- Google I/O reinforces Gemini as the central layer of the entire ecosystem, with new models, agents, and presence in Android, Chrome, Search, and hardware.
- The conference revolves around the cost of AI: Gemini 3.5 Flash and the strategy of mixing models seek to drastically reduce the cost of mass token use.
- Android 17, Googlebook and Android XR are moving towards a more "intelligent" system, with laptops and glasses focused on AI and Gemini Intelligence.
- Search, Workspace, Google Home Speaker, and new developer tools are updated with agentic AI and greater transparency in generated content.
The Shoreline Amphitheater in Mountain View has once again become the global epicenter of technology with the celebration of Google I/O 2026 , the company's major annual developer conference. For two days, Google outlined its roadmap for artificial intelligence, Android, Chrome, hardware, and developer platforms, with a very clear underlying message: AI is no longer measured solely by what it can do, but also by the cost of deploying it at scale.
The inaugural keynote, which could be followed live in Spain from 19:00 PM (peninsular time) via YouTube and the event's official website, confirmed what many already suspected: Gemini is solidifying its position as the core of the entire Google ecosystem. Beyond the spectacular demos, what made the difference this year was the emphasis on the economic efficiency of the models and how they are practically integrated into the products that billions of people use every day.
Gemini, omnipresent and increasingly agentive
From the very first minute, it was clear that Gemini would be the star of the show . Sundar Pichai opened the session by noting that more than 3,2 trillion tokens are already being processed monthly through Google's services, a figure that illustrates both the reach of the company's AI and the economic impact of keeping it running. According to the CEO himself, large Google Cloud customers are processing close to a trillion tokens daily, and that's where the new strategy comes in, with models designed to reduce costs.
The Gemini family has been strengthened with several key additions. One of these is Gemini 3.5 Flash , a model designed for extreme efficiency: Google presents it as four times faster than other reference models, with a cost that can drop to less than half or even a third in certain scenarios. Its role is not to replace the highest-capacity models, but rather to handle the massive volume of tasks where it's not necessary to squeeze every last bit of performance out of the processor.
In parallel, Google has announced the availability of Gemini 3.5 Flash across all its services starting today, while Gemini 3.5 Pro will arrive next month for those needing a more powerful, cutting-edge model. Building on this foundation is Gemini Omni and its first variant, Omni Flash, a multimodal model capable of working with text, images, video, and audio, designed for advanced creative experiences and content generation in multiple formats.
Another key element is Gemini Spark , a new interface that embodies the idea of agentic AI : instead of simply answering questions, it has access to information stored in the user's account and can perform actions on their behalf. Spark will be integrated directly into Google Chrome and Android in the coming months, and will be crucial for Gemini to move beyond feeling like an isolated chatbot and become a cross-platform assistant that interacts with real apps and services.
The Gemini app, which according to Google already boasts over 900 million monthly active users , is also taking a significant leap forward. It's receiving a redesign that had already begun rolling out on Android and iOS, new voice options with more natural accents from different regions—something especially relevant for Europe, where linguistic diversity is key—and the arrival of in-app agents, such as Daily Brief, an assistant that organizes news, tasks, and recommendations to help users start their day in a more structured way.
Between power and price tag: the new AI economy
Beyond the technical aspects, one of Pichai's most striking messages was economic. The executive explained that if large companies currently running workloads on Google Cloud migrated 80% of their workload to a combination of Gemini 3.5 Flash and high-end models like the 3.5 Pro , they could save over $1.000 billion annually in AI costs. This isn't a futuristic demo, but a very concrete proposal for finance departments that are starting to scrutinize token invoices.
Google's logic is simple: reserve the most expensive and complex models for the tasks that truly require them and offload everything else to options like Flash, optimized for cost and speed . The company itself insists that this is the same approach they already apply internally. In March, Google was processing half a trillion tokens a day within its organization. Now, according to Pichai, that figure exceeds 3 trillion daily, relying precisely on cheaper models to maintain the pace without skyrocketing costs.
There is, however, a flip side, which Pichai himself hinted at with compelling figures. In 2022, Google invested around $31.000 billion in capital . This year, the forecast is between $180.000 billion and $190.000 billion, multiplying the effort in data centers, custom chips, and energy sixfold. While promising its clients that they can reduce their AI spending by optimizing the use of the models, the company is making a historic investment to maintain the infrastructure that makes those savings possible.
The emerging business model is clear: Google absorbs the raw infrastructure costs and sells efficiency as a service. For many European companies adopting AI on a large scale, the promise is attractive: outsourcing the complexity of hardware and energy in exchange for a significant reduction in their computing bill. In return, dependence on the Google ecosystem increases, something the company itself doesn't hide, but presents as the logical way to scale AI projects without having to build their own cloud.
In this context, the Google AI Ultra subscription plan , which will cost $200 per month starting next month, is positioned as an option for users and businesses that need access to the most advanced models, while offerings like Google AI Pro and Flash-based APIs target those who want a balance between capability and budget.
