India is creating a financial ecosystem through AI boom: Avendus

India’s artificial intelligence boom is beginning to create a new financial ecosystem around infrastructure such as data centres and computing power.

A new report by Avendus estimates that India’s data centre developers have more than 3GW of capacity lined up over the next five years, requiring investments of nearly $25 billion. But alongside the race to build AI infrastructure, another trend is quietly gathering pace: investors, lenders and infrastructure funds are increasingly treating GPUs and AI-ready data centres as long-term financial assets.

According to the report, GPU deployments can generate equity internal rates of return exceeding 28%, while payback periods can be as low as 2.8 years. That has started drawing in private equity firms, sovereign investors, infrastructure funds, and lenders looking to capitalise on the AI buildout. The report points out that globally, companies such as CoreWeave, Crusoe and Lambda Labs have already raised billions of dollars through debt and equity deals tied to GPU infrastructure. In several cases, GPUs themselves are now being used as collateral for loans.

The report noted that private market activity around data centres is also accelerating globally, with transactions in the sector currently taking place at EBITDA multiples of 20-30x. It added that structures such as REITs and InvITs are increasingly being explored for capital recycling because of the industry’s long-term contracts and predictable cash flows.

Avendus estimates that the domestic data centre industry has witnessed more than $5 billion in transaction activity over the past three years. The deal flow has ranged from private equity infusions and acquisitions to AI-focused infrastructure partnerships.

Among the transactions highlighted in the report, telecom-backed data centre operator Nxtra raised $1 billion in March 2026 to expand capacity and widen its services portfolio. AI cloud startup Neysa secured $600 million earlier this year to build an AI platform with plans to deploy 20,000 GPUs. HyperVault, the AI data centre business launched by TCS, received a commitment of up to $1 billion from TPG.

The report also states that domestic firms such as Yotta, E2E Networks, NxtGen and Neysa are positioning themselves as AI cloud providers rather than conventional data centre operators. Yotta, for instance, has emerged as one of the largest beneficiaries under the IndiaAI Mission’s GPU allocation programme, while E2E Networks has already deployed thousands of Nvidia GPUs and procured Blackwell systems for future expansion, according to the report.

Public markets are also noticing the opportunity as well. Several players, including Sify, ESDS and Yotta, are either preparing listings or scaling AI-focused infrastructure businesses. 

India’s AI ambitions drive demand for GPU and data centre capacity

The investor interest is closely tied to expectations around the scale of India’s AI expansion over the next few years.

Avendus estimates India’s AI market will grow from roughly $13 billion in 2025 to $131 billion by 2032, implying a compound annual growth rate of around 39%. The report attributed the growth to rising enterprise adoption, investments in domestic AI capabilities and efforts to develop indigenous large language models.

More than 80% of Indian firms are already prioritising AI adoption across operations, while the IndiaAI Mission has so far received commitments for over 38,000 GPUs. Of these, around 22,000 GPUs are expected to be allocated to AI workloads in the near term.

The report said deployment of these GPU clusters alone would require nearly 48MW of AI-ready data centre capacity, adding to demand for high-density and liquid-cooled infrastructure.

India’s installed data centre capacity is expected to rise from 1.6GW in 2025 to 5GW by 2030, translating into a 26% compound annual growth rate. Over the same period, the report estimates deployment of 650,000 to 700,000 GPUs across the country, requiring nearly $23 billion in investment.

The expansion is also becoming increasingly concentrated around a few major hubs.

Mumbai currently accounts for nearly half of India’s installed data centre capacity and is expected to attract around 47% of the additional capacity likely to come up in the country over the next five years, according to the report.

The city presently has an installed capacity of about 801 MW, well ahead of Chennai at 268 MW, Delhi-NCR at 161 MW and Hyderabad at 138 MW.

The report also suggests that the next phase of AI infrastructure demand may come from inference workloads, where trained AI systems continuously process user queries and enterprise tasks, and are expected to drive a larger share of AI cloud demand over time. Inference is the action phase of AI which highlights processes that the model goes through when presented with new information to make decisions or generate answers. That matters because inference workloads require persistent compute infrastructure, creating a longer runway for data centre and GPU utilisation.