How to make AI data centres more sustainable
A data centre is a building, often a warehouse, that houses racks of computers. Those computers power services such as artificial intelligence. Their electricity use matters because growing demand can put pressure on power supplies and increase emissions if fossil fuels provide the extra energy. AI tasks require computers to process information, and more computing can mean more electricity demand. Data centres also need power to keep equipment operating. The article notes that grids face extra demand for cooling during hot weather, but it does not give a breakdown of electricity used for data-centre cooling. The scale is already significant: data centres consume more than 1.5% of the world’s electricity supply. Their use has grown rapidly, and the International Energy Agency expects the share to more than double in the next few years. The article argues that this growth should be met sustainably, rather than by burning more fossil fuels.
What is a data centre, and why does running AI in one use so much electricity?
A data centre is a building, often a warehouse, that houses racks of computers. Those computers power services such as artificial intelligence. Their electricity use matters because growing demand can put pressure on power supplies and increase emissions if fossil fuels provide the extra energy.
AI tasks require computers to process information, and more computing can mean more electricity demand. Data centres also need power to keep equipment operating. The article notes that grids face extra demand for cooling during hot weather, but it does not give a breakdown of electricity used for data-centre cooling.
The scale is already significant: data centres consume more than 1.5% of the world’s electricity supply. Their use has grown rapidly, and the International Energy Agency expects the share to more than double in the next few years. The article argues that this growth should be met sustainably, rather than by burning more fossil fuels.
How much electricity do data centres use today, and how much could their demand grow by 2030?
Data centres consume more than 1.5% of the world’s electricity supply today. Their rising demand matters because electricity must come from somewhere, and the source affects emissions and pressure on power systems. The International Energy Agency says the data-centre share of global electricity use will more than double in the next few years.
Growth has been rapid: global data-centre electricity consumption rose by around 12% per year between 2017 and 2024. In 2024, the United States accounted for 45% of demand, China 25%, and Europe 15%. These figures show how concentrated much of the current use is.
The IEA says that by 2030 data centres could consume more electricity than Japan. A separate Electric Power Research Institute projection says continued US growth could add up to 94 gigawatts of demand by 2030, equivalent to 17% of US electricity consumption. These are projections, not certainties, because they depend on assumptions that can change.
What can happen to emissions, air pollution and local resources if data centres’ extra power comes from fossil fuels?
The consequences depend on how the extra electricity is generated. If data-centre demand is met by burning fossil fuels, greenhouse-gas emissions can increase. Communities also worry that added demand for electricity and water could strain local resources, and that data centres may contribute to air pollution and health problems.
Recent research gives a measure of the emissions involved: around 400 US data centres produced an estimated 26 megatonnes of carbon dioxide between May 2024 and April 2025. Another study found that data centres’ energy demands could prolong the lifespan of coal-fired power stations. That links new computing demand not only to new emissions, but potentially to continued use of existing high-emitting power sources.
These outcomes are not inevitable. The article points to government requirements for renewable power and better use of existing grid capacity as ways to limit fossil-fuel demand. The article raises concerns about pollution and health, but does not quantify specific local health impacts or establish that every data centre causes them.
How could data centres use less power from the grid during peak-demand periods?
Electricity demand rises and falls during the day and across seasons. Data centres do not necessarily need to run every task at the busiest moment, so shifting some operations could reduce their call on the grid during peaks. This matters because meeting peak demand can require extra electricity generation.
The mechanism is to curtail data-centre operations when overall electricity use is high, then increase them when demand is lower. Researchers at Duke University analysed this approach in the United States. Their 2025 analysis found that the grid could accommodate 98 gigawatts of extra demand this way without adding new generation capacity.
That amount could cover most, if not all, of expected US data-centre expansion over the next few years. The approach depends on being able to move or delay some computing work, rather than requiring every task to run immediately. US grid operators have been instructed to plan for data-centre demand, while Axiom has announced consultations on working around power-demand peaks.
What alternatives could reduce the need for large cloud data centres?
Large cloud data centres are not the only way to run AI. The article points to models that can run on individual computers or on servers operated by institutions. These alternatives matter because they could reduce reliance on energy-intensive data centres and lower the electricity used for some computing.
The key difference is where the computation happens: instead of sending work to a large cloud service, a model can run on a local computer or an institution’s server. The article says these approaches consume much less power than models run using cloud services. It does not give exact energy comparisons or specify which AI tasks are suitable for each option.
The article argues that development of these alternatives needs better funding. That is one part of a wider effort to meet growing computing demand in smarter, more sustainable ways. The choice is not presented as a complete replacement for large data centres, but as an option that could reduce dependence on them.
How can governments and grid operators help ensure new data-centre power comes from renewable energy?
Governments and grid operators can influence whether new data-centre demand leads to more fossil-fuel use. The central challenge is to provide power without increasing emissions, especially as data-centre electricity needs rise. The article says this outcome is not inevitable and identifies policy and grid management as possible tools.
Governments could mandate that new energy demand be met with renewable sources. Regulators could also examine different ways to manage existing electricity supplies, including using excess grid capacity and shifting flexible demand away from busy periods. In June, the US Federal Energy Regulatory Commission instructed grid operators to draw up a plan for handling extra demand from data centres.
The technology industry is also engaging with the issue. Axiom announced a program to consult US grid operators, regulators and policymakers about working around peaks in power demand. These steps show that planning is under way, but the article does not say that renewable-power mandates have already been adopted or guarantee how future demand will be supplied.
How does an electricity grid balance the power people need with the power generators can supply?
An electricity grid connects power supplies with the people and businesses using electricity. In general, grid operators work to keep supply and demand aligned so electricity remains available as use changes. The source article does not explain the technical controls in detail; this basic description uses established knowledge about how electricity grids operate.
Demand can surge, for example, when people need more cooling on hot days or heating during winter cold snaps. The article says many national grids are designed to accommodate such surges, and that they have excess capacity much of the time. That spare capacity could help serve data centres without requiring as much new energy.
Another option is to shift flexible data-centre operations away from periods of peak use and increase them when demand is lower. A 2025 Duke analysis found the US grid could accommodate 98 gigawatts of extra demand this way without adding generation capacity. Grid planning and demand management can therefore affect how much new power is needed.
This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.
Read more in the JupiteX app
Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.
Or read more news on the web