Google I/O 2026 Keynote: Google Declares the Start of Its Agentic Gemini Era

Google I/O 2026 opened with a clear message: the company no longer sees artificial intelligence as a feature layered onto its products, but as the operating logic behind its next generation of services. At the Shoreline Amphitheatre in Mountain View, Google and Alphabet CEO Sundar Pichai framed the past year as a period of “hyper progress” for AI, using the keynote to introduce new Gemini models, agentic Search, Workspace tools, AI-powered shopping, Android XR smart glasses, and expanded content authenticity systems.

The keynote was not built around one single product reveal. Instead, it showed Google attempting to connect its entire ecosystem — Search, Gemini, Android, Workspace, YouTube, Cloud, Chrome, developer tools and hardware — around a more proactive form of AI. Google’s phrase for that shift is the “agentic Gemini era,” a future in which AI does not only answer questions, but performs tasks, monitors information, builds interfaces, creates media and works across apps under user direction.

A Keynote Built Around Scale

Pichai began by emphasizing the scale of Google’s AI usage. The company said it is now processing more than 3.2 quadrillion tokens every month, up sevenfold from the previous year. Two years earlier, Google was processing 9.7 trillion tokens a month across its surfaces; last year, that figure had grown to roughly 480 trillion.

That growth was presented as evidence that AI is moving from experimentation into daily use. Google said more than 8.5 million developers now build with its models monthly, while its model APIs process roughly 19 billion tokens per minute. The company also said more than 375 Google Cloud customers each processed more than one trillion tokens over the past 12 months.

The numbers were designed to support a broader argument: Google believes its AI advantage comes from owning the full stack, from custom silicon and infrastructure to foundation models and products with billions of users. Pichai said Google has 13 products with more than a billion users each, five of which have more than three billion users.

Gemini’s User Base Becomes a Major Talking Point

The Gemini app was one of the strongest examples of Google’s momentum. The company said the app has crossed 900 million monthly active users, more than doubling from 400 million monthly active users at I/O the previous year. Daily requests in the Gemini app have grown more than seven times over the same period.

Google also said more than 50 billion images have been generated using its Nano Banana image generation models, a figure that signals how deeply creative AI tools are being folded into mainstream product use.

The keynote positioned Gemini not just as a chatbot, but as the connective intelligence across Google’s consumer, enterprise and developer platforms. That strategy appeared repeatedly in the announcements: Gemini in Search, Gemini in Workspace, Gemini in Chrome, Gemini in Android XR, Gemini in coding tools and Gemini as a personal agent.

Gemini 3.5 Flash Arrives, With Gemini 3.5 Pro Coming Next

One of the biggest model announcements was Gemini 3.5 Flash, which Google described as its latest model combining frontier intelligence with action. The model is available across Google’s products and APIs, including the Gemini app, Search, Google Antigravity and the Gemini API.

Google said Gemini 3.5 Flash is four times faster than comparable frontier models when measured by output tokens per second, while delivering frontier-level capabilities at less than half the price of comparable frontier models. Pichai argued that this cost-performance profile could matter especially for companies already consuming large AI token budgets.

The more powerful Gemini 3.5 Pro did not launch during the keynote. Google said it is being used internally and is expected to arrive next month. That made Flash the immediate centerpiece of Google’s model strategy at I/O 2026, while Pro became the next milestone to watch.

Gemini Omni Pushes Google Further Into Multimodal AI

Google also introduced Gemini Omni, a new AI model focused initially on video generation, with broader multimodal ambitions. Google described Omni as capable of generating samples in any output modality from any input, beginning with video outputs and later expanding to image and text.

The first model in the family, Gemini Omni Flash, is available in the Gemini app, Google Flow and YouTube Shorts, with API access for developers and enterprise customers expected in the coming weeks. During the keynote, Google demonstrated realistic physics simulations, advanced video editing and AI-generated modifications inside existing video clips.

The phrase attached to Gemini Omni — “any output from any input” — captures the direction Google is pursuing: AI systems that are not restricted to text prompts or text responses, but that can work across video, images, text and audio in a single creative workflow.

