Google I/O 2026: All the Major AI Announcements

At this year’s Google I/O developer conference, CEO Sundar Pichai laid out what he called the company’s “agentic Gemini era,” unveiling a set of AI products, upgrades, and infrastructure plans designed to make Gemini more active across Search, Android, Workspace, and the web.

Pichai said Google is now focused on showing people “the value in the products they use every day,” after years of rapid advances in artificial intelligence.

Gemini usage surges across Google products

Google used the keynote to highlight how quickly its AI ecosystem has grown over the past year. According to the company, monthly token processing across its products has climbed to more than 3.2 quadrillion per month, up sevenfold year over year. 

Google also said more than 8.5 million developers are now building with Gemini models every month, while its APIs handle around 19 billion tokens per minute.

Pichai pointed to growing adoption across Google’s consumer products as evidence that AI tools are becoming mainstream.

“AI Overviews now has over 2.5 billion monthly active users,” he said. He also described AI Mode in Search as the company’s “biggest upgrade to Search ever.”

Google said the Google Gemini app has now crossed 900 million monthly active users, more than doubling from last year’s figure of 400 million. The company also revealed that users have generated more than 50 billion images using its Nano Banana image generation models.

Search becomes more conversational and more agentic

A major theme throughout Pichai’s keynote was Google’s effort to transform Search from a traditional search engine into a more interactive AI assistant. Pichai said Search now feels “more like an ongoing conversation,” with users asking longer and more complex questions.

Google announced several new AI-powered experiences tied to that vision. One of them, Ask YouTube, is designed to help users navigate videos more efficiently by surfacing relevant clips and jumping directly to the most useful sections. Google said the feature is currently being tested and will launch more broadly in the US this summer.

The company also introduced information agents in Search, AI systems that users can set up to continuously monitor topics and return useful updates or actions in the background.

“These are personalized AI agents you can set up to work in the background, 24/7, to find what you need at exactly the right moment, and help you take action,” Pichai said. Google added that Search will soon generate dynamic layouts, interactive visuals, and persistent dashboards for long-running tasks.

Gemini Spark becomes Google’s personal AI agent

Among the announcements was Gemini Spark, a new persistent AI agent inside the Gemini app.

Google described Spark as a system that can run continuously on dedicated virtual machines within Google Cloud infrastructure. The agent is designed to complete long-running tasks, integrate with tools, and eventually operate directly inside the Chrome browser.

“It’s 24/7 so you don’t need to keep your laptop open,” Pichai said. Google said Spark will first roll out to trusted testers before expanding to Google AI Ultra subscribers in the US. The company also announced Android Halo, a new Android interface that will display live updates from AI agents running in the background.

Gemini 3.5 Flash targets speed and lower costs

On the model side, Google introduced Gemini 3.5 Flash, which the company described as a faster and more affordable frontier AI model focused heavily on coding and agentic workflows.

Pichai said the model performs better than Gemini 3.1 Pro across most benchmarks while remaining significantly faster than competing frontier systems.

“When looking at output tokens per second, it is four times faster than other frontier models,” he said. Google also emphasized pricing, arguing that businesses using Gemini 3.5 Flash could sharply reduce AI infrastructure costs.

“What’s amazing about Flash is how it delivers frontier-level capabilities at less than half the price of comparable frontier models,” Pichai said. The model is now available across Google products and APIs, while Gemini 3.5 Pro is expected next month.

New AI tools arrive for Workspace and creators

Google used the event to unveil a series of new AI-powered productivity and creative tools. Docs Live adds voice-based document creation inside Google Docs, allowing users to verbally “brain dump” ideas instead of typing detailed prompts. The feature is expected to launch for subscribers later this summer.

The company also introduced Google Pics, a new AI image creation and editing platform powered by its Nano Banana models. Google said the system treats individual image elements as editable objects rather than static, flat images.

Meanwhile, Google Flow is receiving new agentic features for brainstorming, editing, and AI-assisted creative workflows.

Google expands AI transparency efforts

Google also focused heavily on AI safety and content authenticity. The company said its SynthID watermarking system has now marked more than 100 billion AI-generated images and videos, as well as tens of thousands of years of audio assets.

Pichai announced that OpenAI, Kakao, and ElevenLabs are adopting SynthID technology alongside existing partner Nvidia. Google added that Content Credentials verification tools will expand into Search and Chrome to help users identify AI-generated or AI-edited media.

Massive infrastructure spending powers Google’s AI push

Behind the scenes, Google signaled that its AI ambitions are coming with a huge financial commitment.

Pichai said the company expects annual capital expenditures to reach between $180 billion and $190 billion this year, compared to $31 billion in 2022. The company also unveiled its eighth-generation TPU chips, including TPU 8t for training and TPU 8i for inference workloads.

Google said its training systems can now scale across more than one million TPUs globally.

Also read: Google is adding Gemini Intelligence to Android, bringing AI features for screen context, Chrome, Autofill, Gboard, and widgets.