The minister explained that the AI ecosystem consists of 5 layers: the application layer, the models layer, the semiconductor or chip layer, infrastructure such as data centers, and the energy layer.
India’s approach to 5 layer AI ecosystem:
- Application and usage layer:This layer offers the highest Return on Investment (RoI). India is aiming to lead in applying AI to enterprise workflows and public services.
- E.g. Kisan e-Mitra is an AI-powered chatbot that helps farmers, Bhashini provides AI-based translation in 20+ Indian languages for accessing digital services.
- Model Layer: Instead of building only massive models, India is developing around 12 focused AI models, which can run on small GPU clusters and deliver AI services at low cost to a very large population.
- Nearly 95 %of AI workloads today are handled by small models, and that a 50-billion parameter model is sufficient for most enterprise requirements.
- Semiconductor: India is focusing on indigenous custom silicon development to reduce dependency.
- India is focusing on mastering manufacturing in the 28nm to 90nm range, covering applications such as electric vehicles, automobiles, railways etc.
- AI Infrastructure: Approx. USD 70 billion of AI infrastructure investment is already confirmed and being rolled out.
- Energy Readiness: India is integrating green energy to sustainably power the country's growing data centre footprint.
- E.g. India has opened nuclear energy to private sector participation through the Shakti Act, which will support the full AI stack.
