Google Unveils Agentic Gemini Era at I/O 2026

Google used its annual I/O developer event in Mountain View to pivot its product strategy toward "agentic" artificial intelligence, aiming to move beyond simple chatbots to systems that actively perform tasks for users [1]. CEO Sundar Pichai announced that the company is entering the "agentic Gemini era," supported by a massive increase in capital spending projected to reach $180 billion to $190 billion this year [1, 2].

Key takeaways

• Google introduced Gemini 3.5 Flash, a new model that the company claims is four times faster than other frontier models [1, 2].
• A new personal agent called Gemini Spark is designed to operate 24/7 on cloud machines to complete tasks on a user's behalf [1].
• Search is undergoing its most significant change in 25 years, incorporating generative UI and agents that monitor topics across the web [1].
• Google is expanding its infrastructure, including the deployment of eighth-generation TPU 8t and TPU 8i chips [1, 2].
• The Gemini app has reached over 900 million monthly users, with daily requests growing sevenfold over the past year [1].

Advancements in Models and Developer Tools

The centerpiece of the event was the launch of the Gemini 3.5 model family, led by Gemini 3.5 Flash [1]. Google reports that this model outperforms its predecessor, Gemini 3.1 Pro, on coding and agentic benchmarks, while maintaining high speeds for inference [1, 2]. Additionally, the company introduced Gemini Omni, a model capable of generating output from various inputs, such as video, with an improved understanding of physics [1]. These models are supported by the updated Antigravity 2.0 platform, which allows developers to orchestrate multiple agents in parallel [1].

Google is also integrating these capabilities directly into its consumer products. The Gemini app is transitioning into an agent-based experience with the introduction of Gemini Spark and the "Daily Brief" feature, which synthesizes inbox and calendar data [1]. Furthermore, Google is rolling out "Universal Cart," a shopping tool that tracks price history and flags product incompatibilities, across its ecosystem [1].

The Shift Toward Agentic Infrastructure

The company’s focus on agents extends to its core Search product, which now allows users to query across text, images, files, and browser tabs simultaneously [1]. By this summer, Google plans to release information agents that monitor web topics and provide synthesized updates, alongside a generative UI that creates custom search layouts [1].

These innovations are backed by significant hardware and cloud investments. Pichai noted that capital expenditures have grown substantially from $31 billion in 2022 to support the development of custom silicon and the processing of 19 billion tokens per minute across model APIs [1, 2]. While these tools represent a major shift in functionality, many features remain in beta or are restricted to specific subscriber tiers, leaving the company to prove that these agents can operate reliably without becoming intrusive to the user experience [1].