5x in Five Years: The Construction Wave Hiding Inside India’s Data Centre Boom

India’s data centre capacity is projected to grow from 1.7 GW at end-2025 to 8–10 GW by 2030, a roughly fivefold build-out compressed into five years, backed by a $180B investment pipeline. For steel buyers and traders, the headline isn’t the digital economy. It’s a concentrated, capital-locked construction cycle landing in a handful of metros, one whose pace will be set less by demand than by power, water, talent, and regulation.

The build-out is real, funded, and front-loaded

Three numbers anchor the case. Capacity stands at 1.7 GW today, with roughly 30% growth expected in 2026 alone (~2.2 GW). The investment pipeline is projected at $180B by 2026. And the 2030 target of 8–10 GW implies a build rate of well over 1 GW of new capacity per year, every year, through the decade’s end.

Each gigawatt of data centre capacity is a construction project first and a compute asset second: heavy foundations, long-span structural frames, raised flooring systems, racking, cable management, chiller plants, plus the substations and transmission connections that feed it. The mechanism matters more than any single tonnage estimate: capacity targets of this scale translate directly into multi-year demand for rebar, structural sections, and fabricated steel, clustered around Mumbai, Chennai, Hyderabad, Pune, and Delhi NCR rather than spread evenly across the country.

Implication: regional buyers in these clusters should expect data centre and allied power projects to compete for the same fabrication capacity, logistics, and material that general construction draws on, and to do so on compressed timelines.

The cost advantage means capital stays committed

India builds data centre capacity at $6–7M per MW, against $12–15M in Singapore and $13–16M in Japan, a 40–50% cost advantage. That gap does two things. It explains why the pipeline is this large. And it makes the pipeline sticky: hyperscalers and operators choosing India on cost economics don’t reverse those decisions on a soft quarter.

Implication: treat this as structural, not cyclical, demand. A cost-driven capex wave with committed capital behaves differently from speculative construction, projects slip, but they rarely die. Plan for a demand floor, not a boom-bust.

The constraints make demand lumpy, not linear

Here’s where the naïve extrapolation fails. Four bottlenecks sit between 1.7 GW and 10 GW:

Power. Data centres run 24/7 and cannot throttle. Their share of national electricity demand could rise from 0.5% to ~3% by 2030, which means grid and transmission build-out (itself steel-intensive) becomes a gating factor for site timelines.

Water. A 1 MW facility consumes ~70,000 litres daily; sector-wide consumption is projected to rise from 150 billion litres (2025) to 358 billion litres (2030), in a country holding 18% of global population but 4% of global freshwater. Expect water availability to shape site selection and permitting speed.

Talent and regulation. Shortages in cooling, cybersecurity, and network operations slow commissioning; regulatory uncertainty (including the powers of the yet-to-be-constituted Data Protection Board) injects timing risk into investment decisions.

Implication: steel demand from this sector will arrive in surges tied to project clearances and grid connections, not as a smooth curve. Buyers who track project milestones (not announcements) will see the demand before the market prices it.

The honest limit

Public, India-specific data on steel intensity per MW of data centre capacity is thin, and this analysis deliberately avoids inventing a tonnage figure. Directionally, the demand pull is real; sized against India’s total steel consumption, data centres alone will not move national HRC or rebar benchmarks. Their significance is regional and sectoral (fabrication capacity, structural sections, and metro-cluster logistics) not macro. Anyone quoting a precise “million tonnes of steel from data centres” number should be asked for their source.

What to watch

• Grid connection approvals in Mumbai, Chennai, and Hyderabad clusters, the truest leading indicator of construction starts.
• The 2026 growth print: does capacity actually reach ~2.2 GW? A miss signals the constraints are binding earlier than expected.
• Water-linked permitting decisions in stressed regions, a proxy for which announced projects convert to steel orders.
• Structural section and fabrication lead times in the five metro clusters, versus national averages.
• Data Protection Board constitution and early rulings, the regulatory trigger that could accelerate or stall committed capital.

Frequently asked questions

How much will India’s data centre capacity grow by 2030?

From 1.7 GW at end-2025 to a projected 8–10 GW by 2030, roughly fivefold, backed by a $180B investment pipeline.

Why is data centre construction cheaper in India?

Build costs run $6–7M per MW versus $12–15M in Singapore and $13–16M in Japan, a 40–50% advantage from lower land and labour costs, tax incentives, and large domestic demand.

What could slow the build-out?

Power (share of national electricity rising from 0.5% to ~3% by 2030), water (150 → 358 billion litres annually by 2030), specialist talent shortages, and regulatory uncertainty.

What does this mean for steel demand?

Concentrated, multi-year demand for rebar, structural sections, and fabricated steel in the Mumbai, Chennai, Hyderabad, Pune, and Delhi NCR clusters, arriving in surges tied to clearances and grid connections, not a smooth curve.

Close

The next 6–12 months will show whether India’s data centre pipeline converts to poured concrete and erected steel at the promised pace, or queues behind its own grid. For buyers, the edge lies in tracking conversion, not announcements; platforms like SteelMath exist to turn exactly that kind of forward signal into procurement timing.