Crisil sees India data centre capacity growing fivefold by 2030

India's installed data centre capacity is set to grow fivefold to roughly 7.5 GW by March 2030, according to Crisil Ratings, which expects the industry to absorb more than INR3 trillion (about US$31.72 billion) of investment over the same period.

The Indian rating agency calls the build-out an investment 'supercycle' in a report published this week. Three forces drive it: generative AI-led computing demand, enterprise digitalisation, and rising data consumption. Nearly half of the upcoming capacity is already committed under long-term contracts, which Crisil says limits speculative development and reduces offtake risk.

What underpins the forecast?

Demand visibility is the report's central claim. Upfront tie-ups with hyperscalers, driven by cloud adoption and compute-intensive gen AI workloads, provide revenue certainty from the outset. That structure de-risks projects even as the scale and speed of the build-out raise execution demands.

Cost is the second pillar. Anand Kulkarni, Director at Crisil Ratings, said: "India's structural cost advantages further strengthen its attractiveness as a data centre destination. Development costs are around one-fifth lower than in several competing Asia-Pacific markets owing to more competitive land, construction and labour costs. Combined with lower power tariffs, this translates into rentals that are 30-35% lower than in peer markets. The ability to develop large, renewable-powered campuses also aligns well with the sustainability goals of hyperscalers, reinforcing India's competitive positioning."

In plainer terms: a developer building in India spends about 20% less to construct a facility than in comparable Asia-Pacific markets, and tenants pay roughly a third less in rent. Whether those margins hold as competition for land, power and skilled labour intensifies during a fivefold capacity ramp is a question the report does not answer.

Where are the risks?

Crisil is careful to separate the strong demand story from the conditions attached to it. The pace of expansion and the credit quality of data centre developers will hinge on three factors: timely power connectivity, regulatory approvals, and disciplined project execution.

The report flags four specific pressure points:

• Power infrastructure bottleneck. The unprecedented growth in development is straining electricity supply, and developers' fortunes depend on adequate, timely power availability.
• Customer concentration. Hyperscaler dominance creates concentration and pricing risk for developers reliant on a small pool of very large tenants.
• Elevated debt. Investments typically precede cash generation by 24-36 months, so leverage will stay high over the medium term. A developer can carry two to three years of spending before revenue starts to offset it.
• Demand sustainability. Crisil conditions its optimism on the durability of gen AI-led demand growth itself, a demand curve that no rating agency can yet verify.

The agency concludes that timely commissioning, cost control, adequate power availability and the sustainability of gen AI-led demand will determine how successfully the industry navigates the cycle.

What is not in the report?

Two gaps stand out. The report acknowledges power supply as a bottleneck but does not engage with the environmental and water-resource concerns that accompany such rapid growth — data centres are heavy water users for cooling, and public opposition has surfaced in other fast-growing markets. That omission matters for a forecast premised on regulatory approvals arriving on schedule.

Second, the fivefold capacity figure rests heavily on hyperscaler commitments, the same customers Crisil identifies as a pricing risk. If a handful of cloud providers control both the demand and the price, the revenue certainty they provide cuts both ways.

The investment cycle, in Crisil's framing, runs to March 2030; whether the power grid, approval processes and gen AI demand all keep pace over those five years is the test the report sets without guaranteeing the outcome.