Can India keep up with AI's soaring demand for infra, GPUs, and talent?

Business Standard

08-05-2025

As India races to stake its claim as a global artificial intelligence (AI) powerhouse, a new Deloitte report has cautioned that the country is staring at infrastructure gaps that could trip its ambitions.
According to the report titled Attracting AI Data Centre Infrastructure Investment in India, the country will require an additional 45–50 million square feet of data centre real estate and 40–45 Terawatt Hours (TWH) of incremental power by 2030 to meet AI-driven demand.
This is no minor obstacle. Despite generating 20 per cent of global data, India accounts for just 3 per cent of the world's data centre capacity currently. With generative AI (GenAI) models demanding massive computing power and storage, India's current infrastructure is underpowered and outpaced.
India's data centre growth promising, but may fall short
The report projected a near 44 per cent compound annual growth in India's data centre capacity over the next five years. But AI data centres are highly energy-intensive—training large language models can consume up to 500 megawatt-hours per model, equivalent to the monthly usage of 150 US homes.
India's dependence on non-renewable energy—still around 55 per cent—adds to sustainability concerns. The report calls for a national policy on renewable incentives, grid modernisation, and captive power infrastructure.
Regulatory delays and land hurdles stall data centre growth
While India offers cost advantages in land and labour, acquiring land for large-scale centres remains slow and complex. Deloitte recommends setting up Data Centre Economic Zones (DCEZs) and Data Centre Facilitation Units (DCFUs) with single-window clearance.
It also suggests classifying data centres as 'essential services' under the Essential Services Maintenance Act to ensure round-the-clock operations.
GPU scarcity threatens AI compute capacity in India
India's AI goals are also challenged by a severe GPU shortage—the computational backbone of GenAI. The US AI Diffusion Framework limits India's import of Nvidia H100-class GPUs to 50,000 units until 2027, which could throttle progress.
Deloitte urges GPU-as-a-Service (GPUaaS) platforms and government-to-government (G2G) dialogue to overcome this bottleneck.
AI workforce growing, but R&D and training still lag
While India's AI workforce is projected to hit 1.35 million by 2027, the report flags a shortage of skilled professionals and dedicated R&D institutions. Deloitte calls for AI-specific training programmes, academic partnerships, and visa incentives to address the gap.
Rural and Tier-2 areas also lag in fibre connectivity, suffering high latency and poor bandwidth. The report recommends relaxing dark fibre regulations and allowing custom network infrastructure.
India's AI push gains momentum with local data storage
India has attracted $29.6 billion in AI investments between 2013 and 2024—still dwarfed by the US and China—but the momentum is building. Over 1,300 AI companies and a rapidly growing GenAI startup ecosystem put India in a strong position—if it can bridge its infra gap.
One step forward came on May 8, when OpenAI announced Indian user data from ChatGPT Enterprise, Edu, and API platforms will be stored locally to align with data sovereignty rules.
Policy support and infra upgrades key to India's AI future
'For India to accelerate its AI capabilities and realise its potential, it is necessary to introduce enabling policies to support the sector,' said S Anjani Kumar, partner at Deloitte India.
He added that India must build AI-ready infrastructure, invest in vernacular datasets, improve talent pipelines, and strengthen research and development to stay competitive in the global AI race.