SBSI · Strategic Insights
Analysis · Energy & Infrastructure · June 2026

Why AI Data Centres Are Measured in Watts, Not Bytes

How the world's newest industry borrowed the units of one that is two centuries old — and what that tells us about where power demand is heading.

BY SACHIN BAXI · JUNE 2026 · ~6 MIN READ
Strategic Insights
415 TWh
Global data centre electricity use, 2024
IEA, Energy and AI Report, April 2025
1.7–2 GW
India's total data centre capacity, 2026
Vestian Research, April 2026 · BusinessToday
945 TWh
Projected global data centre use by 2030
IEA, Energy and AI Report, April 2025
The shift

AI data centres no longer just store data — they continuously process, train models and run specialised chips that generate far more heat. That is why capacity is now measured in MW and GW, not in terabytes.

The cooling problem

Air cooling can no longer keep up. The shift to liquid cooling systems adds to the power load. The defining constraint is now the power supply to operate and cool the facility — not the data it holds.

India's scale challenge

India's data centre capacity in 2026 is about 1.7–2 GW, expected to reach 9–10 GW by 2030–32. But supply projections suggest only around 5 GW may be available by 2032. India has just 1.2 MW of capacity per million people, far behind the US (147 MW) and Singapore (167 MW).

The global picture

The US and China together account for about 70% of global data centre power use. Global consumption is on track to double to 945 TWh by 2030. India's share will grow sharply as AI adoption accelerates.

What this means

The government has asked power utilities to incorporate data centre demand in their capacity plans. With $30 billion of investments already announced in India, getting the energy equation right is no longer optional — it is the central challenge.

The Shift

From Bytes to Watts — What Changed

In a sign of a new age, an industry that did not exist even a decade ago is being measured in the units of an industry that is more than two centuries old.

Today's AI data centre industry is less than five years old, yet it is already attracting investments worth billions of dollars. Its capacity is not measured in bits and bytes, the traditional units of data. It is not MB, TB or PB. Instead, it is measured in units of electricity — MW and GW. A closer look explains why.

The data centres of the 1990s and 2000s were primarily designed for data storage and retrieval. Capacity was measured by the volume of data that could be stored on chips. Processing requirements were relatively modest, and simple air circulation was sufficient to keep systems cool.

As technology evolved, miniaturisation packed more chips into smaller spaces, increasing computing power and heat generation. Cooling systems became larger and more sophisticated, relying on bigger fans and extensive air-conditioning. For years, that was enough. It is no longer sufficient.

"The data itself is almost secondary. The defining constraint is now the power required to operate and cool the facility."

Computation Changes Everything

The data centres of the AI era are not merely storing data. They are continuously processing, analysing, retrieving and training AI models. The critical difference is computation. Instead of handling simple storage tasks, they run highly specialised Graphics Processing Units (GPUs) that execute billions of operations every second. The result is vastly higher heat generation.

Air cooling alone can no longer maintain required temperatures. More powerful and efficient cooling systems are needed. Increasingly, data centres are shifting to liquid cooling technologies, which themselves consume significant amounts of power. As a result, the defining constraint is no longer data storage capacity but the power required to operate and cool the facility. The data itself is almost secondary.

India's Challenge

India: A Supply Problem in the Making

Power demand for AI data centres has surged accordingly. India's total data centre power consumption in 2026 is around 1.7 GW, expected to rise to around 9–10 GW by 2030–32 as large-scale facilities come online, driven by AI demand, localisation requirements and government incentives.

That power will have to come from somewhere — fossil fuels, nuclear energy or renewables. Current projections suggest supply may not exceed about 5 GW by 2032. Recognising the challenge, the government has reportedly asked power utilities to incorporate future data centre demand into their generation and capacity expansion plans.

India also remains significantly below global averages in data centre capacity. The United States has around 147 MW of data centre capacity per million people, China 14 MW, Germany 32 MW and Singapore 167 MW. India stands at just 1.2 MW per million people. While these figures reflect differences in population, they also highlight the relative size of the data centre industry. India has roughly 270 data centres, compared with about 449 in China and more than 5,400 in the United States.

