JupiteX Get the app
Environment & Climate20 Nov 2026 · about 7 min

AI data center power demand will double by 2030 thanks to $4 trillion in spending, adding an entire Great Britain to energy strain

The brief

The expected addition is 120 gigawatts (GW) of data-center capacity by the end of the decade. The International Energy Agency (IEA) estimates that 75% of this new capacity will be dedicated to artificial intelligence. Capacity describes how much power facilities could draw, not necessarily how much they use at every moment. That 75% share works out to roughly 90GW for AI data centers. The IEA also reports that 120GW of data-center capacity is already available worldwide across all types of data centers. The expected build-out would therefore match that existing capacity in size. The scale matters because electricity grids in many regions are already being pressed by proposed data-center projects. The IEA is also tracking 93 large AI factories under construction or expansion. Grid connections and power supply may be limiting factors as these projects move ahead, while the IEA expects spending on AI data-center infrastructure to continue growing.

01

How much additional data-center capacity does the article expect to be built by 2030, and what share is expected to serve AI?

The expected addition is 120 gigawatts (GW) of data-center capacity by the end of the decade. The International Energy Agency (IEA) estimates that 75% of this new capacity will be dedicated to artificial intelligence. Capacity describes how much power facilities could draw, not necessarily how much they use at every moment.

That 75% share works out to roughly 90GW for AI data centers. The IEA also reports that 120GW of data-center capacity is already available worldwide across all types of data centers. The expected build-out would therefore match that existing capacity in size.

The scale matters because electricity grids in many regions are already being pressed by proposed data-center projects. The IEA is also tracking 93 large AI factories under construction or expansion. Grid connections and power supply may be limiting factors as these projects move ahead, while the IEA expects spending on AI data-center infrastructure to continue growing.

02

How much power could the 93 large AI factories use when completed, and how does that compare with Great Britain's power demand?

The 93 large AI factories being built or expanded could have a combined capacity of 50 gigawatts (GW) when complete. The article describes that scale as the equivalent of adding a second Great Britain to power demand. It is a striking measure of how quickly large AI sites could add to electricity needs.

Satellite imagery helps the International Energy Agency track these projects. It estimates that their total capacity grew from under 2GW to 17GW in just two years, with 50GW projected upon completion. These figures describe capacity at the sites, rather than guaranteeing that they will draw that much electricity constantly.

The projected demand matters because grids in many regions are already facing data-center projects that could exceed the systems’ current peak load. The IEA also expects another 120GW of data-center capacity across all types to be completed by the decade’s end. Power supply and grid upgrades will shape how these large projects can be served.

03

What is an AI factory, and why does training and running AI require so much electricity?

The article calls 93 large sites “AI factories” and tracks them as they are built or expanded. It does not give a formal definition of the term. In context, it means large data-center projects designed to support artificial intelligence, rather than a conventional factory making physical products.

In general, training AI involves running many calculations across computer chips, often in large groups of servers. Once a model is trained, answering user requests also requires computing power. Servers produce heat, so data centers may need cooling systems as well; those systems can use additional electricity. These technical details are general background, not explained in the article.

The IEA estimates that the 93 sites’ capacity rose from under 2GW to 17GW in two years and could reach 50GW when complete. Separately, it expects 75% of 120GW in new data-center capacity by the decade’s end to be devoted to AI. That scale helps explain why electricity supply and grid access have become major constraints.

04

When can a new data center raise household electricity bills, and when might it avoid doing so?

A new data center does not automatically mean higher household electricity bills, according to the IEA. The effect depends partly on whether the facility can use capacity already available on the grid or whether the system needs costly upgrades to serve it.

If a data center uses existing capacity and pays its share of the costs, the IEA says it could potentially lead to lower bills for all customers. But if the grid operator must upgrade its system to accommodate the large new customer, bills could rise when those costs are spread across all users. The key mechanism is how upgrade costs are assigned.

The IEA says its analysis does not yet show a clear relationship between data-center capacity growth and electricity-price increases. That means the outcome may differ between places and projects. As more facilities are built, the way utilities plan upgrades and allocate their costs will matter to both data-center operators and household customers.

05

Why can data centers sign renewable-power agreements but still get much of their electricity from coal and natural gas?

A power purchase agreement is a contract to buy electricity, but the source named in a contract does not necessarily match the mix of electricity reaching a data center at every moment. The article reports that renewable energy makes up nearly 50% of data centers’ power purchase agreements, while coal and natural gas are closer to 60% of actual power generation.

The key reason the article gives is reliability. Renewable sources may not always deliver the power a data center needs, so coal- and natural-gas generation can step in. Some projects are also planning or building their own power supplies. The article cites about 180GW planned or under construction in the U.S., including Elon Musk’s Colossus 2 in Mississippi.

Renewable energy is growing to accommodate demand, but the IEA expects coal- and natural-gas emissions from data centers to rise by 70% by 2030. So agreements to buy renewable power do not, by themselves, mean the data center’s actual electricity supply is entirely renewable at all times.

06

Why can building transmission lines take much longer than building a power plant or a data center?

The article gives a striking timeline difference: renewable power plants take one to five years to complete, while transmission work can take three to five times longer. It does not detail the reasons for the delay. The practical effect is clear, though: a new power source does not solve a data center’s needs if electricity cannot be carried to it.

As general background, transmission lines often need long routes across multiple properties and communities. Projects can require land access, environmental reviews, permits, and coordination among different authorities. Those steps can be complex and take time before construction begins. These are well-established reasons such infrastructure can move slowly, but they are not itemized in the article.

The IEA identifies the grid as a significant bottleneck for AI data-center growth. Some large data-center projects can also take years, and hyperscalers are turning to on-site jet engines, though those face shortages too, especially in turbine blades. Slow transmission expansion can therefore make it harder to match new facilities with power.

07

What does an electricity grid do, and why must it balance electricity supply and demand in real time?

An electricity grid is the connected system that moves electricity from power plants and other generators to homes, businesses, and large users such as data centers. It includes transmission lines that carry power over long distances and distribution networks that deliver it locally. The article does not define the grid, so this is general background.

Electricity has to be produced as it is needed because supply and demand on the system must stay balanced in real time. If demand rises, generators or other available sources need to provide enough power; if supply is insufficient, the system can become unstable. This balancing is a basic feature of operating an electricity network.

That requirement matters as AI data centers add large loads. The IEA says projects in many regions could exceed the current peak load their systems can support, and calls the grid a significant bottleneck. Building more generation is not enough if transmission cannot carry power where it is needed, making grid capacity a central issue for future data-center growth.

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