In News:

India has taken a major step toward self-reliance in Artificial Intelligence with the expansion of its national AI compute infrastructure and the selection of three new startups to build indigenous foundation models under the IndiaAI Mission.

Key Highlights

• Compute Infrastructure Boost:India’s total GPU capacity has now surpassed 34,000 units, up from the initial 10,000-target. A fresh addition of 15,916 GPUs to the existing 18,417 empanelled GPUs brings the total to 34,333 GPUs, now available through the IndiaAI Compute Portal (operational since March 2025).
• Subsidised Access:These GPUs are made available at a subsidised rate of ?67/hour, well below the global average of ?115/hour. This has been made possible through private sector empanelment instead of government-built data centres. Service providers receive up to 40% capital subsidy, enabling rapid infrastructure rollout.
• Empanelled Providers:Seven private companies were empanelled for compute provisioning:
  • Cyfuture India Pvt. Ltd.
  • Ishan Infotech Ltd.
  • Locuz Enterprise Solutions Ltd.
  • Netmagic IT Services Pvt. Ltd.
  • Sify Digital Services Ltd.
  • Vensysco Technologies Ltd.
  • Yotta Data Services Pvt. Ltd.

Foundation Model Development

Under the IndiaAI Foundation Model initiative, three new startups have joined Sarvam AI (selected earlier in April 2025) to build India-specific Large Language Models (LLMs):

• Soket AI: Will develop a 120-billion parameter open-source model focused on Indian languages and use cases in defence, healthcare, and education.
• Gnani AI: Building a 14-billion parameter Voice AI model for real-time, multilingual speech recognition and reasoning.
• Gan AI: Developing a 70-billion parameter multilingual TTS (text-to-speech) model aiming for "superhuman" capabilities surpassing global benchmarks.
• Sarvam AI: Previously selected to create a 120-billion parameter Sovereign AI model, following the release of Sarvam-1 (2B parameters) and Sarvam-M (24B parameters).

These foundation models will be trained on Indian datasets and tailored for governance, public service delivery, and regional language support.

AI Kosh& Innovation Initiatives

• AI Kosh: A public dataset platform with 367 datasets uploaded, enabling research and model training using India-relevant data.
• IndiaAI I4C CyberGuard Hackathon: In collaboration with the Ministry of Home Affairs, AI models were developed for identifying cybercrime patterns from complex inputs like handwritten FIRs and audio calls on the National Cyber Crime Reporting Portal.
• Startup Innovation & Skill Development: Funding support, AI labs in Tier-II cities, and talent development programs are part of a broader push to promote innovation and reverse brain drain.

About IndiaAI Mission

• Launched by: Ministry of Electronics & Information Technology (MeitY)
• Cabinet Approval: March 2024 with a budget of over ?10,000 crore
• Objectives:
  • Develop indigenous AI capabilities and infrastructure
  • Democratize AI access for governance, startups, and citizens
  • Promote ethical and safe AI use
  • Position India among the global AI leaders