Seen and Unseen Load: Open-Data Visibility of AI Data-Centre Electricity Demand and Wind–Solar Headroom in China, Japan and South Korea

The debate over whether renewable energy can sustain AI growth relies on demand figures whose origin is rarely examined. We compare open, facility-level records with agency estimates for China, Japan, South Korea and the United States, using the Epoch AI GPU-cluster register (338 clusters), the International Energy Agency, Lawrence Berkeley National Laboratory, South Korea's 11th Basic Plan for Electricity Supply and Demand, and Ember generation data. In 2024, visible load covers 3.8–6.3% of agency-estimated data-centre demand in the United States, 1.4–2.3% in China, at least 0.8–1.3% in Japan and at most 2.6% in South Korea. Data-centre growth absorbs 4–34% of observed wind and solar additions under agency estimates but only 0.13–1.0% under visible load, a ratio of 31 to 53. Reference estimates for China diverge by a factor of 2.6 (100 versus 260 TWh in 2024). The verdict on whether clean power keeps pace with AI therefore depends on what can be observed: while 92–99% of the load remains invisible in open registers, no stable answer is possible. We propose a minimum reporting standard for data-centre electricity.

Authors

Publication Details

Journal
Preprints.org
Published
2026-11-12
DOI
https://doi.org/10.20944/preprints202611.1257.v1
Primary Topic
Green IT and Sustainability
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Seen and Unseen Load: Open-Data Visibility of AI Data-Centre Electricity Demand and Wind–Solar Headroom in China, Japan and South Korea

Alfredo Merlet
Preprints.org
Green IT and Sustainability
preprint

Seen and Unseen Load: Open-Data Visibility of AI Data-Centre Electricity Demand and Wind–Solar Headroom in China, Japan and South Korea

Alfredo Merlet
preprint en

Abstract

The debate over whether renewable energy can sustain AI growth relies on demand figures whose origin is rarely examined. We compare open, facility-level records with agency estimates for China, Japan, South Korea and the United States, using the Epoch AI GPU-cluster register (338 clusters), the International Energy Agency, Lawrence Berkeley National Laboratory, South Korea's 11th Basic Plan for Electricity Supply and Demand, and Ember generation data. In 2024, visible load covers 3.8–6.3% of agency-estimated data-centre demand in the United States, 1.4–2.3% in China, at least 0.8–1.3% in Japan and at most 2.6% in South Korea. Data-centre growth absorbs 4–34% of observed wind and solar additions under agency estimates but only 0.13–1.0% under visible load, a ratio of 31 to 53. Reference estimates for China diverge by a factor of 2.6 (100 versus 260 TWh in 2024). The verdict on whether clean power keeps pace with AI therefore depends on what can be observed: while 92–99% of the load remains invisible in open registers, no stable answer is possible. We propose a minimum reporting standard for data-centre electricity.

Preprints.org
Affordable and clean energy, Climate action, Industry, innovation and infrastructure
Green IT and Sustainability
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon