Google CEO Sundar Pichai has declared a new era in artificial intelligence at the Google I/O 2025 conference, unveiling the powerful capabilities of the Gemini 2.5 Pro model. According to Pichai, the upgraded model has surged to the top of the LMArena leaderboard across all categories, showcasing significant improvements driven by Google’s world-class infrastructure.
At the heart of this progress is Google’s latest seventh-generation TPU, Ironwood, purpose-built to handle complex inferential AI workloads. The new TPU delivers ten times the performance of its predecessor and boasts an impressive 42.5 exaflops of computing power per pod.
One of the key announcements was the launch of AI Mode — a chatbot experience within Google Search that allows users to handle more complex queries. AI Mode, soon to be available to all U.S. users, lives in a separate tab and is designed for tasks such as comparing fitness trackers, finding affordable event tickets, or generating customized charts and graphics. It also supports follow-up questions, enhancing conversational search.
Running on Gemini 2.5, AI Mode is expected to gradually integrate into the core Search experience through AI Overviews. These summaries, which appear at the top of search results, currently reach over 1.5 billion people each month. Google says most users interact meaningfully with these overviews — whether by clicking, reading, or engaging with the content.
Gemini’s influence is also expanding into e-commerce. Users will soon be able to upload a photo of themselves to see how clothing looks on a digital version of their body. Google is also introducing an AI-powered shopping assistant that tracks specific product preferences and alerts users when their chosen items go on sale — even completing purchases on their behalf if desired.
Pichai also introduced a new “Deep Think” mode in Gemini 2.5 Pro, designed for advanced reasoning. With this feature, the AI can turn a grid of photos into a 3D narrated sphere and perform real-time language switching with its text-to-speech functionality.
He emphasized that the latest version comes with enhanced security, transparency, user control, and cost-efficiency — underscoring Google’s commitment to safe and scalable AI development.
