Everyone Is Talking About AI. But the Real Money Is Behind It.

India’s datacentre capacity is set to grow at 30% annually through 2030. The companies supplying the industrial equipment to build them are sitting on multi-year order backlogs — and the market has barely noticed.

There is a version of the AI investment thesis that most people know — semiconductors, hyperscalers, software platforms. And then there is the version that Nomura has just published a detailed report on — the industrial supply chain that physically makes AI infrastructure possible. Transformers, switchgear, cooling systems, UPS units, backup generators, structured cabling. Unglamorous products.

Extraordinary demand. And in India specifically, a market structure that has created what Nomura calls “an enviable seller’s market” with delivery lead times of two to four years and pricing power that most industrial businesses would envy. The bank’s top picks are CG Power and GE Vernova T&D — both rated BUY.

India’s Datacentre Boom 

The numbers behind India’s datacentre industry are striking. IT load capacity has grown from approximately 350 megawatts in 2019 to 1.5-1.6 gigawatts in 2025 — a compound annual growth rate of approximately 29%, comfortably ahead of the global average of 20%. India’s share of global datacentre capacity has grown from 1.5% to 2-3% over the same period. And Nomura’s analysis of announced pipelines suggests visibility on over 15 gigawatts of incremental capacity over the next decade, with India expected to reach approximately 7 gigawatts of total capacity by 2030 — a 30% CAGR from here.

The demand drivers are layered and mutually reinforcing. Rising mobile data consumption, enterprise digitisation, cloud and SaaS adoption, and the step-change in compute intensity that generative AI requires are all pulling in the same direction. India’s data traffic on a per-megawatt basis is approximately 18 petabytes per MW — compared to China’s 4-5 — a gap that quantifies how significantly underpowered India’s datacentre infrastructure is relative to the data being generated. “AI is emerging as a force multiplier for DC demand,” the Nomura report states, as GPU-heavy, AI-optimised datacentres require materially more power and cooling per square metre than traditional facilities.

Geopolitics is adding a tailwind that was not in the original thesis. As hyperscalers diversify away from single-country deployment concentrations, India is emerging as an alternative hub — benefiting from its democratic governance, English-speaking technical workforce and improving power infrastructure.

Why India’s Datacentres 

One concern that frequently surfaces around datacentre investment in emerging markets is whether the economics are viable without the mature power and real estate infrastructure of developed markets. Nomura addresses this directly. Construction costs in India run at USD 6-7 million per megawatt — compared to USD 10-18 million across developed APAC and Western markets. Electricity sourcing through open access arrangements, renewable power purchase agreements and captive power setups costs approximately USD 7-8 cents per kilowatt hour.

These cost advantages, Nomura argues, more than offset the discount to developed market co-location rentals — resulting in stabilised datacentre assets capable of generating “infrastructure-like annuity cash flows and mid-teen equity IRRs.”

The Industrial Supply Chain 

Here is the core of the Nomura thesis — and it is genuinely differentiated from how most investors are thinking about the AI datacentre opportunity in India. The companies that own and operate datacentres — controlling approximately 80% of India’s co-location market — are mostly unlisted or in the process of listing. Direct investment in the datacentre operators is largely not available to public market investors today.

What is available is the industrial supply chain that builds these facilities. Nomura estimates that five product categories together absorb 60-75% of a datacentre’s total capital expenditure budget of USD 10-22 million per megawatt: medium and low-voltage switchgear and transformers; UPS and battery systems; backup diesel and gas generators; precision and liquid cooling distribution units; and rack, busway and structured cabling infrastructure. The companies supplying these products — in a market that Nomura describes as “consolidated and pricing-disciplined” — are the ones capturing the economic value of the boom.

The competitive landscape is oligopolistic. ABB India, Siemens, Hitachi Energy India, GE Vernova T&D, CG Power, Cummins and a handful of others together hold above 40% combined market share in each sub-category. And crucially, delivery lead times of two to four years have created a seller’s market where today’s order bookings are converting into revenue for 2027-29. “We think premium pricing is being captured because DC projects demand higher reliability, tighter customization, accelerated lead times, certifications, and on-site engineering support that traditional commercial or industrial projects do not require,” the report states. Nomura believes this premium pricing environment is likely to sustain for at least the next three to four years.

Of the companies in the datacentre supply chain, Nomura’s top picks are CG Power and GE Vernova T&D — both rated BUY.

CG Power, which supplies power electronics, transformers and switchgear, is seen as one of the biggest beneficiaries given its product breadth across multiple datacentre sub-categories and its strong market positioning. GE Vernova T&D, which focuses on grid infrastructure and transmission and distribution equipment, brings a similar combination of product relevance and market share.