India's Frugal AI Model Offers Alternative to Energy-Intensive Global Approach

Training GPT-4 required over 50 gigawatt-hours of electricity, and global data centres now account for roughly 1.5 per cent of total electricity consumption — a share expected to double by 2030 as AI workloads expand. A public policy analyst argues that India's Digital Public Infrastructure model, which processed over 16 billion UPI transactions in a single month in 2024 at a fraction of Western infrastructure costs, offers a blueprint for ecologically responsible AI governance.

The article, presented in the context of the Raisina Dialogue's Tomorrowland sessions, notes that a single large language model query consumes several times more energy than a standard web search. In 2023, Google reported consuming over 6 billion gallons of water to cool its data centres. India's IndiaAI Mission has committed Rs 10,372 crore toward building compute capacity including shared GPU infrastructure. If built to current global standards, this infrastructure will carry the same energy and water intensity as equivalent facilities elsewhere.

India's existing Digital Public Infrastructure — Aadhaar, UPI and ONDC — was designed under constraint for bandwidth-limited, energy-uncertain conditions, achieving financial inclusion at a fraction of comparable Western deployment costs. Lightweight AI models from institutions including IIT Madras and IISc Bangalore already match larger models on Indian language tasks without equivalent energy overhead. French company Mistral demonstrated that a 7-billion parameter model with careful fine-tuning can outperform much larger models on specific tasks. The author argues this frugal design philosophy constitutes a governance asset India should deploy more deliberately.

The article identifies gaps in India's regulatory response: the National Data Governance Framework and Digital Personal Data Protection Act address data flows but are silent on physical resource implications. No mandatory environmental impact assessment framework exists for AI infrastructure, nor any water use reporting or energy source disclosure obligation tied to AI procurement. The author recommends mandatory energy consumption and carbon intensity reporting for government-procured AI systems, efficiency thresholds in procurement standards, and siting policies that require renewable energy access and adequate water resources.

Key figure — Rs 10,372 crore committed under India's IndiaAI Mission for compute capacity