As AI adoption accelerates and India catches up, all these numbers are set to rise sharply.

India Data Centre Power Capacity — Growth Trajectory
INSTALLED & PROJECTED CAPACITY (GW) · 2022–2032
Sources: Vestian Research (April 2026) via BusinessToday · IEEFA · Outlook Business (April 2026) · Drishti IAS / Ministry of Electronics & IT
Global Perspective

Data Centre Power: India, US, Europe, China

The numbers below put India's position in sharp relief. The US and China together account for nearly 70% of global data centre electricity consumption today. India's share is a fraction — though it is growing faster than most.

Region / Country Electricity Use 2024 Share of Global Use Share of National Grid Projected 2030 Growth to 2030
United States 180 TWh ~44% ~4% of US electricity ~420–430 TWh +130% (IEA base case)
China 102 TWh ~25% ~1% of Chinese electricity ~275–280 TWh +170% (IEA base case)
Europe (incl. UK) ~62 TWh ~15% ~2–3% of European electricity ~105–110 TWh +70% (IEA base case)
India ~10–15 TWh <1% <1% of Indian electricity ~40–45 TWh ~3× by 2030 (est.)
Global Total 415 TWh 100% ~1.5% of global electricity ~945 TWh +128% (IEA base case)

Sources: IEA Energy and AI Report, April 2025 · European Parliamentary Research Service (EPRS), 2025 · Incorrys Power Consumption of Data Centers, November 2025 · Outlook Business, April 2026 · Lawrence Berkeley National Laboratory (LBNL), 2024 · Brookings Institution, April 2026. Note: TWh = terawatt-hours. Europe figures cover EU plus UK. India electricity use figure is an estimate based on installed capacity projections and average utilisation rates.

Global Data Centre Electricity — Share by Region, 2024
PERCENTAGE OF GLOBAL CONSUMPTION (415 TWh TOTAL) · IEA 2025
Source: IEA Energy and AI Report, April 2025 · Incorrys, November 2025
Capacity per Person

How Far Behind Is India?

Absolute numbers can be misleading when populations vary so widely. The per-capita comparison below is more telling: India has just 1.2 MW of data centre capacity for every million people. Singapore has 167 MW, the United States 147 MW. Even China, with its enormous population, is at 14 MW per million.

Data Centre Capacity per Million People (MW)
INSTALLED CAPACITY · APPROX. 2024–2025
Source: Sachin Baxi Strategic Insights, prior research · JLL Data Centre Report 2024 · Cushman & Wakefield Data Centre Market Overview 2024

This gap is why India's data centre investment pipeline is attracting so much attention. The combination of a large and fast-growing digital economy, rising AI workloads, and very low baseline capacity creates an unusual growth opportunity. Not surprisingly, measuring data centre capacity in units of power rather than units of data makes perfect sense.

Country / Region Capacity per Million People (MW) Number of Data Centres Approx. Installed Capacity
Singapore~167 MW~70+~1 GW
United States~147 MW~5,400+~50+ GW
Germany~32 MW~520+~2.7 GW
China~14 MW~449~20 GW
India~1.2 MW~270~1.7–2 GW (2026)

Sources: Sachin Baxi Strategic Insights, prior research · JLL Data Centre Report 2024 · Cushman & Wakefield Data Centre Market Overview 2024 · Vestian Research, April 2026 · EESI (Environmental and Energy Study Institute), March 2025 · Deccan Herald / EESI, 2025

What This Means

The Electricity Equation India Must Solve

As AI adoption accelerates and India catches up, all these numbers are set to rise sharply. Not surprisingly, measuring data centre capacity in units of power rather than units of data makes perfect sense.

The central challenge is no longer about land, or tech, or even capital. It is about electricity — where it will come from, whether it will arrive on time, and how much of it will be clean. With $30 billion in announced investments and a government target of $200 billion in data centre and AI infrastructure by 2030, the race to solve the power equation is already on.

SachinBaxi Strategic Insights. Purely for informational purposes. Not investment advice.