In his May 20, 2025 keynote at Google I/O, CEO Sundar Pichai set the tone with a bold vision of “making AI more helpful with Gemini”. He emphasized that Google is entering a new phase where cutting-edge AI models (notably Gemini 2.5) are being shipped faster than ever and deeply woven into every product. Google’s updated infra (7th-gen TPU “Ironwood” offering 10× the speed) lets it deliver these models at scale. As Pichai put it, “the world is responding, adopting AI faster than ever before” – Google now processes roughly 480 trillion input “tokens” per month (50× last year) and has over 7 million developers building with Gemini (5× more year-over-year). The Gemini app alone exceeds 400 million monthly users. This surge in adoption underpins Google’s strategy: to infuse AI into core experiences (search, Workspace, Android, Chrome, etc.) without sacrificing its existing strengths (search and ads).
The keynote and blogs made clear that Google’s roadmap blends aggressive AI innovation with new hardware horizons (XR glasses and devices). For example, Pichai introduced Google Beam (the successor to Project Starline) – an “AI-first 3D video communications platform” built on Google Cloud infrastructure. Beam uses six cameras and an advanced volumetric model to transform a standard video call into a lifelike 3D experience, enabling “true-to-life 3D video communication” between distant offices. (Beam devices, in partnership with HP and integrators, will roll out to enterprises later this year.) He also described new agentic AI projects like Project Astra and Mariner that bring camera input and web actions to AI assistants – for instance, Gemini’s “agent mode” can now sign you up for an apartment tour by querying Zillow and scheduling a visit on your behalf.
Below, we analyze the biggest AI and XR announcements, their industry and consumer impact, and additional highlights from I/O 2025. Throughout, Google’s own blogs and press materials are cited, along with media coverage and expert commentary.
AI Takes Over Search, Workspace, and Android
Search gets an AI makeover. Google doubled down on “AI Mode” in Search, a chat-like interface powered by Gemini. In a dedicated Search blog, Google announced that AI Mode is now rolling out to users in the U.S. (no signup needed). This mode goes “beyond information to intelligence” – it uses advanced reasoning and multimodal understanding (text, images, voice) to tackle complex queries. Under the hood, AI Mode uses Google’s “query fan-out” technique: it breaks your question into sub-questions and runs many web searches at once, then aggregates the findings. A custom version of Gemini 2.5 (the new top model) is being deployed into AI Mode and AI Overviews. Google demonstrated how AI Mode can plan a trip with one query – it created restaurant lists, custom maps, live music venues, etc. – instead of the user clicking dozens of links. In effect, Google is now doing much of the “Googling” for you.
AI Mode also gains new features:
- Deep Search – an upcoming enhancement that can ingest additional context (e.g. documents, web pages, or spreadsheets) and issue “dozens or hundreds” of queries under the hood for thorough research.
- Live View queries – by tapping a camera icon (in Search or Lens), you can point your phone at text or scenes and ask questions. Google says Search will “become a learning partner that can see what you see,” explaining concepts or suggesting actions based on real-world visual cues.
- Agentic tasks – AI Mode can now take actions on your behalf. For instance, telling it to find baseball tickets will trigger a fan-out search across sites and present options that match your criteria. Google will even let AI Mode complete purchases for you: a new “agentic checkout” feature lets you set a price and have Google add an item to your shopping cart and pay with Google Pay when it hits that price. This tight shopping integration also extends to fashion: users can virtually try on clothes using their own photos, with Gemini-powered outfit recommendations.
Liz Reid, Vice President of Google Search, told The Verge that this approach “goes way deeper than a traditional search,” tapping the Shopping Graph, Maps data, and knowledge graph simultaneously. (Axios observes Google is trying to “make its core products better through AI without displacing its advertising and search businesses”.) In practice, AI Mode and Overviews will start to merge: Google plans to graduate useful AI Mode features into the main search experience over time.
Workspace and apps get smarter. The Gemini assistant is permeating Google’s productivity apps. Real-time speech translation is coming to Google Meet: using a DeepMind audio-language model, Meet will overlay translated speech (with original tone) during calls. Pichai noted this “breaks down language barriers,” and English–Spanish translation is rolling out immediately to Google AI Pro/Ultra subscribers in beta (with more languages and Workspace customer trials soon). In Gmail and Docs, Gemini will provide AI assistance directly. For example, later this year Gmail will offer personalized Smart Reply: if a friend asks you for travel advice, Gemini can “search your past emails and files in Drive (itineraries, tickets, etc.)… and draft a helpful response”. Google says these “personal context” features respect privacy and require opt-in, but they promise to make AI highly customized to you.
