India's AI data centres may deploy up to 700,000 GPUs in 5 years: Report

A report by Avendus Capital estimates that India's data centre capacity will grow at a compound annual growth rate of 26% over the next five years, driven by AI infrastructure demand

Rising adoption of artificial intelligence (AI) could drive the deployment of 650,000-700,000 graphics processing units (GPUs) in India’s data centres over the next five years, creating a $23 billion investment opportunity, according to a report by Avendus Capital.

 

The report estimates that India’s data centre capacity will grow at a compound annual growth rate (CAGR) of 26 per cent over the next five years, driven by AI infrastructure demand, cloud adoption and rapid digitalisation.

 

According to the report, India’s built data centre capacity is expected to nearly triple from 1.6 gigawatts (GW) in 2025 to nearly 5 GW by 2030. Developers currently have an active pipeline of more than 3 GW, including nearly 1 GW of AI-focused data centre capacity, requiring total capital investment of about $25 billion over the next five years.

   

AI boom reshapes India’s data centre business

 

The report noted that traditional data centres were designed primarily for cloud storage and enterprise workloads. However, AI applications require GPU-intensive infrastructure, liquid cooling systems, higher rack density and significantly greater power consumption, fundamentally changing the economics and design of data centres.

 

One of the report’s key findings is the emergence of GPU infrastructure as a high-return segment within India’s data centre ecosystem. At current capital expenditure and pricing levels, large-scale GPU deployments could generate an equity internal rate of return (IRR) of 25.5 per cent, with payback periods of under three years.

 

The report projects India’s AI market to grow from $10 billion in 2024 to $131 billion by 2032, registering a CAGR of 39 per cent. This growth is expected to be driven by rising enterprise adoption and investments in domestic AI capabilities, including the development of indigenous large language models (LLMs).

 

It further noted that more than 38,000 GPUs have already been committed under the IndiaAI Mission, with over 22,000 allocated for AI workloads and expected to be deployed in the near term. This, the report said, is creating significant demand for AI-ready data centre infrastructure across the country.

 

Speaking about the rise of AI data centres, Vaibhav Garg, director, infrastructure and real assets investment banking at Avendus Capital and chief author of the report, said: “AI adoption is emerging as a significant catalyst for next-generation infrastructure investments in data centres, alongside sustained demand from cloud and digital workloads. This dual demand trajectory has already translated into $5 billion of transaction activity over the last three years, with backing from global institutional investors, infrastructure funds, and strategic operators.”

 

He added that public markets and strategic transactions are expected to play a major role in funding the sector’s growth, with three to four initial public offerings (IPOs) likely over the next three years.

 

Private market activity rises

 

The report also highlighted growing private market activity in the sector, noting that global data centre transactions are currently being executed at earnings before interest, tax, depreciation and amortisation (Ebitda) multiples of 20-30 times.

 

Real Estate Investment Trusts (REITs) and Infrastructure Investment Trusts (InvITs) are increasingly being explored as capital recycling structures because of the sector’s long-term contracts and stable cash-flow profile. Global investors such as Blackstone, KKR, Brookfield, Digital Realty, Warburg Pincus and Mubadala are backing Indian data centre platforms.

 

AI factories emerge

 

The report identified the emergence of “AI factories”, the gigawatt-scale compute campuses built specifically for the AI application lifecycle, including training, tuning and inference. Since these facilities require massive GPU clusters, they are often being planned in decentralised locations with abundant land, power and fibre connectivity.

 

Examples cited in the report include:

 

• Reliance Industries planning a 1 GW AI campus in Jamnagar
• Google planning a 1 GW AI hub in Visakhapatnam
• Tata Consultancy Services HyperVault targeting 1 GW of liquid-cooled capacity

 

Mumbai leads as Hyderabad and Chennai gain traction

 

While Mumbai currently accounts for nearly half of India’s total data centre capacity, the report said Hyderabad is emerging as a major AI hub, with Microsoft Azure and Amazon Web Services expanding aggressively in the city.

 

The report also highlighted that three new international subsea cables are expected to land in Chennai during 2026-27, a development that could significantly boost the city’s appeal for hyperscale and AI workloads.