Announced data centre investment in India had passed $180 billion as of September, according to a report released this month. Operational capacity in the first half of 2026 was 1,789 MW, or roughly 1.8 GW. Neither number is wrong. They simply measure different things, and the distance between them is the story.
Most of the money is coming from abroad. About 61% of the equity behind current investments is foreign. When a global company announces an AI investment in India, then, the size of the cheque matters less than how much of it will be running, and by when.
Pledged is not built
Big announcements play out over years. On top of what already operates, 509 MW is under construction and 3,351 MW is planned. Another 15 GW is at an early stage. Of future commitments, only about 24% are linked to projects under construction or planned. The other 76% sits in early-stage projects and land being held for later.
Even the report’s more ambitious scenario is modest about this. It reaches 8 GW by 2030 only by assuming that 15% of the early-stage pipeline becomes ready for service, and the authors take care to call that an assumption, not a forecast.
Other industry counts point the same way. One puts commitments at about $173 billion since 2021, with only about 130 MW of new capacity added in the first half of 2026 and installed capacity expected to reach about 2 GW by the end of the year. The totals differ because each is counted differently, but every measure shows the same gap between money promised and megawatts running.
Five things to check
Start with land and a date. Google announced its $15 billion AI hub in Visakhapatnam in October 2025, with spending planned between 2026 and 2030. The hub will have three data centre campuses. Google acquired the land at the start of April this year and broke ground on April 28, at a ceremony held with AdaniConneX and Nxtra by Airtel. A groundbreaking does not mean anything is running yet. It does mean the project exists beyond a press release.
Then ask about power. Data centres are expected to account for 2.6% of India’s electricity demand by FY2030, up from 0.8% in FY2025. India has a power surplus nationally, but a surplus on paper does nothing for a site that cannot get a grid connection. When Meta agreed in June to lease capacity at a 168 MW AI-enabled data centre that Reliance will build in Jamnagar, it also contracted nearly 1 GW of new renewable energy through CleanMax and Fourth Partner Energy. That is what taking power seriously looks like.
Look for a customer. India’s pre-leased data centre capacity roughly doubled to about 1.15 GW in the first half of 2026, and hyperscalers accounted for more than 80% of absorption over the same period. The Meta deal shows why a signed tenant changes things. Reliance has committed to delivering the facility within two years, with an option to scale, and Meta will lease the capacity.
Check the design. AI-ready facilities have to cope with the power, cooling and connectivity needs of GPU clusters. Higher rack densities can mean stronger electrical distribution, liquid cooling and heavier structural provision. A building put up for ordinary cloud work cannot simply be switched over.
The roughly $200 billion that Commerce and Industry Minister Piyush Goyal said on October 26 India could attract over the next seven to eight years is a wider figure still. Totals of that kind span the whole data centre value chain, from land and construction to IT equipment, network expansion and power.
What comes next
Investors, customers and governments will increasingly judge these projects on megawatts delivered and deadlines met. Companies that report progress phase by phase will find it easier to be believed than those that keep repeating a large total.
States are in a similar position. As of September, eleven had dedicated data centre policies, so having one no longer marks a state out. The report names what will decide delivery: timely power connections, connectivity, predictable approvals and specialist skills. Companies that need AI compute in India would do well to plan around capacity that is built and available to lease, and to count anything still under construction as a bonus.
India’s AI infrastructure push is real, and global capital is behind a large part of it. But an announcement is where delivery begins, not where it ends. Two questions cut through most of the noise: how much is live today, and when is the next phase due?
Also read: Brookfield sees 6.5 GW of AI data-centre capacity coming online in India
Also read: AirTrunk to Invest $30 Billion in India by 2030 to Build 5 Gigawatts of Data Centre Capacity
Also read: Google Chose Andhra Pradesh. Now the World Is Watching
Read our analysis: Can India’s Data Centre Expansion Keep Pace With Its Power Requirements?
