Google Cloud Expands AI Infra in India, Supports IIT Madras

• Google Cloud boosts AI infrastructure in India with Trillium TPUs and Hypercomputer.
• Partners with IIT Madras to launch Indic LLM-Arena for Indian language AI evaluation.
• Offers Gemini model access with full data residency and advanced localization features.

Google Cloud revealed significant AI investments in India, featuring increased local computing power driven by Trillium TPUs and backing for a new community-based platform to assess AI models for Indian languages and contexts.

The technology powerhouse is greatly boosting AI hardware capabilities in India with its AI Hypercomputer framework, allowing companies and government entities to train and implement sophisticated Gemini models locally. This expansion tackles data residency and sovereignty needs while lowering latency for AI applications.

Google Cloud is offering Indian customers early access to its most advanced Gemini models, providing complete data residency support, a first for the company. The rollout features Gemini 2.5 Flash with batch processing functions, Document AI for streamlining document workflows, and grounding capabilities utilizing Google Maps for location-based replies.

In collaboration with IIT Madras, Google Cloud is aiding the initiation of Indic LLM-Arena, a standalone evaluation platform managed by the AI4Bharat center. The platform employs crowd-sourced, blind evaluations where users provide prompts in any Indian language including code-mixed inquiries such as Hinglish and cast votes on replies from anonymous AI systems.

Also Read: IIT Delhi, Simplilearn, Microsoft Launch AI Courses

The initiative tackles significant shortcomings in existing AI assessment, primarily centered on English and Western environments. Indic LLM-Arena assesses models across three areas: linguistic variety encompassing code-switching, cultural relevance to India's regional contexts, and safety measures for India-related issues such as communal misinformation and caste-related stereotypes.

The platform now accommodates text inputs and aims to extend to visual, auditory, and interactive tasks related to documents and online searches. Google Cloud is offering cloud credits to support the infrastructure.