13 Nov 2026
How Does the Power Grid Support AI?
AI runs in data centers that need large amounts of continuous, reliable electricity, and the power grid has become one of the main constraints on how quickly AI capacity can be built. The International Energy Agency estimates data centers used about 415 terawatt hours of electricity in 2024, about 1.5% of global consumption, and projects that figure could more than double by 2030. Delivering that power depends on generation, transmission, substations and transformers, many of which take years to build.
How much electricity do data centers use?
Estimates vary with methodology, but major studies point in the same direction.
Globally, the IEA estimates data center electricity consumption at about 415 terawatt hours in 2024, or about 1.5% of global electricity use, having grown about 12% per year over the prior five years. In its base case, consumption rises to about 945 terawatt hours by 2030, just under 3% of global electricity use.1
In the United States, a 2024 report by Lawrence Berkeley National Laboratory for the Department of Energy found that data centers used about 176 terawatt hours in 2023, or 4.4% of U.S. electricity, and projected 325 to 580 terawatt hours by 2028, representing 6.7% to 12% of U.S. consumption.2 The IEA estimates that the United States accounted for about 45% of global data center electricity use in 2024, and that data centers are on course to account for almost half of U.S. electricity demand growth through 2030.3
Context matters as well. The IEA projects that data centers account for roughly a tenth of global electricity demand growth to 2030, less than industrial motors, air conditioning or electric vehicles.3 AI is one of several sources of rising demand, alongside manufacturing, electrification of transportation and buildings, and population growth. Together, they mark a change for advanced economies where electricity demand had been largely flat for years.
Why is AI’s demand different from other loads?
Data center demand is concentrated. A single large AI campus can require as much power as a small city, and it often needs that power in one location, around the clock. Training clusters run continuously at high utilization. The IEA notes that data centers tend to cluster geographically, which can strain local grids even when national totals look manageable.3
Reliability requirements add to the challenge. Data centers need power that is available nearly all the time, with backup systems for interruptions. That favors locations with strong grid connections and dependable generation.
How does power reach a data center?
Electricity is generated at power plants and moves over high-voltage transmission lines to substations, where transformers reduce the voltage. From there, distribution systems carry it to customers. A large data center typically connects at high voltage through its own substation, then steps the power down further on site and distributes it to its servers. Our article on how data centers work describes the on-site power systems.
Each step requires physical equipment: generators, transmission lines, substations, transformers, switchgear and circuit breakers. Adding a large new load can require upgrades at every step.
Why are grid connections the bottleneck?
Connecting new generation and large loads to the grid is a slow process. Lawrence Berkeley National Laboratory’s annual Queued Up study found about 2,061 gigawatts of generation and storage capacity actively seeking interconnection to the U.S. transmission grid at the end of 2025. The median project built in 2025 took 61 months from interconnection request to commercial operation, compared with 36 months in 2015 and 22 months in 2008.4
Those timelines are often longer than the time needed to build a data center itself. As a result, access to power, rather than access to land or chips, increasingly determines where and when AI capacity comes online.
Why does equipment matter?
The grid depends on equipment with long manufacturing lead times. Wood Mackenzie’s survey for the second quarter of 2025 found average lead times of about 128 weeks for power transformers and about 144 weeks for generator step-up transformers, with switchgear averaging about 44 weeks. The firm estimated a supply deficit of about 30% for power transformers in the United States in 2025.5
Manufacturers have announced new capacity, but building factories, securing specialty materials such as grain-oriented electrical steel and training skilled workers takes time.
Where does the power for data centers come from?
Several sources are expected to contribute. In its analysis to 2035, the IEA projects that renewables meet about half of the growth in data center electricity demand, with natural gas and nuclear each contributing meaningful amounts. Natural gas plays a significant role in the United States, and the IEA expects the first small modular reactors to come online around 2030.3
Data center developers also pursue on-site generation, long-term power purchase agreements and arrangements to locate near existing power plants. Grid operators and utilities are exploring tools that get more out of existing infrastructure; the IEA estimates that grid-enhancing technologies could unlock up to 175 gigawatts of transmission capacity without new lines.3
How is policy responding?
Policymakers are focused on both speed and cost. In November 2026, the U.S. Senate voted 73 to 25 to advance the Bipartisan American Affordability and Jobs Act, which would speed approvals for power plants and transmission lines while requiring large data centers to pay more of the grid upgrades they require.6 States and utilities are adopting rules on how large loads pay for new infrastructure, partly in response to concerns about residential electricity bills.
