Digital Infrastructure Efficiency and Carbon Rebound Risk: Cross−Country Evidence for Sustainable Transitions from 39 Economies, 2018–2024
デジタルインフラ効率とカーボンリバウンドリスク：39か国2018-2024年の持続可能な移行に関する国際比較証拠 (AI 翻訳)
Sirui Li, Xiangdong Liu, Xiangdong Liu, Johnny F. I. Lam, Xieqihua Liu, Xieqihua Liu, Jinghui Zhan
🤖 gxceed AI 要約
日本語
本研究は39か国のパネルデータを用いて、デジタルインフラのエネルギー効率（PUE）と炭素強度の関連を分析。PUE改善がむしろ炭素強度上昇と関連するリバウンド効果を発見し、再生可能エネルギー比率59.82%の閾値を特定。2023-2024年のAI計算能力急増が構造変化をもたらした可能性を示唆し、供給側のグリーン化だけでは不十分で絶対的な炭素上限が必要と提言。
English
Using panel data from 39 economies (2018-2024), this study examines the relationship between digital infrastructure energy efficiency (PUE) and carbon intensity. It finds a rebound effect where PUE reductions are associated with higher carbon intensity, identifies a 59.82% renewable energy threshold, and suggests that the 2023-2024 AI computing surge caused a structural shift. The paper argues that supply-side greening alone is insufficient and absolute carbon caps are needed.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
日本はデータセンターの増加とAI活用の拡大が進むが、本論文はデジタルインフラの効率改善だけではカーボンリバウンドを抑制できず、再生可能エネルギー調達と絶対的な排出上限の併用が必要であることを示唆。SSBJや有報での情報開示において、エネルギー効率指標（PUE）と炭素強度の関係を考慮する重要性を提起している。
In the global GX context
This study provides critical evidence for global climate policy, showing that efficiency gains in digital infrastructure may lead to increased carbon emissions due to rebound effects. It highlights the need for absolute emission caps and renewable energy thresholds, relevant for TCFD/ISSB disclosure frameworks that require companies to report energy efficiency and carbon intensity. The findings also inform the debate on AI's environmental impact.
👥 読者別の含意
🔬研究者:This paper offers a robust empirical framework (fixed effects, threshold regression, IV) for analyzing rebound effects in the ICT sector and the impact of AI computing surges on carbon intensity.
🏢実務担当者:Data center operators and ICT companies should note that improving PUE alone may not reduce carbon emissions without concurrent renewable energy expansion and absolute emission caps; these insights can inform sustainability strategies and disclosure.
🏛政策担当者:The study provides quantitative evidence that supply-side energy transition is insufficient to offset digital sector growth; regulators should consider absolute carbon caps and renewable energy mandates for ICT infrastructure, especially given the AI boom.
📄 Abstract（原文）
The synergistic transition toward digital transformation and green development has been widely regarded as a core pathway to achieving sustainable development in knowledge production. Using balanced panel data from 39 economies covering 2018–2024, this study employed a two-way fixed-effects model to examine the associations of the energy efficiency of digital infrastructure and the energy structure with carbon intensity (CI). The findings showed that: (1) Reductions in Power Usage Effectiveness (PUE) values were significantly associated with higher macro-level CI (coefficient = −2.1564, p < 0.05), which is consistent with the possibility of a rebound effect in the digital sector. Further, time-series discontinuity tests further suggested that the surge in AI computing power, especially in 2023–2024, may have coincided with a structural shift in this relationship (Chow test, p < 0.05). (2) A Panel Threshold Regression (PTR) identified an optimal renewable energy threshold at 59.82%. Crucially, the carbon rebound effect remained highly significant across both high and low green power regimes, demonstrating that supply-side energy transition alone cannot fully absorb the exponential carbon footprint of digital expansion. Furthermore, Instrumental Variable (IV-2SLS) and Placebo Break Tests confirmed the strict validity of these findings. (3) The emission-reduction benefits related to digital knowledge spillovers appeared to be subject to time lags and a possible energy lock in effect, while current environmental policies and carbon pricing mechanisms appear to impose insufficient constraints. This study provides a crucial quantitative framework for monitoring and evaluating the environmental sustainability of the ICT sector. By highlighting the limitations of pure supply-side greening and the necessity of absolute carbon caps, our findings offer integrated policy approaches to align the exponential growth of Generative AI with global sustainable development goals.
🔗 Provenance — このレコードを発見したソース
- openalex https://doi.org/10.3390/su18126216first seen 2026-07-07 04:38:43
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。