Every time someone asks a chatbot a question, generates an image or runs an AI agent, computers in a data centre do the work. Those computers need electricity, and as AI use explodes, so does the power needed to run it. Electricity has become one of the biggest constraints on AI's growth. Here is why AI uses so much power, what the forecasts say and what it means for India.
Why AI uses so much electricity
AI models run on specialised chips, mainly graphics processing units, or GPUs, packed into data centres. These chips perform enormous numbers of calculations and draw a lot of power. They also generate heat, so data centres need cooling systems that use additional electricity and often water.
Power is used in two main phases.
Training. Building a large AI model involves running thousands of chips for weeks or months, consuming huge amounts of energy.
Inference. Every time people use a trained model, it consumes power. With hundreds of millions of users and AI built into more products, inference now accounts for a large and growing share of AI's energy use.
Some tasks use far more power than others. A short text answer uses relatively little energy, while generating images, especially video, and running long reasoning or agent tasks uses considerably more.
What the global forecasts say
The International Energy Agency, in its April 2025 report on energy and AI, estimated that data centres used about 415 terawatt-hours of electricity in 2024, around 1.5 percent of global electricity consumption. It projected that data centre electricity demand could more than double to around 945 terawatt-hours by 2030, with AI as the most important driver of that growth.
For perspective, 945 terawatt-hours is roughly comparable to the annual electricity consumption of a large industrialised country.
Major technology companies have reported rising energy use and, in some cases, rising emissions linked to data centre expansion, even as they invest in renewable energy.
The race for power
AI companies now compete not only for chips but also for electricity. Companies are signing long-term deals for solar, wind, nuclear and other power sources, building data centres near power plants and exploring new energy technologies. Some have revived interest in nuclear power to supply steady, low-carbon electricity. In some regions, grid capacity has become a bottleneck that delays new data centres.
What it means for India
India is becoming a major data centre hub. Global technology companies and Indian conglomerates have announced large data centre investments, driven by growing digital use, data localisation requirements and AI demand.
Estimates of India's data centre growth vary, but they all point sharply upward. McKinsey, as reported in 2025, projected India's data centre capacity rising from about 1.2 gigawatts in 2024 to about 4.5 gigawatts by 2030. Some commentators have suggested that AI-driven data centres could add tens of terawatt-hours of electricity demand annually by 2030.
This creates both challenges and opportunities.
Challenges
Grid capacity. Data centres need large amounts of reliable power, often concentrated in specific cities such as Mumbai, Chennai, Hyderabad and the National Capital Region. This can strain local grids.
Clean energy. India still relies heavily on coal for electricity. If data centres run mainly on fossil fuel power, AI growth could increase emissions.
Water. Many data centres use water for cooling, which is a concern in water-stressed cities.
Land and infrastructure. Data centres require land, transmission lines and backup power.
Opportunities
Renewable energy growth. India has ambitious renewable energy targets and abundant solar potential. Data centres can drive demand for solar, wind and storage through long-term power purchase agreements.
Investment and jobs. Data centres bring investment, construction and skilled jobs.
Digital sovereignty. Local data centres support India's AI ambitions, including the IndiaAI Mission's computing infrastructure.
Efficiency innovation. Indian engineers and companies can contribute to more efficient cooling, chips and software.
How AI could become more efficient
The picture is not only about rising demand. Several trends could slow growth:
- More efficient models. Smaller models, distillation, quantization and mixture-of-experts designs reduce energy per task. Very cheap, efficient models released in 2026 show how quickly efficiency is improving.
- Better chips designed specifically for AI.
- Improved cooling, including liquid cooling and water recycling.
- Smart siting of data centres in places with clean power and cooler climates.
AI can also help the energy sector by optimising power grids, forecasting renewable energy output and improving energy efficiency in buildings and industry.
What individuals can do
- Use smaller, faster AI models for simple tasks.
- Avoid generating content you do not need, especially heavy image and video generation.
- Support companies that are transparent about their energy use and clean energy commitments.
The bottom line
AI is driving a surge in electricity demand because training and running models requires vast computing power in energy-hungry data centres. The International Energy Agency projects that global data centre electricity use could more than double by 2030. For India, the data centre boom brings investment and digital capability, but also challenges for grids, clean energy and water. How quickly efficiency improves and how much clean power India builds will decide whether AI's growth strengthens or strains the country's energy future.
