Emerging evidence
More
Data center power demand may hit up to 1,050 TWh in 2026, review of 18 sources finds
Study title: Impacts of data center expansion on global energy demand: Sustainability challenges in the age of artificial intelligence
- Evidence type
- Commentary, opinion, or narrative review
- Sample
- n = 18
- Sources
- 1 sourcePeer-reviewed
- Published
- 3 min read
- Truvace record
- TRV-2026-2599cite or check it
Study title: Impacts of data center expansion on global energy demand: Sustainability challenges in the age of artificial intelligence
A review of 18 technical reports, standards and regulatory documents published between 2019 and 2025 concludes that electricity demand from data centers, AI and cryptocurrencies could reach between 620 and 1,050 TWh in 2026. The International Energy Agency reference scenario is about 800 TWh, against roughly 460 TWh recorded in 2022, an increase of 160 to 590 TWh in four years.
Read the study abstract
Digital transformation and advances in artificial intelligence have driven the expansion of data centers and increased their share of global electricity demand. This article analyzes the impacts of this growth on power systems and discusses the main challenges to the sector's sustainability. The study adopts a qualitative, exploratory, and descriptive approach and was conducted through a bibliographic and documentary review of 18 sources published between 2019 and 2025. Technical reports, standards, regulatory a…

The problem
A review of 18 reports and standards found that data centers, AI and crypto could use 620 to 1,050 TWh of electricity in 2026 (up from about 460 TWh in 2022), and that they can be built faster than grids can expand.
In brief
The authors argue the strain comes less from the global total than from rapid expansion, geographic concentration, round-the-clock supply needs and mismatched building timelines. More than 8,000 data centers operate worldwide, about 33% in the United States, 16% in Europe and 10% in China. In Ireland, data centers took about 17% of national electricity demand in 2022, and projections put that near 32% in 2026.
The figures are projections drawn from existing reports, not new measurements. Some companies do not disclose installed capacity, consumption or cooling efficiency, so estimates carry significant uncertainty. The authors say sustainability will depend on locational and energy planning, efficiency, data transparency, coordinated grid expansion and low-carbon sources.
Main points
- US data center electricity use rose from about 58 TWh in 2014 to 176 TWh in 2023, and Lawrence Berkeley National Laboratory projects 325 to 580 TWh in 2028.
- The US share of national electricity use was about 4.4% in 2023 and may reach between 6.7% and 12% in 2028.
- China's data centers may use about 300 TWh in 2026 and more than 400 TWh by 2030; EU data centers used about 100 TWh in 2022 and may reach about 150 TWh in 2026.
- Brazil's Ministry of Mines and Energy projects 2.5 GW of data center load by 2037 from new projects in São Paulo, Rio Grande do Sul and Ceará alone.
- AI chip energy efficiency has doubled roughly every two and a half to three years, but growth in devices, users and applications may outpace those gains.
- In some regions, system operators and authorities have restricted or suspended approval of new data centers because of inadequate electricity infrastructure.
- Emergency diesel generators used for redundancy add to emissions, and some cooling systems use large volumes of water in water-stressed regions.
The full story
The authors used a qualitative, exploratory and descriptive method, reviewing technical reports, institutional publications, regulatory acts and efficiency standards in Portuguese and English. The sources included the International Energy Agency, Brazil's Ministry of Mines and Energy, Ireland's Commission for Regulation of Utilities, Lawrence Berkeley National Laboratory, ASHRAE, FGV Energia and the European Parliament.
The regional comparison shows different pressure points. The United States and China see the largest absolute growth. Ireland faces risk from the high share of data centers in national demand. The EU is moving on transparency and efficiency rules, including Directive 2023/1791 on disclosure of energy use, water use and emissions. Brazil must plan grid access and reinforcements before loads connect.
On infrastructure, a data center can be built in two to three years while new generation and transmission face longer planning, licensing and financing. The authors say this can create connection queues, regional constraints and reinforcement costs. Data centers also need continuous supply, which makes solar and wind harder to rely on without storage or complementary generation.
The authors say efficiency alone will not cut total demand if processing volumes grow faster. They describe nuclear power as a firm low-carbon option with high upfront costs and long deployment times, and small modular reactors as an emerging option that still depends on technological, regulatory and economic advances.
Sources
1 on file- World Journal of Advanced Engineering Technology and Sciencesdoi:10.30574/wjaets.2026.20.1.0397
Talk it over
Who carries this cost, and what would have to change to prevent it?
Data center power demand may hit up to 1,050 TWh in 2026, review of 18 sources findshttps://truvace.com/item/impacts-of-data-center-expansion-on-global-energy-demand-sustainability-challeng
Who carries this cost, and what would have to change to prevent it?
The problem A review of 18 reports and standards found that data centers, AI and crypto could use 620 to 1,050 TWh of electricity in 2026 (up from about 460 TWh in 2022), and that they can be built faster than grids can expand.
Answer in the threadHow do we know? What can this kind of evidence show, and what can’t it?
The evidence This rests on commentary or opinion with a sample of 18, from World Journal of Advanced Engineering Technology and Sciences.
Answer in the threadWhat does it leave open? What else would you want to know before acting on it?
The record’s own caveat The findings rest on a review of 18 existing documents and on projections; the authors note that some companies do not disclose detailed consumption data, so estimates carry significant uncertainty.
Answer in the threadHave you seen AI used this way where you study, work or live? Did it go the same way?
Answer in the thread
The questions are the same on every record. The details under them are taken from this record as written.
ace
What readers say