Journal article · 2026 · Open access
Energy use of AI inference, efficiency pathways, and test-time scaling
Joule, 10(8), 102430 · Published 22 April 2026 · DOI 10.1016/j.joule.2026.102430
In brief
“Energy use of AI inference, efficiency pathways, and test-time scaling” is a 2026 journal article by Felipe Oviedo and 7 co-authors, published in Joule. OpenAlex counts 57 citations as of 17 November 2026. It is open access under a CC BY-NC-ND licence: the full text is free to read at Joule.
- Citations
- 57
- Most cited in
- 2026 (56)
- Field-weighted impact
- 49.82
- References
- 46
- Authors
- 8
Abstract
As artificial intelligence (AI) inference scales to billions of queries, estimates of per-query energy use are increasingly important for capacity planning, efficiency interventions, and policy. Yet many public estimates assume non-production settings, leading to systematic overestimation. We introduce a bottom-up framework estimating inference energy from token throughput, node power, and overhead under large-scale deployment assumptions. For frontier-scale models (>200B parameters) on H100 nodes, we estimate a median energy of 0.31 Wh/query (interquartile range [IQR] 0.16–0.60), indicating that widely cited estimates are overstated by 4–20×. In test-time scaling scenarios 15× longer than typical queries, the median energy rises 13× to 3.91 Wh (IQR 2.15–7.05). Across models, serving systems, and hardware, we estimate 8–20× line-of-sight energy reductions. At data-center scale, serving 1 billion queries/day requires 0.7 GWh; if 10% are long queries, demand rises to 1.7 GWh/day. With efficiency interventions, it falls to 0.8 GWh/day, mitigating the energy impact of test-time scaling.
Cite this paper
Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.
APA 7
Oviedo, F., Kazhamiaka, F., Choukse, E., Kim, A., Luers, A. L., Nakagawa, M., Bianchini, R., & Ferres, J. L. (2026). Energy use of AI inference, efficiency pathways, and test-time scaling. Joule, 10(8), 102430. https://doi.org/10.1016/j.joule.2026.102430
MLA 9
Oviedo, Felipe, et al. “Energy use of AI inference, efficiency pathways, and test-time scaling.” Joule, vol. 10, no. 8, 2026, p. 102430. https://doi.org/10.1016/j.joule.2026.102430.
Chicago
Oviedo, Felipe, Fiodar Kazhamiaka, Esha Choukse, Allen Kim, Amy Lynd Luers, Melanie Nakagawa, Ricardo Bianchini, and Juan Lavista Ferres. 2026. “Energy use of AI inference, efficiency pathways, and test-time scaling.” Joule 10 (8): 102430. https://doi.org/10.1016/j.joule.2026.102430.
Harvard
Oviedo, F. et al. (2026) ‘Energy use of AI inference, efficiency pathways, and test-time scaling’, Joule, 10(8), p. 102430. doi:10.1016/j.joule.2026.102430.
IEEE
F. Oviedo et al., “Energy use of AI inference, efficiency pathways, and test-time scaling,” Joule, vol. 10, no. 8, p. 102430, 2026, doi: 10.1016/j.joule.2026.102430.
BibTeX
@article{oviedo2026energy,
title = {{Energy use of AI inference, efficiency pathways, and test-time scaling}},
author = {Oviedo, Felipe and Kazhamiaka, Fiodar and Choukse, Esha and Kim, Allen and Luers, Amy Lynd and Nakagawa, Melanie and Bianchini, Ricardo and Ferres, Juan Lavista},
journal = {Joule},
year = {2026},
volume = {10},
number = {8},
pages = {102430},
doi = {10.1016/j.joule.2026.102430},
url = {https://doi.org/10.1016/j.joule.2026.102430}
}
Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)
Topic and keywords
Physical Sciences › Engineering › Electrical and Electronic Engineering › Green IT and Sustainability
- test-time scaling
- large language model inference
- data center energy consumption
- token throughput
- energy efficiency interventions
- capacity planning
Work with this paper
Most-cited papers in Engineering
- A Closed‐form Equation for Predicting the Hydraulic Conductivity of Unsaturated SoilsMartinus Th. van Genuchten · Soil Science Society of America Journal · 1980 · 28,197 citations
- An improved technique for determining hardness and elastic modulus using load and displacement sensing indentation experimentsW. C. Oliver, George M. Pharr · Journal of materials research/Pratt's guide to venture capital sources · 1992 · 27,379 citations
- Digital communicationsOwen Bishop · Routledge eBooks · 2007 · 24,890 citations
- The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysisNorden Eh Huang, Zhengwei Shen, Steven R. Long and 6 more · Proceedings of the Royal Society A Mathematical Physical and Engineering Sciences · 1998 · 24,008 citations
- Numerical Heat Transfer and Fluid FlowSuhas V. Patankar · CRC Press eBooks · 2018 · 23,491 citations
Questions about this paper
How do I cite this paper in APA style?
Oviedo, F., Kazhamiaka, F., Choukse, E., Kim, A., Luers, A. L., Nakagawa, M., Bianchini, R., & Ferres, J. L. (2026). Energy use of AI inference, efficiency pathways, and test-time scaling. Joule, 10(8), 102430. https://doi.org/10.1016/j.joule.2026.102430
How many times has this paper been cited?
57 times, according to OpenAlex on 17 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.
Is this paper free to read?
Yes. It is open access at Joule: sign in and open the full text from this page.
Who wrote this paper?
Felipe Oviedo, Fiodar Kazhamiaka, Esha Choukse, Allen Kim, Amy Lynd Luers, Melanie Nakagawa, Ricardo Bianchini, Juan Lavista Ferres.
What is the DOI of this paper?
10.1016/j.joule.2026.102430, permanently at https://doi.org/10.1016/j.joule.2026.102430. Sign in and open it from this page.
Bibliographic data from OpenAlex (CC0), last checked 17 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.