India to Build 12 Focused AI Models on Small GPU Clusters for Low-Cost Scale

India’s minister outlines a five-layer AI ecosystem framework and states that India is developing around 12 focused AI models that run on small GPU clusters to provide low-cost AI services at scale. The briefing also lists focus areas across applications, semiconductor manufacturing, AI infrastructure investment, and energy readiness, including private sector participation in nuclear energy through the Shakti Act.

India’s 5-Layer AI Ecosystem Approach :

Dimension	Key Details
AI ecosystem layers	Comprises application layer; models layer; semiconductor or chip layer; infrastructure such as data centers; energy layer
Application and usage layer	The highest Return on Investment (RoI); applies to enterprise workflows and public services
Examples in application layer	Kisan e-Mitra provides for an AI-powered chatbot that helps farmers; Bhashini provides for AI-based translation in 20+ Indian languages for accessing digital services
Model layer	Provides for development of around 12 focused AI models that can run on small GPU clusters and deliver AI services at low cost to a very large population
AI workload distribution	Nearly 95% of AI workloads today are handled by small models
Enterprise model size	A 50-billion parameter model is sufficient for most enterprise requirements
Semiconductor focus	Indigenous custom silicon development to reduce dependency
Semiconductor manufacturing	Mastering manufacturing in the 28nm to 90nm range
Applications covered	Electric vehicles, automobiles, railways
Investment	Approx. USD 70 billion of AI infrastructure investment is already confirmed and being rolled out
Energy readiness	Integrating green energy to sustainably power the country’s growing data centre footprint
Shakti Act	Opening nuclear energy to private sector participation; supports the full AI stack