Sponsored by edgeful

Google I/O 2026 just wrapped, and it’s pretty clear, Google wants Gemini to become the layer sitting inside Search, Gmail, Docs, YouTube, shopping, coding tools, creative apps, and even smart glasses.

Google is now processing more than 3.2 quadrillion tokens per month across its products. Last year, that number was roughly 480 trillion. Two years ago, it was 9.7 trillion. This is AI usage exploding across Google’s ecosystem. Google also said its model APIs are processing around 19 billion tokens per minute, and more than 8.5 million developers are building with its models every month.

Google now has 13 products with over 1 billion users each, and five of those have more than 3 billion users. AI Overviews has more than 2.5 billion monthly active users, AI Mode has passed 1 billion monthly active users, and the Gemini app has now passed 900 million monthly active users, up from 400 million last year.

That is probably the biggest thing to understand from this I/O. How insane the growth trajectory is.

Let’s start with the biggest one: SEARCH.

For more than two decades, Google Search has mostly worked the same way. You type a few words, Google gives you links, and you start clicking. Now Google is changing that. The new Search box is built for longer questions, follow-ups, images, files, videos, and even Chrome tabs. It can help shape your question before you ask it, then carry the conversation into AI Mode. This is the biggest upgrade to the Search box in more than 25 years.

Search can now create. Instead of only giving you text and links, it can build custom layouts, charts, tables, simulations, and even small dashboards based on what you ask. So if you ask about black holes, Search might build a visual explainer. If you are planning a move or a wedding, it could build a tracker you come back to. That is a big shift. Search is no longer just trying to find the answer. It is trying to build the right format for the answer.

Google is also adding Search helpers that can keep watching the web for you. For example, you could ask one to monitor apartment listings, price drops, sports news, or a product launch, and it would keep checking in the background. These will start with Google AI Pro and Ultra subscribers this summer. Search is also getting booking help, so you can ask for things like a private karaoke room or a local service, and Google can pull together availability, pricing, and links to book.

Now, under all of this is GEMINI 3.5 FLASH, Google’s new fast AI engine.

And it’s pretty cool.

Google says Gemini 3.5 Flash is better than Gemini 3.1 Pro across almost all benchmarks, especially in coding and agent-style tasks.

More importantly, Gemini 3.5 Flash is 4x faster than other frontier models by output tokens per second, and the Antigravity-optimized version is 12x faster.

Google also said top companies are processing around 1 trillion tokens a day, and if they shifted most of that work to a mix that includes Gemini 3.5 Flash, they could save more than $1 billion a year.

Gemini 3.5 Flash is available now in the Gemini app, AI Mode in Search, Google Antigravity, Google AI Studio, Android Studio, and enterprise tools. Gemini 3.5 Pro is coming next month.

For developers, the big news was ANTIGRAVITY 2.0.

That is Google’s coding platform where AI helpers can work together on software projects. There is now a desktop app, a command-line version, and a developer kit. Plain English version: Google wants developers to describe what they want built, then have teams of AI helpers work through the code, testing, and project pieces in parallel.

Google AI Studio also got a big update. You can now build native Android apps from prompts, preview them in a browser, test them on a device, and even publish to Google Play’s internal testing track. That matters because it lowers the bar for building real apps, not just mockups or demos.

TPU 8T AND TPU 8I

Google also spent time on the infrastructure behind all of this.

It introduced its 8th-generation TPUs, split into two chips: TPU 8t and TPU 8i. TPU 8t is built for large-scale AI training, while TPU 8i is built for inference, which is the part that matters when AI has to respond quickly in real products.

Google said TPU 8t has nearly 3x the raw computing power of the previous generation, and its training systems can now scale across more than 1 million TPUs globally. TPU 8i is focused on speed and low latency. Both chips are also supposed to deliver up to 2x better performance per watt, which matters because running AI at this scale is expensive and power-hungry.

The most everyday-use announcement was probably GEMINI SPARK.

Spark is Google’s new 24/7 personal AI helper. It runs in the cloud, so it can keep working even if your laptop is closed or your phone is locked. It connects with Google tools like Gmail, Docs, Slides, and more, and it can help with things like organizing plans, pulling notes together, managing projects, or keeping track of updates.

Google says Spark stays under your direction, and it is supposed to ask first before high-stakes actions like sending emails or spending money. Trusted testers get it this week, and Google AI Ultra subscribers in the U.S. are expected to get the beta next week.

