Google I/O 2026: AI Takes Center Stage With New Gemini and Video Generator

When Google opens its developer conference I/O in Mountain View next Tuesday, artificial intelligence is once again expected to take center stage. Anticipated are a new Gemini version as well as a previously unconfirmed video generator called “Omni” – both building blocks of a strategy with which Google aims to gain ground in the race against OpenAI and Anthropic.

Google I/O 2026 takes place on May 19 and 20, with the keynote scheduled for 7 p.m. Central European Time. It is the most important date on the company’s calendar: at I/O, Google traditionally consolidates its product and platform announcements for the coming year – from Android and Chrome to the cloud. Over the past three years, however, the conference has been entirely dominated by AI strategy, and 2026 is unlikely to change that.

A New Gemini Model – But No Frontier Breakthrough

According to a report by tech journalist Alex Heath, Google plans to unveil a new Gemini version on Tuesday. His impression: the model will perform at roughly the same level as OpenAI’s recently released GPT-5.5 – and thus fall well short of Anthropic’s Mythos, which has recently established itself as the new reference point for frontier models, even though it is not yet widely available.

Which version number Google will choose remains open. Given the previous release cadence of three to four months, analysts consider Gemini 3.2 or 3.5 most likely; a jump to Gemini 4.0 is considered possible but less probable. For investors, the question is crucial because Google is expected to demonstrate not only technological progress but increasingly also concrete monetization steps. In the first quarter of 2026, the company’s search business grew by around 19 percent, driven in part by AI Overviews and AI Mode in search.

“Gemini Omni”: A Video Generator That Leaked Early

In parallel, hints about a new AI model called Gemini Omni have been circulating for about a week; it is said to generate and edit videos directly within the Gemini chat. In recent days, some users reported that the model was briefly suggested to them in the Gemini app, followed by the message: “Meet our new video generation model. Remix your videos, edit directly in chat, try a template, and more.”

The information stored in the metadata suggests that Omni is built on Veo, Google DeepMind’s established video generator, which recently launched in the variant Veo 3.1 Lite as a more cost-effective version. It remains unclear how Omni will be positioned relative to Veo. The publisher TestingCatalog suspects that Google could combine Omni’s capabilities with those of the image generator Nano Banana 2 – Omni would then be less a successor to Veo and more a standalone, multimodal Gemini model. Early tests point to high prompt fidelity, better audio quality than Veo 3.1, and embedded background music.

An official confirmation is still pending. However, the fact that the sightings are appearing just days before I/O is widely regarded as a clear signal that Google will present the model on stage.

What Else to Expect

Google already addressed the bulk of its Android announcements on May 13 at the preceding “Android Show: I/O Edition.” The focus there was on “Gemini Intelligence” – a suite of proactive, agentic AI features designed to run across smartphones, wearables, cars, and laptops. Also introduced was the new device category “Googlebook,” a range of Gemini-centric laptops from Acer, ASUS, Dell, HP, and Lenovo, set to launch in the fall. Additionally, Gemini in Chrome with an “Auto Browse” feature and Material 3 updates for Android Auto were presented.

For the main keynote on Tuesday, AI and developer topics are therefore primarily expected: updates to the Gemini family, new tools for agentic AI, expansions to the generative models Veo, Imagen, and potentially Omni, as well as an outlook on Genie 3, Gemma 4, and the robotics model Gemini Robotics ER-1.6. Google has also already confirmed a first look at the Android XR glasses planned for late 2026 – developed in partnership with Samsung, Warby Parker, and Gentle Monster.

Why the I/O Matters

The I/O is more than a product show. For Google, it is the central stage to simultaneously address the narrative of AI transformation to developers, advertisers, and Wall Street. Following recent quarterly figures, the cloud business sits on a backlog of around $462 billion, driven in part by the sale of its own TPU chips. At the same time, OpenAI and Anthropic are pushing into market segments that Google has so far dominated through search, Android, and Workspace – with their own model releases and product initiatives, from OpenAI’s rumored AI phone to Anthropic’s Mythos.

The fact that the Gemini version expected on Tuesday will likely not mark the absolute top of the performance scale is therefore no trivial problem for Google: the company must show that the distribution of its models across billions of devices and their integration into its advertising and cloud infrastructure can compensate for the lead that competitors have built up on the pure model front. I/O 2026 will show whether this calculation pays off.