Gemini Spark Turns the Assistant Into a 24/7 Agent

If Gemini 3.5 Flash was the model story, Gemini Spark was the agent story. Google introduced Spark as a personal AI agent inside the Gemini app, designed to work in the background and act on a user’s behalf under their direction.

Spark runs on dedicated Google Cloud virtual machines, meaning it can continue long-running tasks without requiring a user’s laptop or phone to remain active. It is powered by Gemini 3.5 and the Google Antigravity harness, and it will initially integrate with Google’s own tools before adding third-party support through MCP in the coming weeks.

Google said Spark is rolling out to trusted testers first, with a beta coming to Google AI Ultra subscribers in the U.S. next week. The company also said users will eventually be able to interact with Spark through the Gemini app, email and chat.

The significance of Spark is that it shifts Gemini from a reactive assistant toward a persistent AI worker. Instead of asking a question and waiting for an answer, users could assign longer tasks and receive progress updates while the agent continues operating in the background.

Search Becomes More Agentic

Google Search also received one of its most consequential updates. Google announced information agents in Search, described as personalized AI agents that can operate in the background, find information at the right moment and help users take action. These agents are rolling out this summer, beginning with Google AI Pro and Ultra subscribers.

The company also said Search will gain agentic coding capabilities powered by Gemini 3.5 Flash and Google Antigravity. This will allow Search to generate custom experiences for individual questions, including dynamic layouts and interactive visuals. Google said those generative UI capabilities will be available for everyone in Search this summer at no charge.

For longer-running tasks, Search will also be able to build persistent dashboards or trackers — described as mini apps for specific tasks — that users can return to over time. These custom experiences will arrive in the coming months, starting with Google AI Pro and Ultra subscribers in the U.S.

This marks a major change in Google’s original search model. Search is no longer being presented only as a box that returns links, but as a workspace that can monitor, generate, summarize, visualize and act.

Workspace Gets More Voice and Creative AI

The keynote also expanded Google’s AI push into productivity. Docs Live allows users to create documents by speaking to Gemini Live. Google said the feature can handle natural speech patterns, pauses, corrections and incomplete thoughts while turning them into structured documents.

Docs Live is expected to roll out this summer for subscribers, with related voice capabilities also coming to Gmail and Keep.

Another Workspace-related product, Google Pics, is a new AI image creation and editing tool built on the Nano Banana model. Unlike basic image editors, Google said Pics treats every element of an image as an individual object rather than a flat static image, allowing users to create, swap or refine specific details. It is available to trusted testers and will roll out later this summer to Google AI Pro and Ultra subscribers in Workspace.

These tools show Google moving AI deeper into everyday productivity: writing, editing, designing, summarizing and planning are increasingly being framed as conversational workflows.

YouTube, Shopping and Daily Brief Get AI Upgrades

Google also introduced Ask YouTube, a conversational search feature designed to help users locate relevant moments inside YouTube videos without knowing the title of a video. Google said it is beginning to test Ask YouTube and will roll it out broadly in the U.S. this summer.

For commerce, Google announced Universal Cart, a shopping hub that works across Search, Gemini, YouTube and Gmail. It includes AI-powered price tracking, smart recommendations and simplified checkout features through a unified interface.

Another new feature, Daily Brief, is coming to the Gemini app as an out-of-the-box agent that synthesizes information from a user’s inbox, calendar and tasks. Google said it will not simply summarize information, but prioritize, organize and suggest next steps in a concise morning digest.

Together, these announcements suggest Google wants Gemini to become a layer that sits above fragmented digital activity — videos, shopping, messages, calendars, tasks and documents — and turns that activity into actionable context.

Android XR and Smart Glasses Return to the Spotlight

Hardware also had a major place in the keynote. Google confirmed two categories of Android XR smart glasses: audio glasses, which offer spoken assistance, and display glasses, which show information directly in the user’s line of sight. Audio glasses are expected to launch first, arriving this fall.

Google also announced partnerships around Android XR, including work with Samsung. The first audio glasses will support both Android and iOS devices.

