Data centers drive 18% of new global power demand
Projections show cooling, factories, and electric transport will claim the majority of grid expansion through 2030 as physical bottlenecks mount.

Data centers will account for less than a fifth of new worldwide power demand over the rest of the decade, leaving the majority of grid expansion to manufacturing, air conditioning, and transportation.12
Projections from S&P Global indicated that data centers will represent 18% of net global electricity demand growth between 2024 and 2030.1 The remaining 82% will stem from cooling systems, industrial factories, and the electrification of vehicles and buildings.1 Separate research from the International Energy Agency found that data centers will drive roughly one-tenth of global power growth across the same span, trailing industrial motors, household air conditioning, and electric vehicles.2 Worldwide data center electricity consumption stood at 415 terawatt-hours in 2024, or 1.5% of global electricity use, and is projected to reach approximately 945 terawatt-hours by 2030.23
The concentration of computing growth differs across regions.2 In advanced economies, where electricity consumption remained flat for several decades, the International Energy Agency projected that data facilities will drive more than 20% of net power demand growth through 2030.24 In the United States, data centers are on track to absorb nearly half of all electricity demand growth over the decade.24 By 2030, American facilities will consume more electricity than domestic manufacturing for steel, aluminium, cement, and chemical goods combined.24

Sector investors maintain that expansion plans reflect durable commercial requirements. "If there's a single narrative in the market that is most starkly divorced from the fundamentals we're seeing on the ground, it's that the demand for AI infrastructure and the energy that supports it is in some way fragile," Connor Teskey, chief executive of Brookfield Asset Management, said during Climate Week meetings in New York.1 Capital expenditure by five large technology companies exceeded $400 billion in 2025, according to the International Energy Agency, with spending projected to increase by another 75% in 2026.56
The scale of individual computing projects has also altered utility procurement. "Large loads have been a catalyst to so much new innovation because we've never had the opportunity to think about adding net new clean power at a big scale," said Veery Maxwell, founding partner of clean technology investment firm Galvanize.1 Technology companies accounted for approximately 40% of all corporate renewable power purchase agreements signed worldwide in 2025, according to agency records.5
That procurement has expanded to advanced firm power sources. The pipeline of conditional offtake agreements between data center operators and small modular reactor nuclear projects expanded from 25 gigawatts at the end of 2024 to 45 gigawatts by early 2026.5 Hyperscalers have also emerged as financial backers for geothermal and battery storage systems, seeking continuous generation to match server loads that swing by more than 50% within a second.56 The agency estimated that data centers could install between 20 and 25 gigawatts of battery storage globally by 2030 to manage those sharp fluctuations.6

Engineering demands inside the facilities are escalating in tandem. Between 2020 and 2025, the power density of artificial intelligence server racks increased by 11 times.6 By 2027, an individual server rack the size of a household refrigerator is projected to concentrate peak power demand equivalent to 65 homes, requiring cooling systems capable of dissipating heat equivalent to 30 natural gas boilers.6 Because nearly half of US data center capacity sits in just five regional clusters, the resulting power loads place intense pressure on localized distribution networks.2
Severe physical constraints continue to challenge project timelines. Constructing high-voltage transmission lines requires four to eight years in advanced economies, while procurement wait times for critical equipment such as transformers and substation cables doubled between 2022 and 2025.2 Confronted by long interconnection queues, developers have pursued between 15 and 27 gigawatts of onsite natural gas generation in the United States. Satellite tracking from the International Energy Agency showed that roughly one-fifth of those gas-powered projects have commenced land clearing or construction.6 Yet equipment availability has constrained that alternative as well, as global orders for natural gas turbines climbed 70% in 2025.6
Those operational delays have produced public disruptions. Oracle declared force majeure on a massive US data center project under construction, illustrating the vulnerabilities of rapid project execution.1 At the same time, municipal authorities and residents have mounted resistance to data campus proposals over local electricity rate increases, noise pollution, and water use for server cooling systems.15
The regulatory and supply bottlenecks now visible in computing reflect the broader hurdles confronting the international power sector. As factories, heating systems, and transport fleets switch to electricity, grid operators must integrate loads far larger than those generated by server farms. Resolving interconnection delays, expanding manufacturing capacity for heavy electrical gear, and securing community consent will determine whether power grids can accommodate the wider transition.
What this rests on
24 sentences trace to 6 sources.
- 1 View: Enough about the data centers already See the source
- 2 Executive summary – Energy and AI – Analysis - IEA See the source
- 3 Energy demand from AI – Energy and AI – Analysis - IEA See the source
- 4 AI is set to drive surging electricity demand from data centres while offering the potential to transform how the energy sector works - News - IEA See the source
- 5 Data centre electricity use surged in 2025, even with tightening bottlenecks driving a scramble for solutions - News - IEA See the source
- 6 Executive summary – Key Questions on Energy and AI – Analysis - IEA See the source