Google Unleashes AI's Next Frontier At I/O 2025: From Beam To Gemini, It's All About Real-World Impact
Unlike previous years, when Google held back its biggest announcements for I/O, Pichai made it clear that the "Gemini era" is redefining the company's rollout strategy.
At the annual Google I/O 2025 event, CEO Sundar Pichai took centre stage to outline how the tech giant is turning years of AI research into powerful, real-world applications — faster than ever before. The company’s keynote, by CEO Sundar Pichai, highlighted sweeping advances in its Gemini model family, cutting-edge infrastructure, and bold experiments that are quickly becoming mainstream features across its ecosystem.
AI Is No Longer Waiting for the Spotlight — It’s Shipping Weekly
Unlike previous years, when Google held back its biggest announcements for I/O, Pichai made it clear that the "Gemini era" is redefining the company's rollout strategy. “We want to get our best models into your hands and our products ASAP. And so we’re shipping faster than ever,” he said.
At #GoogleIO, we shared how decades of AI research have now become reality.
— Sundar Pichai (@sundarpichai) May 20, 2025
From a total reimagining of Search to Agent Mode, Veo 3 and more, Gemini season will be the most exciting era of AI yet.
Some highlights 🧵 pic.twitter.com/2n9rbGNj0Q
That speed is backed by serious infrastructure muscle. Google’s new Ironwood TPUs, the seventh-generation processors tailored for inferential AI, are delivering 10x the performance over previous versions, with a jaw-dropping 42.5 exaflops per pod. This enables Google to not only scale faster but also lower the cost of AI services dramatically. The results? Gemini 2.5 Pro now dominates the LMArena leaderboard, marking a 300-point Elo score leap from the original Gemini Pro.
From Starline to Beam: Making Communication More Human
Google's futuristic 3D video call experiment, Project Starline, has evolved into something even more ambitious — Google Beam. Built in collaboration with HP, Beam blends AI with an array of six cameras to turn flat video feeds into life-like 3D conversations. The platform includes millimetre-precise head tracking at 60 frames per second for an ultra-immersive feel.
This isn’t Google’s only push in redefining virtual interaction. “In Google Meet, we’re adding near real-time speech translation that matches voice tone and even facial expressions,” Pichai announced. Initially rolling out in English and Spanish to AI Pro and Ultra subscribers, this feature will expand to more languages soon and become available to business users for testing.
Your AI Sidekick Just Got Smarter: Gemini Live and Agent Mode Debut
Project Astra — Google's exploration into context-aware AI assistants — has now materialised as Gemini Live. Users can interact with it through their phone’s camera or screen sharing, using it for tasks as varied as mock interviews to fitness planning. Gemini Live is now available on Android and rolling out on iOS.
Meanwhile, Google’s prototype agent system, once dubbed Project Mariner, is being upgraded with multitasking abilities and a new “teach and repeat” method. The experimental Agent Mode in the Gemini app will soon help users complete complex tasks like apartment hunting — even booking a tour — using tools like Zillow. Trusted testers like Automation Anywhere and UiPath are already building solutions on it.
Search Reimagined, Creativity Supercharged
AI is also reinventing Search with a new “AI Mode” — a dedicated tab that lets users issue longer, more complex queries and refine results through follow-ups. Pichai confirmed that AI Overviews now reach over 1.5 billion users globally and are boosting query engagement in major markets like the U.S. and India.
Creatives have plenty to be excited about, too. Imagen 4, Google’s most capable image model yet, and Veo 3, a video model with native audio generation, are both available in the Gemini app. For filmmakers, Google introduced Flow — a tool that expands short clips into cinematic scenes.
AI That Knows You: Personal Context in Gemini
One of the most personal touches to Google’s AI push is the idea of “personal context.” With permission, Gemini can now use your data across Gmail, Drive, and Docs to craft responses that sound uniquely like you. A standout example: smart replies in Gmail that mirror your tone, word choices, and past trip itineraries.
“It will match your typical greeting and capture your tone, style and even favourite word choices,” said Pichai. This feature will become available to subscribers later this year.
From Inspiration to Everyday Utility
Reflecting on the impact of AI, Pichai shared a touching moment riding in a Waymo with his elderly father, now a top tourist attraction in San Francisco. “I saw the progress in a whole new light,” he said. “It was a reminder of the incredible power of technology to inspire, to awe and to move us forward.”
As Google races ahead in the AI arms race, its focus remains clear: make AI not just powerful, but deeply personal and profoundly useful.
Top Headlines
