Artificial intelligence needs enormous amounts of electricity, and the chips that run it produce intense heat. Supplying reliable power and removing heat efficiently have become two of the biggest challenges, and costs, in building AI data centers. This article explains why, as of October 2026.
Why AI uses so much power
AI chips perform trillions of calculations every second, and racks of AI servers draw many times more power than traditional server racks. Training large models can require thousands of chips running continuously for weeks, and serving AI to millions of users runs around the clock. As a result, AI data centers are planned in tens or hundreds of megawatts, and electricity has become a key factor in where they can be built.
Where the power comes from
Data centers connect to the electricity grid and keep backup generators and batteries (UPS) for uninterrupted supply. To control costs and emissions, operators increasingly sign long-term contracts for solar and wind power, and some explore other sources, including nuclear. Rising data center demand is prompting investment in power generation, transmission lines and grid upgrades in many countries.
Why cooling matters
Nearly all electricity used by computers turns into heat. If heat is not removed, chips slow down or fail. Cooling can consume a significant share of a data center's total energy, so efficient cooling directly lowers costs.
Cooling methods
Air cooling uses fans, raised floors and air conditioning, the traditional method, but it struggles with today's dense AI racks. Liquid cooling carries heat away far more efficiently. In direct-to-chip cooling, liquid flows through plates attached to the hottest components. In immersion cooling, servers are submerged in special non-conductive fluids. Evaporative cooling uses water to cool air, which saves energy but uses water. Many new AI facilities are designed for liquid cooling from the start.
Measuring efficiency
PUE (Power Usage Effectiveness) compares total facility power with power used by computing equipment; closer to 1.0 means less energy is spent on cooling and other overheads. WUE (Water Usage Effectiveness) measures water use. Operators also report renewable energy share and carbon emissions.
The Indian context
In India, high temperatures and humidity in many regions make cooling more demanding, and water availability is a concern in several cities. Reliable power supply and renewable energy access are major factors in data center location decisions. Growing data center demand is one driver of investment in India's power, renewable energy and transmission sectors, alongside government targets for clean energy expansion.
Environmental debate
AI's energy and water use has raised concerns about emissions and local resource strain. Supporters point to efficiency gains, renewable energy commitments and AI's potential to optimise energy systems. Critics worry that growth in demand could outpace clean supply. This balance is an active area of policy discussion.
Why it matters for investors
Power and cooling needs link AI to many industries: electricity generation and transmission, renewable energy, electrical equipment, cooling systems, batteries and construction. These can benefit from AI demand, but they are also affected by regulation, project delays, input costs and the pace of AI investment.
Common misconceptions
"AI is just software." It depends on vast physical infrastructure.
"Cooling is a minor cost." It can be a major share of data center energy.
"Renewables solve the problem easily." Matching round-the-clock demand with clean supply is challenging.
Frequently asked questions
Why do AI chips get so hot?
Almost all the electricity they use becomes heat.
What is liquid cooling?
Using liquid instead of air to remove heat, far more efficiently.
What is PUE?
Total facility power divided by computing power; closer to 1.0 is better.
Do data centers use water?
Some cooling methods do; others are designed to use little water.
How PositoAI fits in
Research companies in power, renewable energy and electrical equipment on PositoAI, with company pages, red flags and financial history. Read our guides to data centers and AI and the economy.
Summary
AI needs huge amounts of electricity, and its chips produce intense heat. Data centers rely on grid power, backups and growing renewable contracts, and are shifting from air to liquid cooling. Efficiency is measured by PUE and WUE. In India, heat, water and power supply shape where data centers are built. Power and cooling link AI to many physical industries.