MUMBAI: Forget slow-burn product cycles and hush-hush unveilings. At Google I/O 2025, Google parent Alphabet CEO Sundar Pichai made one thing very clear: in the Gemini era, Google isn’t waiting for the stage — it’s shipping on Tuesdays.
“We’re moving faster than ever,” Pichai told the crowd, revealing a dizzying array of AI-powered updates, model breakthroughs, and bleeding-edge products — all underpinned by Google’s latest powerhouse: Gemini 2.5 Pro. The model has surged 300 Elo points since its first generation and now sweeps the LMArena leaderboard. The driving force? Ironwood, Google’s seventh-gen TPU, serving up a brain-melting 42.5 exaflops per pod.
The pace of adoption is just as staggering. Over the past year, Pichai explained that:
* Token usage exploded from 9.7 trillion to 480 trillion a month.
* More than seven million developers are building with Gemini — up five times.
* The Gemini app has clocked 400 million monthly users, with a 45 per cent spike among 2.5 Pro users.
With scale like that, Pichai says, “we’re in a new phase of the AI platform shift — turning decades of research into daily reality.”
Say hello to Google Beam — the spiritual sequel to Project Starline. The new AI-first video chat platform uses six cameras, advanced head-tracking and 3D lightfield displays to create a shockingly lifelike experience. In short: Zoom calls, eat your heart out. With HP on board, Beam hardware hits early testers later this year.
Also in the mix: real-time speech translation for Google Meet, matching tone, voice and facial expressions. English and Spanish are live in beta now; more tongues to come.
Google’s AI assistant ambitions are now tangible. Agent Mode, based on Project Mariner, is headed to the Gemini app. Think digital butler with brains: it browses listings, tweaks filters, and can even schedule a house tour — all while chatting like a pro.
Expect agents to be everywhere soon: Search, Chrome, Workspace — even chatting with each other using the new Agent2Agent protocol. Google also confirmed Gemini API and SDK support for Anthropic’s Model Context Protocol.
On the personalisation front, Smart Replies in Gmail are getting a glow-up. With permission, Gemini will mine your emails and files to craft eerily you-like responses, complete with tone and pet phrases. Yes, your AI is about to start writing better emails than you.
AI is now baked right into Google Search. The newly announced AI Mode allows for longer, more complex queries and natural follow-ups — already available in the US. With Gemini 2.5 Pro now powering it, Search just got a serious IQ boost. AI Overviews are already live in 200 countries and growing fast, especially in the US and India., revealed Pichai.
Developers, meet your new best friend: Gemini 2.5 Flash, a lightning-fast, low-cost version built for speed and scale. But the headline act is Deep Think, a new reasoning mode for Pro users that layers in parallel thinking and long-context analysis. Consider it AI with a PhD.
Creatives, your time has come. Google dropped Veo 3, its most advanced video model yet — now with native audio generation. It also rolled out Imagen 4 for image generation, and Flow, a slick tool for building cinematic clips on the fly.
Throw in Canvas integration, support for quizzes, infographics and multilingual podcasts, and the Gemini app is shaping up to be a Swiss Army knife for creators.
Pichai closed with a personal anecdote: a Waymo ride in San Francisco with his 80-something father. “He was amazed,” Pichai said. “It reminded me just how powerful technology can be — not just to dazzle, but to bring people along.”
The message was clear: AI at Google is no longer a moonshot — it’s now a movement.