Google I/O 2026: Sundar Pichai Unveils Agentic Gemini Era with New Models, Infrastructure, and AI Agents

Google CEO Sundar Pichai opened the 2026 I/O conference with a clear message: we've entered the "agentic Gemini era." In a keynote that highlighted both technological advancements and practical applications, Pichai outlined how Google's full-stack AI approach is transforming products, infrastructure, and developer experiences to create more intelligent, capable AI agents that can take action on behalf of users.

The Evolution to an Agentic Future

"It's been an extraordinary year since our last I/O, a period of relentless shipping, technology advances and hyper progress," Pichai stated. "We're now in the part of the AI cycle where people want to see the value in the products they use every day."

This focus on tangible value comes as Google marks a decade since pivoting to an "AI-first" approach. The company's commitment to this vision is reflected in its infrastructure investments, with capital expenditure projected to reach $180-190 billion in 2026—six times the $31 billion spent in 2022.

The scale of AI adoption is staggering. Two years ago, Google processed 9.7 trillion tokens monthly; today, that number has jumped to over 3.2 quadrillion tokens monthly. This growth is supported by 8.5 million developers building with Google's models monthly and APIs processing roughly 19 billion tokens per minute.

Infrastructure Advancements: TPU 8t and 8i

Supporting this massive scale requires significant infrastructure investment. Google announced its eighth-generation Tensor Processing Units (TPUs), marking a fundamental shift with a dual-chip approach:

• TPU 8t: Optimized for large-scale pretraining, delivering nearly three times the raw computing power of the previous generation
• TPU 8i: Designed specifically for inference, dramatically improving speed at every step

Both chips offer up to two times better performance-per-watt, addressing efficiency concerns as AI workloads grow.

"With JAX and Pathways, our training is no longer constrained by the limits of a single, massive data center," Pichai explained. "Instead, we can now seamlessly distribute training across multiple sites, scaling training across more than 1 million TPUs globally. This gives us the ability to create the largest training cluster in the world."

This infrastructure enables training larger, more capable models in weeks rather than months—a critical advantage in the rapidly evolving AI landscape.

Model Innovations: Gemini Omni and 3.5 Flash

Google introduced several model advancements, headlined by:

Gemini Omni

Gemini Omni represents a significant leap forward in world understanding, capable of generating samples in any output modality from any input. Starting with video outputs and expanding to image and text over time, this model combines Gemini's intelligence with Google's generative media models.

The first model in the Omni family, Gemini Omni Flash, is available immediately through the Gemini app, Google Flow, and YouTube Shorts, with broader API access for developers and enterprise customers coming in the weeks ahead.

Gemini 3.5 Flash

Perhaps the most significant model announcement is Gemini 3.5 Flash, which Pichai described as "our first in a series of models combining frontier intelligence with action."

Compared to the previous 3.1 Pro model, 3.5 Flash shows improvements across almost all benchmarks, with extraordinary progress in coding and real-world economically valuable tasks (as measured by GDPVal). What sets 3.5 Flash apart is its combination of frontier-level capabilities with exceptional speed—four times faster than other frontier models in output tokens per second.

"When you look at the intelligence versus output speed, it's in a league of its own in the top right quadrant," Pichai emphasized.

The model has already transformed Google's internal development processes. "In March we were processing half a trillion tokens a day internally across our AI developer tools, and we've been doubling every few weeks. Now, we're processing more than three trillion tokens a day," Pichai shared.

Perhaps most compelling for businesses is the cost advantage. "Gemini 3.5 Flash delivers frontier-level capabilities at less than half the price of comparable frontier models," Pichai noted. "If companies used a mix of Flash and other frontier models they could save a lot of money. If top companies processing about 1 trillion tokens a day shifted 80% of their workloads from other frontier models to 3.5 Flash, they'd save over $1 billion dollars annually."

Antigravity 2.0 and the Agent Platform

To support the development of AI agents, Google announced Antigravity 2.0, an evolution of its agent-first development platform that expands beyond coding to become a comprehensive platform for developing and managing cohorts of autonomous AI agents.

The new version includes:

• A standalone desktop application serving as a central hub for agent interaction
• Enhanced orchestration capabilities for diverse agent tasks
• An optimized version of Gemini 3.5 Flash that's 12x faster than other frontier models

"Antigravity is expanding beyond the coding environment, turning it into a platform to develop and manage cohorts of autonomous AI agents," Pichai explained. "This includes Antigravity 2.0, a new standalone desktop application that acts as a central home for agent interaction, where anyone can orchestrate agents for all sorts of tasks."

Gemini Spark: The Personal AI Agent

The most consumer-facing announcement is Gemini Spark, a personal AI agent within the Gemini app designed to navigate users' digital lives, taking action on their behalf and under their direction.

