India's data centre capacity may reach 3-3.6 GW by 2030, hyperscalers to drive growth: Axis Capital
New Delhi, August 2
India's data centre capacity could rise to around 3-3.6 GW by 2030, with hyperscalers and emerging AI-focused cloud players expected to become the dominant drivers of new demand, according to a report based on expert assessment cited by Axis Capital. However, execution and supply-chain constraints could keep actual operational capacity below the 6-8 GW targets publicly announced by industry players.
The report estimated that hyperscalers could account for around 90 per cent of data centre demand by 2030, compared with the current 60 per cent share, as companies such as Amazon Web Services, Microsoft and Google expand capacity and emerging neo-cloud and large language model hosting requirements gather pace. Enterprise demand, meanwhile, is expected to remain broadly flat in absolute terms.
India's installed data centre capacity currently stands at around 1.3-1.4 GW. Third-party colocation facilities account for the largest share at 60-65 per cent, followed by hyperscaler captive or leased capacity at 25-30 per cent. Enterprise and edge data centres account for around 8-10 per cent and 2-3 per cent, respectively.
The report noted that publicly announced capacity targets of 6-8 GW by 2030 appear aggressive. A more realistic estimate is 3.4-3.6 GW of planned capacity by mid-2030, while actual operational capacity could be closer to 2.8 GW. This reflects challenges around execution, power availability, equipment supply and construction timelines.
The data centre expansion pipeline is nevertheless gaining momentum. Around 900 MW to 1.2 GW of capacity had already been awarded in the last six to seven months of calendar 2026, although an award does not immediately translate into operational capacity. The report said it typically takes at least 32-36 months to roll out even the first phase of a project.
Supply-chain bottlenecks could remain a key constraint. Hyperscaler-grade backup generation equipment from leading global suppliers is reportedly booked roughly two years ahead, while shortages of EPC contractors and manpower and power-sanction timelines of around 18 months could slow project execution.
Capital intensity is also rising as AI-enabled facilities require more expensive infrastructure. Traditional colocation builds cost around Rs 44-46.6 crore per MW, excluding IT hardware, while AI-enabled infrastructure costs around Rs 48-54 crore per MW, representing a 10-15 per cent premium.
India's data centre capacity could reach 7-8 GW by 2035 if currently announced projects are fully realised, pointing to significant long-term opportunities across power, cooling, backup infrastructure, construction and electrical equipment, the report said.
— ANI
Reader Comments
The supply chain constraints are real. My husband works in infrastructure procurement and backup generator lead times have doubled since last year. The 32-36 month timeline for even the first phase seems optimistic in some cases. Better to under-promise and over-deliver, ya!
This is excellent news for Indian engineering and construction companies. With Rs 44-54 crore per MW investment, imagine the job opportunities in tier-2 cities where these centres are coming up. But we need to ensure local talent gets trained for these high-tech roles rather than relying on foreign expertise.
Hyperscalers taking 90% share by 2030 is a bit concerning for Indian companies. We need our own local cloud players to grow alongside. Remember when we said Jio would disrupt telecom? We need that kind of disruption in the data centre space too!
Power sanction timelines of 18 months!? That's the real bottleneck. We need bureaucratic reforms at the state level. Every month of delay costs crores. The government should create a single-window clearance mechanism for critical digital infrastructure projects. Otherwise, we'll keep lagging behind China and Singapore.
Finally a realistic analysis! All these companies announcing 100 MW projects without having power connections is pure drama. The 2.8 GW operational capacity figure sounds much more plausible. Also, the cost difference for AI-enabled infrastructure shows why only the big players can afford this game 🤔
We welcome thoughtful discussions from our readers. Please keep comments respectful and on-topic.