Pith. sign in

REVIEW 3 major objections 5 minor 56 references

Modelling Scenarios for Carbon-aware Geographic Load Shifting of Compute Workloads

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Geographic load shifting of compute workloads delivers only about 5% emissions reduction under realistic scenarios, even with optimistic assumptions.

desk verdict A transparent analytical model and closed-form upper bound that likely gets the qualitative answer right — real geographic load shifting saves only a few percent — but the exact ~5% figure rests on an unjustified averaging step. read the letter →

arxiv 2509.07043 v4 pith:OBCSTRER submitted 2025-09-08 cs.OH

Wim Vanderbauwhede This is my paper
classification cs.OH
keywords geographicloadshiftingcarbon-awarecomputingdatacentreemissionsembodiedcarbonintensityrenewableenergyflexibilitymodelling
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks how much greenhouse-gas emissions geographic load shifting—moving compute workloads to data centres in lower-carbon regions—can actually save. It builds an explicitly optimistic analytical model that counts both operational emissions and embodied carbon, then applies it to commercial AI data centres and HPC centres. Across realistic scenarios the model puts the reduction at roughly 5% (4.6–5.3% for the AI cases, and 3.3–13.6% for HPC depending on flexibility). The paper concludes that this is far below what is needed to offset the growth in data-centre capacity, so carbon-aware geographic load shifting alone is not a climate solution.

What carries the argument

The carrying object is the effective shifted fraction alpha_eff, a capped function of the high-site load, low-site free capacity, and the number of sites. It determines how much work actually moves after constraints are applied and feeds into the shifted-emissions equation that produces the reduction ratio r = (C_b − C_gls)/C_b. The model is deliberately linear and optimistic: it ignores grid capacity, demand, and curtailment, so its estimates are upper bounds on what geographic load shifting can achieve.

What would settle it

Run the same scenarios hour by hour with real grid carbon-intensity traces and load data, applying the cap equation at every time step, and compare total emissions with the annual-average formula. If the hour-by-hour total differs materially from the ~5% estimate, the averaging assumption is the cause.

Watch

Extended reading notes

Core claim

The central claim is that geographic load shifting, even under favourable assumptions, yields only small emission reductions—typically around 5% for realistic parameters—because embodied carbon, idle power, a limited shiftable fraction of workload, time constraints, and the nonzero carbon intensity of renewables all dilute the operational gain. The paper formalises this in a linear model where baseline and shifted emissions are functions of per-node embodied carbon, operational carbon at high- and low-emission sites, load, idle factor, shiftable fraction, time fraction, and overhead. It derives an ideal-case upper bound: with identical loads at or below half capacity, and no embodied carbon,

Load-bearing premise

The load-bearing premise is that yearly average load and yearly average carbon intensity can replace the real time-varying values without changing the result, even though the shiftable fraction is a capped nonlinear function of those values.

Editorial extensions

If this is right

  • If the claim is correct, data-centre operators should expect only single-digit percentage emission savings from spatial shifting alone, even in ideal grid conditions.
  • To counter data-centre growth, reductions must come from other levers such as lower embodied carbon, longer server life, higher efficiency, heat reuse, or operational changes that go well beyond moving workloads.
  • As grids decarbonise, the gap between high- and low-emission regions shrinks, further reducing the ceiling on geographic load-shifting benefits.
  • The ideal-case formula gives a quick rule-of-thumb for any region pair: the maximum possible reduction from shifting is (C_hi − C_lo)/(C_hi + C_lo) when loads are below half capacity.
  • The paper's growth-compensation analysis shows that a 5% reduction is undone in less than a year of medium-range data-centre capacity growth, so the benefit is short-lived.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The annual-average approximation may not be neutral: the shiftable load is a capped, nonlinear function of load and carbon intensity, so averaging before applying the cap could bias the ~5% result; the paper does not quantify this bias.
  • The same dilution factors—embodied carbon, idle power, limited shiftable fraction, and nonzero renewable carbon intensity—would likely also bound purely temporal or spatio-temporal shifting, suggesting those approaches may face similar ceilings.
  • A direct testable extension is to run the model hour-by-hour with real carbon-intensity and load traces for different region pairs, which would show whether the annual-average shortcut under- or overstates the true reduction.
  • The model's focus is emissions only; geographic load shifting may still have value for other goals such as 24/7 carbon-free-energy matching or grid flexibility, but that value is not an emissions reduction and should not be counted as one.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper presents an analytical model for estimating the CO2e-emission reductions obtainable by geographic load shifting of compute workloads between a high-emission and a low-emission data centre. The model separates embodied and operational emissions, and includes load, idle-power fraction, movable-workload fraction, time-fraction, and overhead. A closed-form ideal upper bound is given in Eq. (19). The model is applied to two AI data-centre scenarios and five HPC scenarios. The central quantitative claims are that realistic reductions are small (4.6% and 5.3% for the AI scenarios; between 3.3% and 13.6% for the HPC scenarios), and that such reductions cannot offset the projected growth in data-centre emissions (Fig. 1).