The Gemini app itself also got a full redesign. Google is calling the new look “Neural Expressive,” which basically means brighter visuals, smoother motion, new typography, haptics, and a more interactive feel. The bigger change is that Gemini is moving away from long walls of text. Responses can now include images, timelines, narrated videos, and dynamic graphics that are built around your question.

There is also a new feature called DAILY BRIEF. Think of it like a morning rundown built from your connected apps. It can pull urgent updates from Gmail, upcoming events from Calendar, and follow-up details into one skimmable briefing. It can also suggest next steps, which is where Google keeps pushing this idea that Gemini should not just tell you what happened, but help you decide what to do next.

On the creative side, Google announced GEMINI OMNI.

This is its new model for creating from different kinds of input, starting with video. You can bring in text, images, video, or audio, then use normal language to edit or create a video. Google says Omni can keep characters consistent, understand physical movement better, and let you refine a scene over multiple turns. Gemini Omni Flash is rolling out to the Gemini app, Google Flow, and YouTube Shorts.

Google Flow and Flow Music also got upgrades. Flow now has a creative helper that can brainstorm, create variations, organize assets, and batch edit. Flow Music can now edit songs section by section, rewrite or translate lyrics, change a beat drop, create covers, and use Gemini Omni to make music videos. There are also mobile apps coming for Flow and Flow Music.

YouTube got its own AI moment too.

ASK YOUTUBE is a new conversational search feature that lets you ask more specific questions and get structured answers from across YouTube videos and Shorts. So instead of searching five different phrases and scrubbing through videos, you can ask something like how to teach a child to ride a bike, then refine from there. It is available now for U.S. Premium members 18 and older, with a broader rollout planned.

YouTube Shorts is also getting Gemini Omni remix tools. Users can remix eligible Shorts by adding prompts and images, changing the scene, changing the style, or inserting themselves into a video. These remixes will include watermarks and metadata, link back to the original video, and creators can opt out. YouTube is also expanding likeness detection to creators 18 and older.

Workspace also got a lot of practical updates.

Gmail Live lets you ask your inbox questions by voice, like “What’s my flight gate?” or “What’s happening at my kid’s school this week?” Docs Live lets you talk through an idea and turn it into a cleaner draft. Probably the one new feature that will get used a lot. Keep is getting similar voice features. Google also announced Google Pics, a new image creation and editing tool for things like flyers, invitations, social posts, and photo edits. It lets you move or change specific objects, edit text inside images, translate text, and work inside apps like Slides and Drive.

Shopping also got a major AI update with UNIVERSAL CART.

This is a cart that follows you across Search, Gemini, YouTube, and Gmail. Add something once, and the cart can watch for deals, price drops, restocks, payment perks, and even product compatibility. Google gave the example of building a PC, where the cart could flag parts that do not work together. Universal Cart rolls out in the U.S. across Search and Gemini this summer, with YouTube and Gmail coming later.

Google also talked a lot about labeling AI content. SynthID, its watermarking system, is expanding across Search, Gemini, Chrome, Pixel, and Cloud. Google is also adding C2PA Content Credentials, which can help show whether something came straight from a camera or was changed with AI tools. You will be able to ask things like “Is this made with AI?” in Search features such as Lens, AI Mode, and Circle to Search. OpenAI, Kakao, and ElevenLabs are also adopting SynthID for some AI-generated content.

Then came the SMART GLASSES.

Google showed off new “intelligent eyewear” built with Samsung, Gentle Monster, and Warby Parker $WRBY ( ▼ 10.96% ). The first versions are audio glasses coming this fall. They let you talk to Gemini, ask about what you are seeing, get walking directions, send texts, take photos and videos, translate speech or text, and tap into apps like Uber or DoorDash. They will pair with both Android and iOS phones.

Google also previewed Android Halo, a small status area at the top of your phone that shows what your AI helper is doing. So if Spark is working on a task, going live, or sending you a message, you can see it without leaving what you are doing.

There were a few more updates worth knowing. Project Genie is getting Street View support, so people can create interactive worlds based on real U.S. places. It is rolling out to eligible $200 AI Ultra subscribers. Google also announced Gemini for Science, a collection of tools for research, hypothesis generation, computational experiments, and literature review. And Stitch, Google’s AI design tool, now lets people steer and reflow designs in real time using text, voice, files, or code.

And yes, the pricing changed. Google introduced a new $100-per-month AI Ultra plan aimed at developers, advanced creators, and power users. It also dropped the top AI Ultra plan from $250 to $200 per month. The $100 plan includes higher usage limits than Pro, priority access to Antigravity, 20TB of storage, YouTube Premium, and access to new features like Gemini Spark in the U.S.

That’s all!