At Google I/O 2026, CEO Sundar Pichai’s keynote promoted the company’s next phase around enterprise adoption and the agentic systems needed to run AI at massive scale. His pitch in a nutshell: Google wants Gemini to become a daily operating layer across development, search, productivity and consumer services.
An ambitious goal, and the company laid out impressive stats to support it. Google said its AI systems now process more than 3.2 quadrillion tokens per month across its products, up sharply from last year, and more than 8.5 million developers use Google models monthly.
Also remarkable: Google Cloud now has more than 375 customers that each processed over one trillion tokens in the past year.
Among the most significant enterprise news was Google’s solution for an agent control plane, said Mitch Ashley, VP, Software Lifecycle Engineering at The Futurum Group.
“The AI stack consolidation, from custom silicon through Antigravity to Spark, positions Antigravity as a vertically integrated agent control plane. The competitive question in agentic AI has moved from model quality to who governs agent execution across the stack,” he said.
“Operations teams are already managing agents proliferating vertically and horizontally across the stack. Without a unifying control plane, those environments fragment into disconnected agent populations with inconsistent identity, telemetry, and policy enforcement, and the operational burden compounds with every new agent deployed.”
AI and Agentic AI
The star of the show was Gemini 3.5 Flash, a new AI model Google describes as built for both reasoning and action. Pichai said it outperforms Gemini 3.1 Pro on most benchmarks while running faster than rival frontier models. Google touts Flash as a cost-control tool for enterprises facing rapidly growing token bills.
Gemini 3.5 Flash is available now across its products and APIs. Gemini 3.5 Pro is being tested internally and is expected next month.
Gemini has also expanded into content creation with Gemini Omni, a model designed to generate output from multiple input types. The first release, Gemini Omni Flash, focuses on video generation and editing through conversational instructions. Google said the model will be available in the Gemini app, Flow and YouTube Shorts, with enterprise and developer API access coming later.
For developers, Google introduced Antigravity 2.0, a desktop application for managing teams of AI agents. As Ashley noted, the platform moves beyond a coding assistant and toward an environment where users can direct autonomous agents across longer tasks. Google claims that an optimized version of Flash in Antigravity runs far faster than other frontier models in that setting.
Gemini Spark, Google’s new personal AI agent, will run on Google Cloud virtual machines and continue working even when a user’s device is closed. It will connect first to Google tools, then to third-party applications through MCP support.
Spark will begin with testers, then move to Google AI Ultra subscribers in the US. It is also slated to work inside Chrome later this summer.
Search
As ChatGPT gained traction, doubts emerged about the long-term dominance of Google Search. But Google is rebuilding the product for the AI era. Search is no longer framed only as a query engine, but as a system that can synthesize and act.
Search is being rebuilt around the agentic model: Google announced agents that can monitor topics and alert users. Search will also gain generative interface features that can build dashboards and trackers layouts in response to complex questions.
Content Verification
Addressing generative AI’s ability to create hyper-realistic content, Pichai devoted part of the keynote to authentication and content verification tools. The company expanded its SynthID initiative, a watermarking system designed to identify AI-generated media without visibly altering the content itself.
Google said SynthID has been used to watermark more than 100 billion AI-generated images and videos, along with tens of thousands of years of audio content.
Content Credentials verification and SynthID detection are being integrated directly into Search and Chrome. Users will be able to determine whether content originated from a camera, was generated entirely by AI or was modified using generative tools.
Consumer Market
The company also unveiled several consumer-facing AI features, all of which support AI that is embedded and conversational:
Ask YouTube introduces AI-powered video search that surfaces the most relevant sections of videos instead of requiring users to manually scrub through content.
Docs Live enables users to create and edit Google Docs through spoken prompts, turning voice interaction into a primary workflow interface.
Daily Brief uses Gemini agents to generate personalized summaries from Gmail, Calendar and task data, with suggested actions and prioritization.
For Apple users, the Gemini Mac app will let users issue voice commands tied to files and folders directly from Finder, including generating emails and organizing documents.
