India has taken a decisive step forward in its artificial intelligence strategy, expanding its national AI infrastructure and endorsing indigenous foundation models aimed at serving India-specific needs. At the ‘IndiaAI – Make AI in India, Make AI Work for India’ event held in New Delhi, Union Minister for Electronics and Information Technology Shri Ashwini Vaishnaw announced the addition of 15,916 GPUs, bringing India’s total compute capacity to 34,333 GPUs.
The announcement aligns with Prime Minister Narendra Modi’s vision of democratizing technology. “Technology should not be left in the hands of a few,” Vaishnaw emphasized, reiterating the government’s focus on equitable access, innovation, and reversing brain drain. The event also celebrated the onboarding of 367 datasets to the AI Kosh platform and the growth of compute infrastructure to support AI models trained on Indian data.
Three Startups to Lead India’s Foundation Model Race
Following a rigorous expert evaluation, three startups were selected under the IndiaAI Foundation Model pillar:
- Soket AI: Will build India’s first open-source 120 billion parameter foundation model, addressing the country’s linguistic diversity and focusing on defense, healthcare, and education.
- Gnani AI: Will construct a 14 billion parameter multilingual Voice AI model capable of real-time, speech-enabled services with advanced reasoning capabilities.
- Gan AI: Will develop a 70 billion parameter multilingual model with “Superhuman” text-to-speech (TTS) abilities aiming to compete with global leaders in TTS.
These projects complement the previously announced Sarvam AI, which is building India’s Sovereign LLM with open-source 120B parameters for citizen services under initiatives like “2047: Citizen Connect” and “AI4Pragati.”
GPU Price Discovery and Cyber AI Innovation
The government also published the commercial bids for AI compute usage, categorized by GPU type and usage durations (on-demand, monthly, 6-month, and 12-month slots). CyFuture India Pvt. Ltd. emerged as the lowest (L1) bidder across most categories, offering highly competitive pricing across AMD, Intel Gaudi, and NVIDIA H100 GPUs. Other notable bidders included Vensysco Technologies, Yotta Data Services, Netmagic, Sify Digital, and Locuz Enterprise Solutions.
Meanwhile, under the IndiaAI I4C CyberGuard AI Hackathon, four winning teams showcased AI-based tools for classifying cybercrime complaints and identifying criminal patterns on the National Cyber Crime Reporting Portal (NCRP). These models demonstrated the ability to process complex inputs such as FIRs, screenshots, and audio callswith high accuracy—signaling India’s intent to integrate AI into core law enforcement operations.
Conclusion: A Holistic National Strategy
India’s AI mission now operates with a unified vision: indigenous technology, scaled compute, public-sector AI applications, and a firm ethical foundation. By combining foundational research, local data, affordable compute access, and industry participation, India is crafting a globally relevant yet locally rooted AI ecosystem.