Gemini is also coming to Chrome. Starting immediately, a limited preview of “Gemini in Chrome” lets you summon the assistant on any webpage to explain content or automate tasks. Google marketing says users can “understand complex information and complete tasks on the web” using page context. This blurs the line between Google Search and the AI assistant – effectively embedding Gemini into the browser.
On Android phones, Gemini (via Google Assistant) is expanding to new tasks. The keynote showed an “Agent Mode” in the Gemini app, where you can ask the assistant to do chores like apartment hunting: it will query sites like Zillow, filter listings, and even schedule tours. This powerful, agentic mode (powered by Project Mariner) will be coming to subscribers soon. And on Android Auto, AI will soon power in-car voice commands that use your phone’s understanding. In all these ways, Google is pushing Gemini to be a “universal assistant” that can see, hear, and act across contexts.
Android 16 and devices. Unsurprisingly, Android also got several upgrades. Google previewed Android 16, focusing on adaptive UIs for foldables/tablets. There’s a new Jetpack Compose for TV (bringing Gemini capabilities to Google TV this fall) and a refreshed Wear OS 6 with “Material 3 Expressive” theming for watch faces. Notably for XR, Google revealed Android XR – a new platform for headsets and glasses built in the Gemini era. Developer Preview 2 of the Android XR SDK launched, alongside an ecosystem roadmap: Samsung’s Project Moohan headset and an upcoming Xreal “Project Aura” glasses are in the works. Google also announced partnerships with eyewear brands (Gentle Monster, Warby Parker, and future Kering Eyewear) to make “stylish glasses with Android XR”. By year’s end, developers will be able to build apps for these AR glasses, suggesting Google expects this form factor to explode.
Generative AI and Creative Tools
Gemini 2.5 – bigger and better. Under the hood, Google continued to press its Gemini model family. The Gemini 2.5 series (Flash and Pro) was previewed with stronger reasoning abilities. In particular, “Deep Think” – an enhanced reasoning mode for Gemini 2.5 Pro – is designed to consider multiple hypotheses before answering, making it suited for complex tasks (math, coding). Google Cloud says Deep Think will appear in Vertex AI for trusted testers first. Importantly, Google emphasized security: a new privacy-preserving approach makes Gemini 2.5 “our most secure model family to date”. Both Gemini 2.5 Flash and Pro will become generally available (on Vertex AI for enterprise) in the coming weeks.
In consumer apps, Google highlighted its multimodal generation models. The Gemini app will gain Imagen 4 (new image generation with better text rendering) and Veo 3 (video generation). Google demonstrated Flow, an AI-powered video-editing tool (built on Veo 3) for filmmakers. Flow lets creators craft cinematic clips via text prompts, with camera controls and storyboarding. For example, a YouTuber could say “show me a sunset timelapse in New York,” and Flow generates video accordingly. These tools will be available to Google AI Ultra subscribers and, in limited form, to Pro users. Google also unveiled Jules, an autonomous coding agent in public beta. Jules can connect to your code repo and perform tasks like writing tests or fixing bugs on its own, freeing developers to focus on higher-level work. Meanwhile, Gemini Code Assist (chat-based coding help) is now freely available inside VS Code, JetBrains IDEs, and Google Cloud Shell. In short, Google showed an all-out assault on creative domains – from writing code to composing videos and images – powered by its latest models.
New subscription: Google AI Ultra. To monetize these advances, Google introduced a top-tier AI subscription – Google AI Ultra – a $249.99/mo tier for “the highest usage limits and access to our most capable models and premium features”. This builds on its existing Google One/AI Pro plans. AI Ultra users get unlimited access to the best models: Gemini app with Deep Research and Veo 3 video, Flow at 1080p, the Whisk image-to-animation tool (highest limits), enhanced NotebookLM notebook features, and always-available Gemini in Workspace apps (Gmail, Docs, Vids). They also get early access to new features like Gemini in Chrome and Deep Think. In exchange, Google AI Ultra comes with bonuses: an individual YouTube Premium subscription and 30 TB of cloud storage. (Google will rebrand its old $20/month “AI Premium” to Google AI Pro, which gets Flow (Veo 2) and Chrome Gemini immediately at no extra cost.) These tiers signal Google’s strategy to cater both to casual users (AI Pro) and power users/enterprises (AI Ultra) who want enterprise-grade assistance.
