cloud.without the data having to go to Google data centers identifiable as such. That is, Apple licenses the model, not the service.In this way, the company seeks to demonstrate that it can use the foundation of an LLM as powerful as Gemini without personal data being mixed with other Google workloads.
To reinforce this approach, Apple and Google have indicated that, in addition to these cloud-based models, devices will be able to use Highly compressed versions, similar to Gemini Nano, on localThis approach breaks, in part, with the idea that powerful AI necessarily requires abandoning any control over the information being processed. Apple is thus trying to position itself at a middle ground: taking advantage of the most advanced "brain" available, but enclosed in what they describe as a kind of proprietary "cryptographic safe."
ChatGPT, OpenAI and the rest of the players: what changes now
The investment in Gemini doesn't erase Apple's other collaborations in the field of AI. As of today, Apple Intelligence already uses ChatGPT In certain circumstances: when Siri detects that a query requires very broad general knowledge or creative content generation that its models do not cover well, it can ask the user if they want to send the request to OpenAI's AI.
With the arrival of Gemini as Siri's central engineThis scheme becomes more complicated. On paper, the new assistant should be able to handle many of the requests currently routed to ChatGPT, reducing the practical need for that bridge. Even so, sources close to the industry suggest that Apple could maintain the integration with OpenAI for very specific scenarios or as a secondary option for those who already use the ChatGPT account in other contexts.
From a strategic point of view, the alliance with Google confirms that Apple has opted for a multi-vendor modelIt licenses technology from various players, adapts it to its ecosystem, and retains control of the final layer of experience and security. This position differs from that of companies like Microsoft, which is much more aligned with a single main partner (OpenAI), and also contrasts with the approach of companies like Meta, which promote open models.
This hybrid approach has advantages and risks. On the one hand, It provides flexibility to change partners if one falls behind. or if regulation forces adjustments. On the other hand, it adds technical complexity and can generate some confusion about exactly who is doing what in the background. For now, Apple is sticking to a simple message: The user will speak to Siri, not to Gemini or ChatGPT.and the system will decide which model to use at any given time.
In parallel, the agreement It strengthens Google's position in the battle for AI leadership.The news has had a direct impact on the markets: Google's parent company, A, has managed to surpass the $4 trillion market capitalization in intraday trading, a club that so far has only included companies like Nvidia, Apple and Microsoft.
Reactions, criticisms and potential regulatory impact
Beyond the positive reception on Wall Street, the Apple-Google deal has caused critical reactions within the technology sector itselfThe most visible one has been that of Elon Musk, founder of xAI —creator of the Grok AI— and owner of X (formerly Twitter), who has used the social network to rail against what he considers an excessive concentration of power in the hands of Google.
In response to Google's official announcement of the deal, Musk stated that "This seems like an unreasonable concentration of power for Google, given that they also own Android and Chrome."Their concern centers on the fact that the same company controls the AI that powers both the Android ecosystem and a large part of the smart features of Apple devices, reducing the space for alternatives like their own.
These criticisms add to regulatory debates are already underway in the United States and the European Union regarding the dominant position of large technology platforms. Competition authorities have closely monitored the agreement that makes Google Search the default search engine on Apple devices for years, and it wouldn't be surprising if now Take a close look at this new collaboration in the field of AI.
In Europe, the context is particularly sensitive. AI Regulation and Digital Markets Law They set strict standards on transparency, data use, and potential anti-competitive practices. An agreement under which A single AI provider becomes a central component of the platforms of two giants like Apple and Google It may raise questions about technological dependence, interoperability, and real choices for consumers and developers.
For now, both Apple and Google emphasize that Apple controls the experience, infrastructure, and privacy safeguardsand that Gemini is simply providing the underlying model. Even so, the true scope of the collaboration and its impact on competition will become clearer as the new Siri and upcoming Apple Intelligence features roll out to users' devices.
With this movement, Apple moves from a cautious stance on generative AI to a decisive commitment, albeit supported by an external partner.Meanwhile, Google is solidifying its Gemini technology as a benchmark for third parties. The new Siri, expected to begin rolling out throughout 2026 with iOS 26.4 and later, will be the first visible test of just how much this combination can deliver a truly more useful, privacy-conscious, and competitive experience in a market where assistants and language models have become the cornerstone of almost everything we do with our phones.