The company also introduced Android Halo, a new UI space that will show live updates and task progress from agents like Gemini Spark. Google said Android Halo is coming later this year.

The smart glasses announcement is notable because it brings Google back into a category it helped popularize years ago, but this time with Gemini as the central interface. The pitch is less about a standalone gadget and more about ambient AI: directions, messaging, photos and hands-free help without constantly looking at a phone.

Antigravity 2.0 Targets Developers and AI Agents

For developers, Google announced Antigravity 2.0, an upgrade to its AI coding platform. The company said Antigravity is expanding beyond a coding environment into a platform for developing and managing groups of autonomous AI agents.

The keynote described a future where developers can hand over complex coding tasks to Gemini, with multiple AI sub-agents simultaneously handling coding, testing and execution. Google said an optimized version of Flash inside Antigravity is not just four times faster, but 12 times faster than other frontier models.

This aligns with the keynote’s larger agentic theme: Google is not only building agents for consumers, but also giving developers tools to create, supervise and deploy agents of their own.

TPU 8 and the Infrastructure Behind the AI Push

Google’s AI ambitions depend heavily on infrastructure, and Pichai spent part of the keynote discussing investment and custom silicon. Google said it expects capital expenditure this year to reach approximately $180 billion to $190 billion, about six times its annual capex in 2022.

The company highlighted its eighth-generation tensor processing units, including TPU 8t for training and TPU 8i for inference. Google said TPU 8t offers nearly three times the raw computing power of the previous generation, while both chips deliver up to two times better performance-per-watt.

Google also said it can now scale training across more than 1 million TPUs globally, using JAX and Pathways to distribute training across multiple sites instead of relying on a single massive data center.

That infrastructure message matters because Google’s consumer AI announcements require huge compute capacity. Agents running 24/7, multimodal video models, AI Search, Workspace generation and developer tools all increase demand for fast, efficient inference and training.

SynthID Expands as AI Content Concerns Grow

Google also used the keynote to address AI transparency. The company said SynthID has now watermarked more than 100 billion images and videos, along with 60,000 years of audio assets.

Google is expanding SynthID and Content Credentials verification to Search and Chrome, making it easier for users to identify whether content originated from AI, a camera or generative editing tools.

The company also announced new SynthID adoption by OpenAI, Kakao and Eleven Labs, adding to previous support from Nvidia.

This was one of the most socially significant parts of the keynote. As generative media becomes more realistic, detection and provenance tools are becoming essential to public trust, journalism, elections, education and online safety.

What Google I/O 2026 Means for the AI Industry

The Google I/O 2026 keynote showed a company trying to move faster than the traditional software cycle. Instead of shipping isolated AI features, Google is building a multi-layered AI system: models, chips, agents, developer platforms, consumer apps, browser tools, productivity products and wearables.

The biggest strategic shift is the rise of agents. Gemini Spark, information agents in Search, Daily Brief, Flow agents and Antigravity 2.0 all point in the same direction: Google wants AI to become more autonomous, persistent and embedded in workflows.

For users, that could mean less time switching between apps and more help completing tasks. For developers, it could mean new ways to build software with cohorts of AI agents. For businesses, it could mean lower AI operating costs if models like Gemini 3.5 Flash deliver the speed and pricing Google claims. For society, it raises new questions about trust, transparency, user control and the boundaries of automated decision-making.

Conclusion: Google’s AI Future Is No Longer Theoretical

Google I/O 2026 was not simply a showcase of new AI tools. It was a declaration that Google’s next major platform shift is already underway. Search is becoming agentic. Gemini is becoming more proactive. Workspace is becoming conversational. Android is preparing for live agent visibility. Smart glasses are returning with AI at the center. Developers are being invited to build with and manage agents rather than just write code manually.

The keynote’s most important message was that Google sees AI as a full-stack transformation of computing. Whether users experience that through Gemini Spark completing background tasks, Search generating custom dashboards, Docs Live turning speech into structured documents or smart glasses offering hands-free assistance, the company’s direction is clear: Google wants Gemini to become the intelligence layer across its entire ecosystem.