Generative media on Cloud. On the Google Cloud side, I/O highlighted new generation models for media: Imagen 4, Veo 3, and a new Lyria 2 (for music synthesis) are now available via Vertex AI. These let businesses generate video, images, and audio from text prompts, with built-in watermarking and safety filters. The Cloud blog stressed these models “give you excellent ways to create visual and audio content” under enterprise-friendly policies. In addition, Vertex AI gains user-friendly features like thought summaries (to audit an agent’s reasoning steps) and the aforementioned Deep Think mode.
For developers, Google Cloud announced many productivity tools. Firebase Studio – a new cloud-based “AI workspace powered by Gemini” – can turn design mockups (from Figma) into full-stack apps. It now auto-detects if your app needs a backend (Auth, Firestore, etc.) and will provision it for you. For deployment, Cloud Run added one-click publishing from Google AI Studio and support for directly deploying Google’s generative models (Gemma 3) to scalable GPU endpoints. Cloud Run also launched an “MCP server” so that AI agents (via the new Model Context Protocol) can auto-deploy services as needed. Altogether, these updates aim to let enterprises build and deploy AI applications with minimal ops work.
XR and Immersive Computing
Android XR: Glasses and Headsets. One of I/O’s biggest surprises was Google’s progress on AR/VR (XR). After laying groundwork last year, Google now showed a running prototype of Android XR smart glasses. These glasses (un-named consumer prototype) have a camera, mic, speakers and an optional in-lens display. In a demo, Pichai and team had people send each other text messages, get turn-by-turn walking directions, and crucially use real-time translation through the glasses. The live translation demo let two people speak English and Spanish, with the glasses overlaying subtitles (“giving you subtitles for the real world”). Google’s message is that Gemini’s camera understanding can let glasses assist naturally – for example, knowing your context to automatically adjust appointments or fetch info. As the Android blog notes, pairing these glasses with Gemini “means they see and hear what you do, so they understand your context…and can help you throughout your day”. In short, Google envisions a lightweight AR interface where AI is always listening and guiding hands-free.
However, Google acknowledges style matters for wearables. It announced partnerships to build fashionable smart glasses. Starting now, Google is working with eyewear firms Gentle Monster and Warby Parker to design frames that run Android XR. (Samsung is also broadening its XR partnership – beyond its Moohan headset – to help jumpstart an ecosystem of wearables.) The plan is to release reference hardware and SDKs later this year, so third-party companies can make Google-powered AR glasses. On stage, Google said it is already gathering feedback from prototype testers to ensure privacy and usability.
Samsung headsets and Android XR platform. On the VR side, Google confirmed the first Android XR headset (Samsung’s Project Moohan) will ship later this year. With Qualcomm, Google is building the OS for powerful VR headsets with “infinite screen” environments. (Qualcomm’s new chip for XR was also announced, promising high performance.) The Android XR SDK Developer Preview 2 is now available for app makers. In other words, Google is throwing its weight into an AR/VR platform that is open (Android-based) rather than closed like Apple’s Vision Pro.
Google Beam (3D video conferencing). As noted, Google Beam (née Starline) is another XR innovation. Officially, Google describes Beam as “the next chapter of Project Starline” – an AI-powered 3D telepresence system. Beam uses advanced compression and Google Cloud AI to reconstruct a 3D image of each person from ordinary video cameras, enabling eye contact and lifelike depth cues. Beam will initially be pitched at enterprises; Google is working with Zoom, HP, and integrators to deploy Beam stations at corporate offices. This could impact sectors like telemedicine and remote collaboration by making virtual meetings feel more natural than flat video calls.
Immersive experiences and emerging tech. Beyond hardware, Google showcased consumer-facing XR and AI demos. For example, the Gemini app now has a “Gemini Live” mode (from Project Astra) where you can share your phone’s camera live with an assistant – e.g. showing a real-world object and asking questions about it. Pichai noted Gemini Live is already on Android and rolling out to iOS, letting the assistant “understand the world around you” via vision. In Google Maps, we saw hints of AR directions (walking guidance overlaid on the camera view). Google also teased a partnership with Xreal (formerly Nreal) – on I/O we learned Xreal’s new AR glasses, “Project Aura,” will run Android XR and Gemini. (TechCrunch reported a $1000 price for that device.) These moves signal that Google’s XR plans span both internal R&D and an ecosystem of partners.
