Google I/O 2026: 100 Bets on One Stack

Google made 100 discrete product announcements at I/O 2026, a volume that is less a product event and more a territorial claim. The centrepiece is Gemini Omni, a unified multimodal architecture capable of generating text, image, audio, video, and code from a single model. That is not an incremental upgrade. It is a direct assault on the fragmented multi-model workflows that most AI-native startups have built their products around. For founders and investors watching from the outside, the event functions as the clearest map yet of where Google intends to own the stack.

The context matters. For the past two years, the prevailing startup strategy in AI products has been to stitch together best-in-class models: one for text, one for image generation, one for video, one for code. That complexity created real moats, or so it seemed. Gemini Omni collapses that logic. With a single unified architecture handling every content type, Google is betting that integration beats specialisation, and that developers will take the path of least resistance when the quality ceiling is high enough. Gemini 3.5 Flash, also announced at I/O, signals that speed and cost efficiency are part of the same push.

The product surface is enormous. Gemini Omni Flash is now live inside Google Flow, the AI creative studio launched at last year's I/O, which has since expanded to over 140 countries. The Gemini app itself received a full redesign under the Neural Expressive system, featuring fluid animations, new typography, and haptic feedback, with Gemini Live launching inline. A new agent called Daily Brief pulls from your inbox, calendar, and tasks to surface the most actionable information each morning. These are not experiments. They are consumer-scale deployments happening right now.

For investors, the 100-announcement breadth forces an uncomfortable question: which portfolio companies just had their roadmap neutralised? Standalone AI video tools, multimodal API wrappers, and productivity agents built on top of third-party models are the most exposed categories. The more urgent question is where greenfield opportunity actually opens up. Gemini Omni's API surface, and the agentic infrastructure Google is building around Search and Workspace, creates integration layers that startups can build on rather than compete with. The companies that move fast to sit on top of these primitives, rather than underneath them, are the ones worth watching.

Over the next twelve months, the real test is execution at scale. Google has a long history of impressive I/O demos that diffuse slowly into actual user behaviour. But the deployment of Gemini Omni Flash into Google Flow across 140 countries, alongside aggressive consumer-facing redesigns, suggests this cycle is different. Expect a shakeout in the mid-tier AI tools market by mid-2027, as products that cannot meaningfully differentiate from what Google now ships for free come under serious pressure. The startups that survive will be the ones that identified the white space in the 100 announcements, not the ones that competed head-on with them.