Sign In
By signing in or creating an account, you agree with Associated Broadcasting Company's Terms & Conditions and Privacy Policy.
By signing in or creating an account, you agree with Associated Broadcasting Company's Terms & Conditions and Privacy Policy.
New Delhi: The Office of the Principal Scientific Adviser to the Government of India has issued a White Paper describing a national vision to democratise the access to artificial intelligence infrastructure. The paper contends that with AI becoming the core of innovation, economic expansion, and government capacities, compute capacity, data sets, and AI models need to be widespread, cheap, and inclusive.
Currently, these essential resources are concentrated in some of the global companies and major cities. This concentration prevents the involvement of the startups, researchers, and institutions in the lower cities and regions. The White Paper advances the concept of AI infrastructure as a national resource much like other types of Digital Public Infrastructure (DPI) to bring innovation that is India-diversified in language, social, and economic terms.
AI infrastructures are the physical foundation of the current AI systems. It has data centres, high-performance computing clusters, and dedicated processors like GPUs, TPUs, and NPUs. Data centres are highly central to the system, as they store huge amounts of data and provide the computer needed to train and deploy AI models.
The article observes that strong data storage and processing power are key to developing an AI ecosystem. Absence of domestic compute capacity would lead to countries being highly dependent on external cloud providers with the risk of increased costs, sovereignty, and resilience.
India produces almost one-fifth of the globe's data but also contains merely three per cent of the global data centre capacity. The existing installed capacity is about 960 MW, which is projected to rise drastically to reach 9.2 GW by 2030. The largest hubs are Mumbai and Navi Mumbai with good connectivity and support of policies; next come Chennai, Bengaluru, Hyderabad, and the Delhi NCR region.
The National Supercomputing Mission has placed more than 40 petaflops of computing capacity on the academic and research side. There are already systems like PARAM Siddhi-AI and AIRAWAT used in applications like language processing, weather prediction and drug discovery.
The White Paper includes a number of digital programmes that are intended to expand AI access. In 2025, the Indian AIKosh was launched and currently contains thousands of datasets and hundreds of AI models in 20 different fields. Platforms like Bhashini are also creating large-scale datasets of Indian language models, and Bhashini offers a platform where, through APIs, it allows controlled sharing of spatial data.
The IndiaAI Compute Portal has also enabled access to computers. More than 38,000 GPUs and over 1,000 TPUs can be used by researchers and startups at subsidised rates, which are usually fewer than half the world market rate. Under this model, users are able to train and fine-tune models without the need to own physical infrastructure.
One of the fundamental suggestions of the White Paper is to implement a Digital Public Infrastructure approach to AI. The DPI model emphasises a single platform instead of building registries, metadata standards, access protocols, and consent frameworks on an interoperable, modular structure. The layers are designed to cut down on costs and enhance transparency and make access predictable by the smaller players.
Further sophisticated systems such as federated access of data and coordinated computer exchanges can be introduced over time. The article emphasises that it is a progressive method of balancing aspiration and reality as well as preserving privacy, safety, and responsibility.
Significant challenges are also identified by the White Paper. The growth of AI infrastructure will lead to energy and real estate consumption rates being extremely high, and sustainability will be a burning issue. The adoption is also lopsided in other sectors like agriculture, healthcare, and education, where data and compute access are yet to be opened up.
The cooperation of the state and the business, efficient data centres and increased governmental efforts will be necessary. The paper draws the conclusion that the democratisation of AI infrastructure cannot be based only on the ownership by the state. Rather, it needs equitable, transparent and reproducible government-good layers that enable innovators in India to create and deploy AI in a responsible way.