Other Highlights
- Shopping with AI: Google is embedding AI into shopping. In AI Mode, the Shopping Graph (50+ billion listings) is fully integrated, and Gemini helps narrow choices visually. Users can virtually “try on” clothes using their own image. The “agentic checkout” (discussed above) promises to automate price tracking and buying. This could reshape e-commerce by making Google an active shopping agent.
- Android & Wear OS: Aside from XR, Google noted a few platform updates. Wear OS 6 (coming soon) will introduce “Material 3 Expressive” on smartwatches – customizable watch faces and motion effects. Google also previewed Android TV improvements: a stable Compose for TV release, new content APIs, and bringing Gemini to TV by fall (so you could ask your TV for recommendations via voice).
- Cloud and Enterprise: On the cloud side, Google’s big announcements (AI Studio, Vertex updates, FireSat, drones/space projects) were covered at Google Cloud Next rather than I/O, but the Cloud blog summarised them. In brief, enterprises get more AI tools (AutoML Vision for data tables, Fire Satellite imagery on Google Cloud, etc.). Importantly, Google emphasized scalability and security for business AI: e.g. the firm boasts Gemini 2.5 is now “the most secure model family to date”, addressing corporate concerns about AI risk. It also expanded free Google AI Pro for education in more countries.
- Developer Tools: I/O gave many developer-centric updates. Android developers got new ML Kit GenAI APIs (Gemini Nano) for on-device AI tasks, along with improved Jetpack libraries. Kotlin Multiplatform got a new shared module template. Firebase’s AI-assistant features (in Studio) were highlighted above. Google also announced Jules (beta) for coding and Gemini Code Assist (free) for all developers in major IDEs. As a whole, Google is arming devs with AI helpers to write code, build apps, and deploy them without managing servers.
Throughout these announcements, Google executives pitched clear benefits: AI to boost productivity, creativity and accessibility. Translations and contextual assistants promise to help users across languages and abilities. Flow and Imagen help creators tell stories in new ways. And by tying data (with permission) into AI, Google aims to personalize experiences while keeping control at the user’s hands.
Industry observers have noted the broader impact. Axios’s Ina Fried summarized that Google’s I/O was “more AI in more places” – a sign that Google is prioritizing its AI-driven roadmap to stay ahead of rivals. Venture investors and enterprise CIOs will be watching how Google Cloud’s new AI services (Vertex models, Cloud Run agents, etc.) simplify operations. Developers will likely welcome tools like Firebase Studio and the new AR SDK to build next-gen apps. For consumers, the change will be gradual: AI Mode in search and Translate for Meet arrive in the coming weeks, and XR devices are still months away. But Google’s message is that AI assistance will soon be woven into most products you use – from your browser to your sunglasses.
# All speakers at the Google I/O conference 2025
Conclusion: The Road Ahead
Google I/O 2025 mapped out a future where artificial intelligence and immersive computing converge. As Sundar Pichai put it, we are moving “from research to reality” – innovations once in labs (giant AI models, 3D video calls, AR glasses) are now being productized. If successful, these announcements could reshape how we interact with technology: search will do the searching for us, our calendar and email will write themselves, shopping will be guided by AI, and screens may give way to glasses.
At the same time, Google faces challenges – ensuring user privacy, handling data responsibly, and giving developers time to adapt. Investors will judge whether Google’s AI bets pay off without undermining its core ad business, as Axios notes. Consumers will weigh the convenience of AI (e.g. auto-completed tasks) against concerns about handing over control to algorithms. Tech communities will watch how the Android XR ecosystem grows amid competition from Apple and Meta.
One thing is clear: Google’s I/O 2025 signals that AI is the connective tissue across Google’s ecosystem. From deep servers in data centers (new AI chips, Vertex AI) to devices on our faces (AR glasses), AI is in every layer. The announcements suggest Google intends to keep pushing the frontiers of computing. As Pichai said, Google’s goal is to deliver “more intelligence… for everyone, everywhere”. How well this plays out will depend on execution – and on how end users and businesses embrace the new reality Google is crafting.
Sources: Official Google I/O press releases and blog posts (keynote transcript, technology blog, Google for Search/Shopping, Android blog, Cloud blog, Google One blog); coverage from Axios, TechCrunch and The Verge. (All statements are supported by cited sources above.)
Hi I'm Olly, Co-Founder and Author of CybaPlug.net.
I love all things tech but also have many other interests such as
Cricket, Business, Sports, Astronomy and Travel.
Any Questions? I would love to hear them from you.
Thanks for visiting CybaPlug.net!