Outside the United States, governments in the Middle East, Europe and Asia are funding domestic AI computing capacity, which requires power as well. The result is that data center capacity tends to go where power and permitting are available and where it is welcome. Our article on the AI supply chain describes these sovereign programs.
Which industries are involved?
Supporting AI’s power needs draws on a wide range of industries: power generation, including gas turbines, nuclear, renewables and storage; transmission and distribution equipment such as transformers, switchgear, cable and circuit breakers; engineering and construction firms; utilities and independent power producers; and electrical equipment inside data centers, including power distribution and cooling systems. Many of these businesses also serve broader electrification trends, which some describe as a long investment cycle; our article on supercycles in investing discusses how to evaluate such claims.
What are the risks?
Demand uncertainty. Projections span wide ranges. Efficiency gains in chips, cooling and software could reduce power needs, while faster adoption could raise them. The IEA itself publishes several scenarios.
Execution. Permitting, interconnection and equipment delays can postpone projects.
Regulation and cost allocation. Rules on who pays for grid upgrades, and how utilities earn returns, can change.
Financing. Generation and grid projects require substantial capital and are sensitive to interest rates.
Valuation. Companies tied to power demand may already reflect strong expectations. Investments involve risk, including possible loss of principal.
Frequently asked questions
How much of U.S. electricity do data centers use?
Lawrence Berkeley National Laboratory estimated that data centers used about 4.4% of U.S. electricity in 2023 and projected a range of 6.7% to 12% by 2028, depending on how quickly AI servers are deployed and how efficiently they run. Other studies use different methods and assumptions, so estimates vary.
Why does it take so long to connect to the grid?
New generation and large loads must go through interconnection studies to determine what upgrades the grid needs, followed by permitting, equipment procurement and construction. Lawrence Berkeley National Laboratory found the median project built in 2025 took 61 months from interconnection request to commercial operation. Large transformers and other equipment can take more than two years to deliver.
What is a power transformer, and why is there a shortage?
A power transformer changes electricity from one voltage to another, allowing power to move efficiently over long distances and then be stepped down for use. Demand from new generation, data centers and grid replacement has outpaced manufacturing capacity, and production depends on specialized steel and skilled labor, so lead times have stretched beyond two years.
Are data centers the main driver of electricity demand growth?
They are a major driver in some regions, including the United States, where the IEA expects data centers to account for almost half of demand growth through 2030. Globally, the IEA projects data centers account for about a tenth of demand growth, with industry, air conditioning and electric vehicles contributing more.
What sources of power are expected to supply data centers?
The IEA projects that renewables meet about half of the growth in data center electricity demand, with natural gas and nuclear also contributing significantly through 2035. In the United States, natural gas plays a large role. Many developers combine grid power with long-term contracts, on-site generation and battery storage.
- International Energy Agency, “Energy and AI” (April 2025), chapter “Energy demand from AI.”
- Lawrence Berkeley National Laboratory, “2024 United States Data Center Energy Usage Report,” prepared for the U.S. Department of Energy (December 2024).
- International Energy Agency, “Energy and AI” (April 2025), key findings.
- Lawrence Berkeley National Laboratory, “Queued Up: 2026 Edition, Characteristics of Power Plants Seeking Transmission Interconnection as of the End of 2025.”
- Wood Mackenzie, “Making the Connection: Meeting the Electric T&D Supply Chain Challenge” (September 2025), and Q2 2025 lead time survey, as reported by POWER Magazine.
- Reuters, U.S. Senate vote on the Bipartisan American Affordability and Jobs Act (November 2026); VistaShares, “The Deployment Gap” (November 2026).
This material is provided for educational purposes only. It is not investment, legal or tax advice, and it is not a recommendation or an offer to buy or sell any security. Information is drawn from sources believed to be reliable as of the date of publication, is not guaranteed as to accuracy or completeness, and may change without notice. Figures attributed to third parties are their estimates. Projections and forecasts are not guarantees of future results. Investors should consult their own financial, legal and tax professionals before making investment decisions. Investments involve risk, including possible loss of principal. Companies in a single industry or theme may be affected by common economic, regulatory or technological developments and may be more volatile than the broad market. References to companies, governments or programs are for illustration only.