Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 105 of the 105 references it cites were found in a registry or an index; 39 of the 103 sources its claims cite were checked for support.

From Efficiency Gains to Rebound Effects: The Problem of Jevons' Paradox in AI's Polarized Environmental Debate

Open access
Luccioni, Alexandra Sasha, Strubell, Emma, Crawford, Kate · 2025 · Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency

Abstract

As the climate crisis deepens, artificial intelligence (AI) has emerged as a contested force: some champion its potential to advance renewable energy, materials discovery, and large-scale emissions monitoring, while others underscore its growing carbon footprint, water consumption, and material resource demands. Much of this debate has concentrated on direct impacts—energy and water usage in data centers, e-waste from frequent hardware upgrades—without addressing the significant indirect effects. This paper examines how the problem of Jevons’ Paradox applies to AI, whereby efficiency gains may paradoxically spur increased consumption. We argue that understanding these second-order impacts requires an interdisciplinary approach, combining lifecycle assessments with socio-economic analyses. Rebound effects undermine the assumption that improved technical efficiency alone will ensure net reductions in environmental harm. Instead, the trajectory of AI’s impact also hinges on business incentives and market logics, governance and policymaking, and broader social and cultural norms. We contend that a narrow focus on direct emissions misrepresents AI’s true climate footprint, limiting the scope for meaningful interventions. We conclude with recommendations that address rebound effects and challenge the market-driven imperatives fueling uncontrolled AI growth. By broadening the analysis to include both direct and indirect consequences, we aim to inform a more comprehensive, evidence-b…

Overview

CoChat Summary

TL;DR

This conceptual paper argues that the polarized debate over AI's environmental impact is stuck on direct effects (data center energy, water, e-waste) while ignoring indirect "rebound effects," in which efficiency gains spur higher overall consumption—Jevons' Paradox. The authors call for interdisciplinary lifecycle-plus-socioeconomic analysis and for transparency, regulation, and business-model reform rather than efficiency alone.

Key findings

  • Direct impacts are rising despite efficiency gains: the IEA puts data center electricity at 2% of global electricity, more than doubling by 2026 (surpassing Canada's national power use); Google reports a 48% GHG increase since 2019, Baidu 32.6% over 2021, and Microsoft 29.1% since 2020, while one analysis found real data center footprints can exceed reported values by over 600%.
  • Water and materials burdens are concentrated and opaque: Microsoft reported a 34% increase in global water consumption between 2021 and 2022 (over 1.7 billion gallons) and Google a 20% uptick; one study suggests 10–50 GPT-3 queries consume about half a liter of water; 54% of technology-critical materials come from Indigenous or peasant territory and 62% from drought-prone zones.
  • E-waste and hardware demand scale with AI: e-waste reached 62 million tonnes in 2022 with only about 22% formally collected and recycled, and AI is estimated to add 1.2–5 million metric tons of e-waste by 2030; NVIDIA shipped 3.7 million GPUs in 2024, more than a million more units than in 2023, even as hardware efficiency improved.
  • The paper taxonomizes indirect effects into material (substitution, scale, space), economic (direct, indirect/real income, economy-wide), and societal/behavioral (induction, time, indirect policy) rebound effects—e.g., targeted advertising revenue reached $36 billion in 2024 and Amazon's AI recommendation tool generates almost a third of its annual sales, while AI-assisted navigation has been shown to cut emissions by an average of 3.4%.

Methods

  • Conceptual/position paper: no new empirical data collection or experiment; it synthesizes existing literature, corporate ESG reports, and prior studies to reframe the AI-and-environment debate around indirect effects.
  • Builds an analytical taxonomy by adapting Börjesson Rivera et al.'s second-order ICT effects taxonomy and Kaack et al.'s work on AI's indirect impacts, organizing rebound effects into three themes: material objects and physical spaces, micro-/macro-economic processes, and society and human behavior.
  • Proposes mitigation and tracking approaches (qualitative/quantitative research, transparency, regulation, efficient devices and distributed compute) rather than testing them.