Security, transparency and the fight against deepfakes
With the advancement of generative tools, one of the most serious challenges is the detection of synthetic content. At Google I/O 2026, the company reinforced its commitment to SynthID , its technology for identifying AI-generated images, audio, and video. In addition to existing partners like Nvidia, OpenAI, Eleven Labs, and Kakao have now joined the initiative, participating as promoters of this markup and identification standard.
In practice, this means that more and more platforms and models will incorporate invisible watermarks or metadata that allow users to recognize when a piece of content has been created or modified by AI. Google emphasizes that this is not just a technical decision, but a response to regulatory and social concerns, particularly intense in the European Union, where legislation on synthetic content and disinformation is becoming increasingly stringent.
Gemini itself is also becoming a verification tool. From now on, users can use Google's models to check if an image or video has been generated or altered with AI , with contextual explanations of why the system reaches that conclusion. This type of functionality fits with the company's strategy of presenting itself not only as a provider of models, but also as a responsible actor in managing the social impact of AI.
Pichai and Demis Hassabis repeatedly emphasized that the advancement of these models must go hand in hand with greater transparency and security mechanisms. It's no coincidence that the segment dedicated to SynthID and content identification came right after the presentation of the most spectacular capabilities of models like Gemini Omni for video and audio generation . For Google, simply flexing its technical muscle is no longer enough; it also needs to demonstrate its ability to manage the associated risks.
Android 17 and the shift towards the "intelligence system"
Although Android 17 had revealed much of its potential the previous week at The Android Show: I/O Edition, the operating system once again played a prominent role in the keynote. Google no longer describes it simply as a classic OS, but as a true "intelligent system" in which Gemini is embedded at every level: from the home screen to notifications, including voice dictation and content management.
One of the most talked-about features is the deep integration of Gemini Intelligence , the layer that automates tasks, summarizes web pages, fills out complex forms, transforms filler-filled dictations into clean text, and creates custom widgets from natural language instructions. Google has confirmed that these capabilities will first arrive on the latest Samsung Galaxy and Google Pixel models throughout the European summer, and will then be rolled out to other devices, provided they meet the hardware requirements.
Among the new features for everyday use are tools designed for social media creators and users: improvements to Instagram posting , new AI-assisted editing features for photos and videos, and the arrival of Adobe Premiere on Android for more professional workflows. Screen Reactions has also been announced, allowing simultaneous recording of the screen and the user's reaction, which can be useful for both content creators and educational settings.
However, the company admits that many of these advancements won't be available on all devices. The processing and memory requirements of the most ambitious features are high, which could exclude even some relatively recent devices. This is a sensitive issue in regions like Spain, where mobile phone replacement cycles tend to be longer than in other markets.
Googlebook, the bet to reinvent the laptop
Another focus of the event was Googlebooks , the new generation of laptops that Google positions as a middle ground between traditional Chromebooks and a PC with AI designed from the ground up. These devices run on a system still shrouded in mystery—known internally as Aluminium OS —that combines the best of Android and ChromeOS, and places Gemini at the heart of the experience.
Far from being a simple name change, Googlebooks are designed to integrate much more closely with the user's Android phone, sharing apps and files, and offering features like Magic Pointer , a cursor that, thanks to Gemini, suggests contextual actions based on what's being pointed at on the screen. The idea is that the laptop itself suggests shortcuts, summaries, or automations without the user having to explicitly request them each time.
The company has confirmed that manufacturers such as ASUS, Acer, Dell, HP, and Lenovo will participate in this category, opening the door to models with varying price points and features. While details on all configurations and specific dates for Europe haven't been released yet, Google has made it clear that I/O is just the beginning and that more information about hardware, availability, and the operating system's final name will be revealed throughout the year.
The initial impressions of Googlebook that had leaked in previous weeks were, at best, mixed. Therefore, the company took advantage of the Mountain View stage to showcase more concrete demos of real-world use cases : from syncing documents between mobile and laptop to quickly creating custom dashboards that combine information from Gmail, Calendar, and other desktop applications.
For the European laptop market, which has been relatively flat and dominated by Windows for years, Googlebook's offering is a way to compete with the so-called AI PCs being launched by other manufacturers. The difference, according to Google, lies in the fact that its offering relies on Gemini as a unifying layer and on smoother integration with the Android ecosystem, something that could be particularly appealing to those already using the company's services.
Android XR and smart glasses: from prototype to product
The field of extended reality has also been part of the conversation. Android XR , the platform that Google has been working on for some time with partners such as Samsung, Qualcomm, Gentle Monster, and Warby Parker, is back in the spotlight as the foundation upon which smart headsets and glasses with integrated Gemini will be built.