Key features include:

• Operation on dedicated virtual machines in Google Cloud
• 24/7 availability without requiring users to keep their laptops open
• Integration with Gemini 3.5 and the Google Antigravity harness for long-horizon tasks
• Seamless integration with tools, starting with Google's own and expanding to third-party tools through MCP
• Multiple interaction methods: through the Gemini app, email, and chat
• Android UI integration through a new space called "Android Halo"
• Direct operation within Chrome as an "agentic browser"

"Gemini Spark is the first experience made possible by 3.5 models and Antigravity," Pichai stated. "This combination gives us new ways to accelerate our mission and transform our products to be radically more helpful."

Gemini Spark is beginning rollout to trusted testers this week, with Beta access for Google AI Ultra subscribers in the U.S. starting next week.

Search Transformation in the Agentic Era

Search, Google's flagship product, is undergoing significant transformation to incorporate agentic capabilities:

• Information Agents: Personalized AI agents that work in the background to find information and help users take action at the right moment
• Agentic Coding Capabilities: Building custom experiences with dynamic layouts and interactive visuals
• Persistent Dashboards: Custom trackers for longer-running tasks that users can return to and make progress on

These enhancements will roll out starting this summer, with information agents initially available to Google AI Pro and Ultra subscribers, while generative UI capabilities will be free for all users.

Transparency and Content Authentication

As AI-generated content becomes more prevalent, Google announced expanded efforts to ensure transparency:

• SynthID: The invisible watermark has now marked over 100 billion images and videos, along with 60,000 years of audio assets
• Content Credentials Verification: Expanding across products to show content origin and editing history
• New Partners: OpenAI, Kakao, and Eleven Labs joining Nvidia in adopting SynthID

"Research shows people can correctly identify high-quality deepfake videos only about a quarter of the time," Pichai noted. "Three years ago, we launched SynthID, our watermark that is invisible to the naked eye. Since launch, SynthID has now watermarked over one hundred billion images and videos."

Additional Product Announcements

Beyond the core AI and agent announcements, Google revealed several new products and features:

• Ask YouTube: A new feature that reimagines the YouTube experience, making information more digestible and jumping directly to relevant video sections
• Docs Live: Voice-powered document creation that allows users to verbally "brain dump" ideas
• Google Pics: AI image creation and editing tool that treats elements as individual objects rather than flat images
• Daily Brief: An agent for the Gemini app that synthesizes information from inbox, calendar, and tasks
• Google Flow Updates: New agent capabilities for planning and reasoning through complex tasks
• Intelligent Eyewear: Audio glasses offering spoken help and display glasses showing information hands-free
• Gemini for Science: Collection of AI tools to accelerate scientific research

Market Position and Competitive Landscape

Google's announcements come at a critical juncture in the AI market, where companies are differentiating through both model capabilities and practical applications. While Google's Gemini models compete with offerings from OpenAI, Anthropic, and others, the company is emphasizing its full-stack approach as a key differentiator.

"We've been taking a differentiated, full-stack approach to AI innovation, from our custom silicon and secure foundation, to our world-class research and models, to our products and platforms that touch billions of people," Pichai stated. "This approach enables us to iterate and innovate faster in ways that are lighting up every part of the company."

This integrated approach allows Google to offer solutions that span infrastructure, models, and applications—a comprehensive offering that appeals to both enterprises and developers.

Business Impact and Strategic Direction

The announcements at I/O 2026 signal Google's strategic focus on making AI more actionable through agents. Rather than simply providing smarter conversational interfaces, Google is building systems that can take action on behalf of users across its product ecosystem.

For businesses, the cost-performance advantages of models like Gemini 3.5 Flash offer significant operational efficiencies, potentially reducing AI infrastructure costs by billions annually for large organizations. The Antigravity platform provides tools for developing custom agents tailored to specific business processes, potentially unlocking new levels of automation and productivity.

For consumers, the shift toward agents like Gemini Spark represents a fundamental change in how we interact with technology—moving from query-response to ongoing assistance that works proactively in the background.

As Pichai concluded, "As we look across the full stack of innovation, from the infrastructure behind TPU 8i to the frontier capabilities of Gemini 3.5 and Antigravity, it's clear we're firmly in our agentic Gemini era. I'm excited to see how it will unlock new ways to accelerate our mission and transform our products to be radically more helpful, for everyone everywhere."

The announcements at Google I/O 2026 demonstrate not just technological advancement but a clear vision for how AI will evolve from conversational interfaces to proactive agents capable of taking meaningful action on behalf of users—a vision that could reshape the competitive landscape in cloud computing and AI services for years to come.