Limitations

  • Stated: the authors note that applying Jevons' Paradox to AI has "notable conceptual and empirical limitations," citing Sorrell's critique that it oversimplifies causal relationships among efficiency gains, economic expansion, and behavioral responses; they also state that rebound effects are characterized by uncertainty and heterogeneity, and that measuring them is hampered by severe data scarcity, trade-secret opacity, and non-standardized corporate reporting.
  • Stated: the paper acknowledges that claims about AI's climate-positive applications "remain largely hypothetical, with little explicit evidence or quantitative analysis," and that existing AI-vs-human LCA comparisons (e.g., 130–1500× less CO2e per page of text, 310–2900× per image) do not account for social impacts, legality, or rebound effects.
  • Evident from the text: the paper offers no original quantification of rebound magnitudes for AI and does not empirically validate its taxonomy; several cited figures come from corporate self-reports or third-party estimates whose reliability the authors themselves question.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1145/3715275.3732007

Cites

105 of 105 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

28 Valid · 8 Not valid · 44 Not checked · 94 Own results and interpretations

70 more claims · 11 with their cited sources checked

Create a free account to see all 174 claims and to verify the 31 references still open.

See all claims — free account
  • It may state two separate claims as one. 47
  • Probably not a substantial claim. 73
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 87 references

Research works printed but not linked: 8

  1. Yoshua Bengio Salem Lahlou Tristan Deleu Edward J Hu Mo Tiwari and Emmanuel Bengio. 2023. Gflownet foundations. The Journal of Machine Learning Research 24 1 (2023) 10006–10060.
  2. Ye Chen Michael Kapralov John Canny and Dmitry Pavlov. 2009. Factor modeling for advertisement targeting. Advances in neural information processing systems 22 (2009).
  3. Peter Clutton-Brock, David Rolnick, Priya L Donti, and Lynn Kaack. 2021. Climate change and AI. recommendations for government action. Technical Report. GPAI, Climate Change AI, Centre for AI & Climate.
  4. Jaume Freire-González and Ignasi Puig-Ventosa. 2015. Energy efficiency policies and the Jevons paradox. International Journal of Energy Economics and Policy 5 1 (2015) 69–79.
  5. W Stanley Jevons. 1866. The coal question. In The Economics of Population. Routledge, 193–204.
  6. Eric Masanet Arman Shehabi Nuoa Lei Sarah Smith and Jonathan Koomey. 2020. Recalibrating global data center energy-use estimates. Science 367 6481 (2020) 984–986. · DOI · Paper page
  7. Zeeshan Tariq Murtada Saleh Aljawad Amjed Hasan Mobeen Murtaza Emad Mohammed Ammar El-Husseiny Sulaiman A Alarifi Mohamed Mahmoud and Abdulazeez Abdulraheem. 2021. A systematic review of data science and machine learning applications to the oil and gas industry. Journal of Petroleum Exploration and Production Technology (2021) 1–36.
  8. Sunita Tiwari and Pratibha Singh. 2011. Environmental impacts of e-commerce. In International conference on environment Science and engineering, Vol. 8. 202–207.

Cited by

See all 71

Metadata

Authors
Luccioni, A. S., Strubell, E., & Crawford, K.
Affiliations
  • Hugging Face, Montreal, Canada
  • Carnegie Mellon University (US)
  • Microsoft Research New York City (United States) (US)
Venue
Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency
Publisher
ACM
Published
2025-06-23
Pages
76-88
OpenAlex
W4411541921
Cited by
71
References
87
Citations by year
2026: 53 · 2025: 18
FWCI
150.99
Type
conference paper
Language
en
Open access
Locations
  • ACM Conference on Fairness, Accountability, and Transparency (FAccT) (publishedVersion, cc-by)
  • arXiv (Cornell University) (submittedVersion, cc-by-nc-sa)
  • arXiv (Cornell University)
Funders
NSF (National Science Foundation) (2326610)
Topics
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Green IT and Sustainability
  • Physical Sciences › Environmental Science › Environmental Engineering › Environmental Impact and Sustainability
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Energy, Environment, Economic Growth
Keywords
Jevons paradox, rebound effect, artificial intelligence, environmental impact of AI, efficiency improvement, energy consumption, environmental debate, technology and sustainability

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-27, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-27 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: From Efficiency Gains to Rebound Effects: The Problem of Jevons' Paradox in AI's Polarized Environmental Debate. https://app.cochat.ai/papers/10.1145/3715275.3732007
The paper
Luccioni, A. S., Strubell, E., & Crawford, K. (2025). From Efficiency Gains to Rebound Effects: The Problem of Jevons' Paradox in AI's Polarized Environmental Debate. Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency. https://doi.org/10.1145/3715275.3732007

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.