This year's I/O focused particularly on smart glasses, a format that had already been showcased in previous demos but has yet to reach the mass market. Google clarified that there will be two main types of glasses : those with an integrated screen, similar to those shown last year, and those without a screen, focused on audio and with Gemini acting as a constant-present audio assistant. These latter glasses, it was announced, will be available this fall, at least in some markets.
The promise is of glasses capable of "seeing" the same as the user and offering contextual information in real time : simultaneous translation of conversations, navigation directions, help to identify objects or places, sending messages without taking the phone out of the pocket... All this with AI processing the environment and returning responses in a more proactive way.
Meanwhile, Samsung took advantage of the event, albeit indirectly, to offer a first glimpse of its own Android XR glasses , developed in collaboration with Gentle Monster and Warby Parker. For now, it's more of a statement of intent than an immediate launch, but it demonstrates that the partner ecosystem surrounding Android XR is solidifying and that we'll see a variety of designs and approaches.
The big question for Europe and Spain remains the specific availability. Google usually rolls out these types of devices first in the United States and some Asian markets, so we'll have to see when the Android XR glasses arrive in Europe and which features will be activated from day one, also taking into account European regulatory requirements regarding privacy and the handling of data captured by cameras and microphones.
Search, shopping, and productivity: AI is everywhere
The search engine, still the company's biggest business, has also received its share of AI. Google has announced that AI Overviews in Search , those AI-generated summaries that accompany some searches, will now be powered by Gemini 3.5. The goal is to offer more accurate and contextual answers, while maintaining control mechanisms to prevent major errors.
This is the foundation of AI Mode , the AI-assisted search service that has received interface and functionality updates. The key innovation is the integration of agentic AI into this search mode , allowing the search engine itself to generate personalized interactive tools: animated charts, calendars, maps, or small tracking panels created on the fly based on user needs. This functionality, which will be available globally for free starting this summer, is poised to become one of the most visible applications of AI for those who simply want to "search on Google."
Another striking feature is the new universal shopping cart , integrated directly into the search engine. Instead of having lists scattered across different stores, users can keep a centralized record of the products they're interested in, with prices, availability, and recommendations. For European e-commerce, especially for small businesses that rely on search visibility, this type of functionality can change how users compare and make purchasing decisions.
In the area of productivity, one of the most talked-about announcements has been Docs Live , a feature integrated into Gemini that allows users to generate complete documents from a brief voice description and subsequently edit them using their voice. It will arrive this summer for users of the Google AI Pro and Ultra plans, and fits into the trend of blurring the lines between dictation, assistants, and traditional text editing.
Google also showcased new features for Google Pics , its AI-powered image manipulation tool, and highlighted Stitch, a platform capable of generating complete application and web page interfaces from prompts. Stitch relies on Antigravity, the company's agentive development infrastructure, and an extreme example was demonstrated: engineers instructing Gemini to create an operating system from scratch, something that—according to Google—took 12 hours and cost less than $1.000 in API credits.
Antigravity, developer tools and the role of Europe
Beyond the keynote address, the Google I/O 2026 program was packed with technical sessions focused on Google AI Studio , Antigravity, Gemma, TPU, Flutter, Web UI, Chrome DevTools for agents, and quantum computing applied to science. For the European developer community, which followed the event online at io.google/2026, the main interest was how these tools can lower the barrier to entry for complex AI projects.
Antigravity, in particular, is positioned as a unified platform for AI-powered software engineering , aiming to attract both professionals and enthusiasts of the "coding vibe." The idea is to centralize tools for code generation, testing, deployment, and automation, all supported by models like Flash and Pro, depending on the needs of each project. With the integration of Google AI Studio—which is launching a new icon and mobile app currently in pre-registration—the goal is to enable the design, testing, and deployment of agents and interfaces almost entirely within the Google ecosystem.
In Europe, where many tech SMEs operate on tight budgets, the emphasis on efficient models and simplified development cycles is particularly relevant. Google knows that much of its AI business growth in the region will come from companies that can't afford to build their own computing infrastructure but can leverage APIs and managed tools if the costs are predictable and reasonable.
In this sense, the message that has permeated the entire Google I/O 2026 is that a new phase in the AI cycle is beginning. After a period dominated by fascination with the capabilities of the models, the debate is now shifting towards cost, efficiency, and real-world use cases . Google has attempted to position itself at the center of this conversation with a proposal that blends powerful models, agentic tools, and a very specific narrative about cost savings, while also addressing regulatory pressure and expectations of transparency, which are particularly strong in the European context.
What this year's Google I/O leaves us with is the image of a company that wants to demonstrate that Gemini is not just a flashy chatbot, but the layer that connects Android, Chrome, Search, Workspace, laptops, connected cars, and future smart glasses. It remains to be seen to what extent users and businesses in Spain and the rest of Europe will embrace this vision, but the message Google has sent from Mountain View is clear: the next wave of innovation will not be measured solely by spectacular demos, but by the actual cost of putting artificial intelligence to work every day.