Google doubles down on India’s AI future with local data centers and Trillium TPUs

Google Gemini

Google is making a major push to strengthen India’s AI future. The company announced a massive expansion of its local AI hardware capacity, powered by its advanced AI Hypercomputer architecture and the latest Trillium TPUs. The move is aimed at helping Indian businesses, startups, and government organizations train and deploy advanced Gemini AI models locally — a step that underscores Google’s commitment to AI sovereignty and data residency within India.

In simple terms, this means that instead of relying on servers located abroad, Indian developers and companies can now build and run powerful AI systems entirely within India’s borders. This will not only reduce delays in processing (or latency) but also help meet the country’s growing regulatory and compliance needs around data storage and privacy.

Google says the goal is to make its most advanced AI tools available faster than ever to Indian users, with full control over their data. The company is bringing Gemini 2.5 Flash, its next-generation AI model, to India with local machine learning support. It’s also opening early access for customers to test even more powerful upcoming Gemini models — all with complete data residency. This marks the first time Google Cloud is launching its top-tier AI models directly in India.

To make these tools more useful for Indian users, Google is also rolling out several new features designed around local needs. These include batch support for Gemini 2.5 Flash, allowing companies to run large-scale AI tasks affordably within India; a preview of Document AI to automate document-heavy workflows; and a unique capability called Grounding on Google Maps, which will help AI models provide accurate, location-based answers in real time.

But Google’s plan isn’t just about infrastructure — it’s also about people. The company is investing in India’s developer and research ecosystem through collaborations like its new partnership with IIT Madras and AI4Bharat. Together, they’re launching Indic Arena, an open platform that lets anyone in India test and benchmark AI models across different Indian languages. Google Cloud is powering the project with cloud credits, helping researchers create more inclusive AI for India’s diverse linguistic landscape.

“Indic Arena will be a neutral benchmark for understanding how models perform across our many languages,” said Mitesh Khapra, associate professor at IIT Madras.

Google says it wants to empower startups, researchers, and enterprises alike to build AI “by Indians, for Indians.” The company is inviting the country’s growing AI community — from universities to government bodies — to take advantage of this expanded Gemini capacity and sovereign-ready infrastructure.

With this move, Google is signaling that India’s AI journey is not just about using global technology — it’s about shaping it.