India’s GPU Expansion and the Quest for Sovereign AI At the India AI Impact Summit in New Delhi, Union Minister for Electronics and Information Technology Ashwini Vaishnaw announced the addition of 20,000 GPUs to the existing 38,000‑unit cluster, with a long‑term target of 2 lakh GPUs . The move underscores the centrality of computational power in modern governance, defence and economic productivity. Key Developments Immediate boost of 20,000 GPUs to the national cluster. Long‑term goal of 2 lakh GPUs to stay competitive with global AI leaders. Implementation through the IndiaAI Mission and its seven pillars. Partnerships such as L&T‑Nvidia for large data‑centre development. Budget provision of tax holidays for data centres achieving optimal PUE . Why GPUs Matter for AI A GPU can perform parallel matrix multiplications, the core operation of neural networks and LLMs . Unlike the sequential processing of a CPU, GPUs accelerate training and inference, making them indispensable for modern AI applications. Geopolitics of Compute The AI era has created a new hierarchy where access to high‑end silicon is as strategic as oil or rare earths. Over 90% of cutting‑edge GPUs are designed by Nvidia (US) and fabricated in Taiwan, exposing the ecosystem to supply‑chain risks and export‑control policies. The US has imposed restrictions on advanced GPU exports to curb China’s AI capabilities, prompting nations like India to pursue sovereign AI as a strategic hedge. IndiaAI Mission – Seven Pillars IndiaAI Compute Pillar : Subsidised GPU access at Rs 65 per hour for startups and researchers. Application Development Initiative : Supports AI solutions for India‑specific challenges. AIKosh Dataset Platform : Consolidates government and private data for model training. Foundation Models : Development of Indian LLMs in regional languages; example – Sarvam AI . FutureSkills : AI‑skill development from undergraduate to PhD level, with AI labs in educational institutions. Startup Financing : Funding under the IndiaAI Startups Global programme (launched March 2025). Safe and Trusted AI : Emphasis on bias mitigation, privacy, explainability and governance. Challenges in Building a Sovereign AI Ecosystem Scaling to lakhs of GPUs raises environmental and infrastructural concerns. A single high‑end GPU rack consumes >100 kW, equivalent to the peak demand of 80‑100 Indian households, stressing urban grids. Data‑centres also need large volumes of water for cooling, a critical issue for a water‑stressed nation. Moreover, high‑quality, structured datasets are scarce, and a significant share of Indian user data is processed abroad. Human capital is another bottleneck. Building and maintaining AI models requires specialised talent, many of whom migrate to the US or Europe for better remuneration and research facilities, creating a brain‑drain. Way Forward – Holistic Chip‑to‑Grid Strategy India must develop a self‑sufficient ecosystem covering silicon design, GPU manufacturing, renewable‑energy‑powered data‑centres and water‑efficient cooling technologies. The ISM 2.0 can foster indigenous GPU IP, reducing reliance on Nvidia and US supply chains. Policy steps include: Extending tax incentives for data‑centres achieving low PUE and using renewable energy. Funding R&D for water‑less cooling solutions. Strengthening AI talent pipelines through scholarships, research grants and industry‑academia collaborations. Ensuring data localisation and creation of high‑quality, Indian‑language datasets. By addressing these challenges, India can transform its GPU expansion from a hardware tally into a robust foundation for a sovereign AI future, enhancing national security, economic growth and global competitiveness.