Data Centres

A data centre is a large, secure building full of computers called servers, along with storage machines and network equipment. These machines store and process the data behind everything you do online: UPI payments, WhatsApp messages, YouTube videos, government portals and AI chatbots. When you "save something to the cloud", it is actually saved on a server in some data centre. In simple words, data centres are the factories and warehouses of the digital economy.

Why do we need them?

Every online action needs a computer somewhere to receive, store and answer it. One server cannot serve crores of people, so companies put thousands of servers together in one place with steady power, cooling and security. As more people use digital payments, streaming, cloud services and AI, the need for data centres rises very fast.

How does a data centre work?

Picture a giant, cold library where every shelf is a computer.

1. IT equipment: racks of servers, storage and network switches do the actual computing.
2. Power: electricity comes from the grid, with backup from batteries (UPS) and diesel generators, so the servers never switch off.
3. Cooling: servers produce a lot of heat, so they must be cooled all the time. This is done by air-conditioning, chilled water, or evaporative cooling towers that use water, or newer liquid cooling that brings cold liquid close to the chips.
4. Connectivity: high-speed fibre links connect the centre to the internet and, for international traffic, to undersea cables.

Types of data centres

• Enterprise: owned and used by one company for its own needs (for example, a bank).
• Colocation: a company rents space, power and cooling in a shared building and puts its own servers there.
• Hyperscale: huge facilities run by big cloud companies (like Amazon, Microsoft, Google), with tens of thousands of servers.
• Edge: small centres placed close to users to reduce delay, useful for things like video streaming and self-driving systems.
• AI data centres: built around GPUs (graphics processing units), special chips that train and run AI models. They use far more power per rack than ordinary servers and produce much more heat.

How is size measured?

Data centres are measured by the electricity their IT equipment can draw, in megawatts (MW) or gigawatts (GW), not by floor area. One GW equals 1,000 MW. Reliability is often described by the Uptime Institute's Tier I to Tier IV levels. Tier I is basic, with no backup paths. Tier IV is "fault tolerant", meaning the centre keeps running even if one part fails.

How do we measure their efficiency?

• Power Usage Effectiveness (PUE): total electricity used by the whole facility divided by the electricity used by IT equipment alone. A PUE of 1.2 means that for every 1 unit used by computers, 0.2 units go to cooling, lighting and losses. Closer to 1.0 is better. It was introduced by The Green Grid in 2007.
• Water Usage Effectiveness (WUE): litres of water used per kilowatt-hour (kWh) of IT electricity. Lower is better.

Why are they resource hungry?

Data centres need three scarce things together: steady electricity, a lot of water and land with good connectivity.

• Electricity: Globally, data centres used about 415 TWh in 2024, around 1.5% of the world's electricity, and the International Energy Agency (IEA) expects this to roughly double to about 945 TWh by 2030, mainly because of AI.
• In India, data centres used about 13 TWh in 2024, around 0.8% of the country's electricity demand; the Council on Energy, Environment and Water (CEEW) projects about 57 TWh (2.6%) by 2030.
• Water: Evaporative cooling consumes large amounts of fresh water, which is a serious concern in a water-stressed country like India. Water is also used indirectly at power plants that generate the electricity.

India's position and Indian examples

• Infrastructure status: On 11 October 2022, the Department of Economic Affairs added data centres with at least 5 MW of IT load to the Harmonised Master List of Infrastructure, under the "Communication" category. This helps them get long-term loans more cheaply. It had been announced in the Union Budget 2022-23.
• Draft Data Centre Policy, 2020: MeitY released a draft policy to make India a global data centre hub, simplify clearances and give the sector infrastructure status.
• Data localisation: The RBI's April 2018 rule requires payment data to be stored only in India. Such rules increase demand for domestic data centres. The Digital Personal Data Protection Act, 2023 also governs how personal data is handled.
• Leading states: Maharashtra (Mumbai, Navi Mumbai) and Tamil Nadu (Chennai) lead because of undersea cable landing points, power and demand. Gujarat, Uttar Pradesh, Telangana, Andhra Pradesh, Odisha and Madhya Pradesh are also competing for investment, often with state data centre policies offering cheap land, power and tax benefits.
• Scale (as of 2026): about 271 data centres (January 2026) and about 1.57 GW of capacity (August 2026), expected to reach 4.5 to 6.5 GW by 2030.

Commonly confused concepts

• Data centre vs cloud: The "cloud" is a service (renting computing or storage over the internet). The data centre is the physical building where the cloud actually runs.
• Data localisation vs data sovereignty: Localisation means data must be stored within a country. Sovereignty is the broader idea that data is subject to the laws of the country where it is collected.
• PUE vs WUE: PUE measures energy efficiency (electricity overhead). WUE measures water efficiency (litres per kWh).
• MW of IT load vs total power drawn: IT load is power for the servers alone; total power also includes cooling and losses. Total power = IT load × PUE.

Issues, criticism and the way forward

• Water stress: Many hubs, like Chennai, Hyderabad and Bengaluru, already face water shortages. Large evaporative-cooling plants can compete with farmers and households for groundwater or reservoir water.
• Power and emissions: Most of India's electricity still comes from coal. Unless data centres use renewable energy, their growth adds to carbon emissions and stresses the grid.
• Local opposition: Residents near proposed projects have objected to water use, noise, diesel generators and closeness to homes, schools and hospitals.
• Incentive race between states: States offer cheap land, power and tax breaks to attract projects, but the jobs created after construction are relatively few, so the net gain may be small.
• E-waste and land use: Servers are replaced every few years, creating electronic waste.
• Way forward: Experts suggest linking approvals to verified water and power availability, using treated wastewater and closed-loop or liquid cooling, buying renewable power, setting PUE and WUE standards, publishing resource-use data, and tying incentives to measurable outcomes like jobs and clean energy use.

Concepts to Know

• Server: A powerful computer that stores data and runs programs for many users at once, usually over the internet.
• GPU (Graphics Processing Unit): A chip that can do many calculations at the same time. It is the main hardware for training and running AI models.
• Terawatt-hour (TWh): A unit of energy. 1 TWh = 1 billion kWh (units). A household using 100 units a month uses 1,200 units a year.
• Evaporative cooling: Cooling by letting water evaporate, which takes heat away, like sweat cooling your body. Effective but consumes water.
• Undersea cable landing station: The coastal point where an international internet cable from the sea connects to a country's network.
• Harmonised Master List of Infrastructure: The official list of sectors the government treats as "infrastructure", which makes them eligible for easier, cheaper long-term finance.