Google has announced a major expansion of its artificial intelligence (AI) infrastructure in India, marking one of its largest local capacity investments to date. The initiative will allow Indian businesses and public sector organisations to train and deploy advanced Gemini models entirely within the country, meeting local data sovereignty and residency requirements.

Powered by Google’s AI Hypercomputer architecture and the latest Trillium TPUs, the upgraded infrastructure is designed to support high-performance, low-latency AI applications. According to the company, this will help Indian enterprises develop AI models that capture the nation’s distinct business logic and cultural context.

“We’re committed to bringing our latest AI advancements to India faster than ever, with the controls you need,” Google stated in its announcement blog.

Bringing Next-Generation Gemini Models to India

Earlier this year, Google Cloud introduced Gemini 2.5 Flash for regulated Indian clients with local machine-learning processing support. Now, the company is opening early testing for its latest and most powerful Gemini models in India, the first time Google Cloud has done so globally. Future Gemini releases will also feature full data residency support.

The company is simultaneously expanding its AI tools and services locally:

Batch support for Gemini 2.5 Flash is now available, helping organisations manage high-volume, non-real-time AI workloads in India more cost-effectively.

Document AI, currently in preview, introduces localised document automation capabilities for Indian businesses.

Grounding on Google Maps, a new feature, enables AI applications to deliver real-time, location-aware insights based on Google Maps data.

Building a Sovereign AI Ecosystem

Beyond infrastructure, Google emphasised its commitment to nurturing India’s AI talent and research ecosystem as part of its broader push towards digital sovereignty. To this end, Google Cloud and Google DeepMind have partnered with IIT Madras to launch Indic Arena, an open benchmarking platform developed by the AI4Bharat Centre for evaluating AI model performance across India’s diverse languages.

Indic Arena will allow users nationwide to anonymously test, rank, and benchmark AI models, creating a neutral framework for assessing multilingual AI capabilities. Google Cloud is providing cloud credits to power the platform’s computational needs.

“A critical part of building AI for India is having a neutral, standardised benchmark to understand how models perform across our many languages,” said Mitesh Khapra, Associate Professor at IIT Madras. “Indic Arena will be that platform.”

Empowering Indian Innovation

Google’s expanded compute capacity and sovereign-ready Vertex AI platform are designed to support startups, universities, enterprises, and government agencies in developing the next generation of AI applications — built by Indians, for Indians.

The initiative underscores Google’s long-term commitment to strengthening India’s digital infrastructure, fostering local innovation, and enabling the country to emerge as a global leader in responsible AI development.