PSA Paper on Democratising AI Infrastructure: Bridging the Compute and Data Gap in India
SDG 9: Industry, Innovation, and Infrastructure | SDG 10: Reduced Inequalities | SDG 17: Partnerships for the Goals
Office of the Principal Scientific Adviser (PSA) | Ministry of Electronics and Information Technology (MeitY)
The PSA White Paper on Democratising Access to AI Infrastructure (December 2025) presents a comprehensive strategy to move AI resources from concentrated urban hubs to a distributed, equitable national utility. India currently faces a structural paradox: while the nation generates nearly 20% of the world’s data, it hosts only 3% of global data center capacity. To address this, the paper proposes a “whole-of-ecosystem” shift that treats compute, datasets, and models as Digital Public Goods (DPGs).
The current state and future roadmap of India’s AI infrastructure include:
Physical Infrastructure Surge: India’s data center capacity is projected to grow from 960 MW to 9.2 GW by 2030. Regional dominance is currently led by Mumbai (25%), followed by Bengaluru (22%) and Hyderabad (22%), with emerging hubs in Chennai and Delhi NCR.
National Compute Resources: The IndiaAI Mission is operationalizing a national GPU pool featuring 38,000 GPUs and 1,050 TPUs. These are offered via a “Compute-as-a-Service” portal at subsidized rates of under Rs. 100/hour, compared to global market rates exceeding Rs. 200/hour.
Data Sovereignty & Repositories: The IndiaAIKosh platform serves as a central repository, having onboarded 5,722 datasets and 251 AI models across 20 critical sectors like healthcare and agriculture. Specialized initiatives like Bhashini host 350+ language models across 17+ Indian languages to ensure linguistic inclusivity.
High-Performance Computing (HPC): The National Supercomputing Mission has deployed over 40 petaflops of capacity. India’s flagship AI supercomputer, AIRAWAT, is now the fastest AI-dedicated machine in the country, supporting distributed research for drug discovery and climate modeling.
State-Level Digital Public Infrastructure (DPI): Telangana has launched India’s first state-led DPI for AI (TGDeX), aiming to create 2,000 AI-ready datasets by 2030 using a federated model that allows data sharing without central pooling.
Sustainability Mandates: With data centers expected to consume 3% of India’s total electricity by 2030, states like Maharashtra, Tamil Nadu, and Karnataka are mandating a minimum of 30% renewable energy for new facilities.
What are Digital Public Goods (DPGs) in AI? They are open-source software, open data, open AI models, and open standards that are designed to be universally accessible and reusable. By positioning compute clouds and high-quality datasets as DPGs, India aims to ensure that the core building blocks of AI are shared public utilities rather than proprietary assets, enabling innovators even in remote areas to train models without needing to own expensive physical hardware.
Policy Relevance
The PSA’s roadmap validates India’s transition to “Sovereign AI”, ensuring that domestic innovation is not dependent on foreign cloud providers or proprietary global datasets. The integration of AI with the existing DPI architecture—including the Geospatial Data Sharing Interface (GDI) and GODL-India—standardizes how startups and researchers access sensitive public-interest data securely. For policy stakeholders, the focus on public-private partnerships and regional edge facilities in cities like Jaipur and Coimbatore is critical to ensuring that AI adoption is not restricted to technology-mature sectors like pharmaceuticals, but extends to agriculture and education.
Follow the full report here: Democratising Access to AI Infrastructure