WEMO 2025 (complet) - Flipbook - Page 8
W E M O 202 5
O U T LO O K
• The rapid growth in data centers’ electricity consumption
is triggered by AI. In 2024, data centers 20,21,22 contributed to
global and regional electricity demand, mainly driven by the
rapid expansion of Arti昀椀cial Intelligence (AI) but also by cloud
computing, streaming and cryptocurrency mining. Globally,
data centers consumed approximately 415 TWh of electricity,
accounting for about 1.5% of total electricity consumption.
The United States accounted for the largest share of global
data center electricity consumption in 2024 (45%), followed
by China (25%) and Europe (15%). While data centers remain
a small fraction of global electricity growth (less than 10%
of total demand increase), their concentrated nature strains
local grids, particularly in regions with high data center
density, such as Ireland where they account for 20% of total
electricity consumption. This is prompting the government
20https://www.iea.org/news/ai-is-set-to-drive-surging-electricity-demand-from-data-centreswhile-offering-the-potential-to-transform-how-the-energy-sector-works
21https://www.jefferies.com/insights/the-big-picture/
how-data-centers-are-shaping-the-future-of-energy-consumption/
22https://www.iea.org/reports/energy-and-ai/executive-summary
23https://www.datacenterdynamics.com/en/news/
ireland-plans-legislation-for-data-centers-to-build-and-run-own-power-lines-report/
In the United States, 24 data centers consumed around
4.4% of total electricity in 2023,25 rising to an estimated
6% in 2024. This surge was driven by AI, cryptocurrency
mining, and cloud services, with hyperscale data centers
in states like Virginia and Texas leading consumption.
The US Department of Energy projects that data center
demand could reach up to 12% of US electricity by 2028.
According to the IEA report, 26 data center electricity
consumption is set to more than double, reaching
approximately 945 T Wh by 2030 (slightly more
than Japan’s total electricity consumption today).
The energy-intensive nature of AI, with ChatGPT queries
requiring nearly 10 times the power of a Google
search (2.9 Wh vs. 0.3 Wh), has amplified data
center electricity demand. Additionally, streaming
platforms like Netflix and YouTube consume over 200
TWh annually, further boosting data center loads.
Moreover, Large Language Model (LLM) training requires
thousands of GPUs to operate in lockstep and perform the
24https://escholarship.org/uc/item/32d6m0d1
25https://www.energy.gov/articles/
doe-releases-new-report-evaluating-increase-electricity-demand-data-centers
26https://www.iea.org/reports/energy-and-ai/executive-summary
27 https://www.eenewseurope.com/en/nvidia-details-ai-power-supply-adds-innoscience-gan/
same computation on di昀昀erent data. This synchronization
results in power fluctuations at the grid level. Unlike
traditional data center workloads, where uncorrelated
tasks smooth out the load, AI workloads cause abrupt
transitions between idle and high-power demand. Those
transitions are likely to become increasingly abrupt as
more LLMs come online, and more di昀케cult for grids to
balance demand and supply under those conditions.
To address this, the new Nvidia AI system27 uses several
mechanisms across different operational phases,
WEMO 2025
The buildings sector, driven by air conditioning, accounted
for nearly 60% of demand growth (600 TWh), while industry
contributed 40% (400 TWh), and transport, boosted by electric
vehicle (EV) sales rising over 25%, added 8% to demand.
to draft a new “private wire” policy enabling private operators
to build and run direct power links to data centers. 23
7
Natural gas generation increased by 103 T Wh,
particularly in the US and Middle East, while coal added
149 TWh, with 90% of this rise in China and India.