Significance. If the headline figures hold, the paper provides a useful and largely accessible counterweight to optimistic claims in the carbon-aware-computing literature. Its strengths include a transparent model, public source code, an explicit parameter-free upper bound in Eq. (19), and scenario parameters that are stated in detail rather than hidden. The qualitative conclusion that geographic load shifting yields only small reductions under realistic assumptions is likely robust. However, the numerical '~5%' claim depends on an unjustified averaging of a nonlinear capped function, and the embodied-carbon value is internally inconsistent between text and Table I. These issues need to be resolved before the quantitative conclusions can be accepted as stated.

major comments (3)
  1. [Section IV-F, Eqs. (12), (15)] The statement that 'α, β and λ evolve over time, we use yearly averages without loss of generality' is not justified. α_eff is given by α_eff = min(α, (1−λ_lo)/λ_hi, 1), which is a nonlinear, capped function of λ_hi and λ_lo. In general E[α_eff] ≠ α_eff(E[λ_hi], E[λ_lo]); because the function is neither globally convex nor concave, Jensen's inequality does not bound the error. The reported 4.6–5.3% reductions are computed from yearly-averaged λ and α_eff. A simple two-point load distribution with the same mean λ=0.83 (e.g., 80% at λ=0.8 and 20% at λ=0.95 for both sites) lowers α_eff from 0.20 to ≈0.17 and reduces the solar-scenario saving by about one percentage point. The qualitative conclusion is likely robust, but the precise '~5%' figure is not established without quantifying this bias or redoing the calculations time-resolved.
  2. [Section V-A and Table I] The embodied-carbon value is inconsistent. The text states 'Our embodied carbon model yields an estimate for this configuration of 5,730 kgCO2e/y per node,' but Table I lists c_emb = 2,066 kgCO2e/y per node. The accompanying statement that the estimate 'corresponds to 0.45 ktCO2e/MW/y' matches 2,066 kg/year per node at 4.55 kW/node, not 5,730. Since embodied carbon is one of the factors that the paper identifies as limiting reductions (see Fig. 7), these two values lead to different baselines and different reduction percentages. The authors must reconcile this discrepancy and recompute the affected scenario results if 5,730 is the intended value.
  3. [Section V-A.3, discussion of Fig. 7] The text says the kink in the scenario curves is 'triggered when the load exceeds 0.5.' For the scenario parameters α=0.2, Eq. (12) gives the cap condition αλ_hi > 1−λ_lo, so with λ_hi=λ_lo=λ the kink occurs at λ > 1/(1+α) = 0.833, not 0.5. The 0.5 threshold applies only to the ideal α=1 case in Eq. (19). This mischaracterizes the behaviour of the curves in Fig. 7 and should be corrected.
minor comments (5)
  1. [Eq. (3)] The variable ν appears in the expression '(1+ν).PUE·CI' but is never defined. Please define it or remove it. In Eq. (4) there is also a missing closing parenthesis.
  2. [Section IV-A] In the parameter list, 'λlo load of high-emission site' should read 'low-emission site.'
  3. [Section V-A.1 / Table I] The scenario text gives the average fossil CI for the high-emission regions as 448 gCO2e/kWh, but c_hi in Table I corresponds to about 410 gCO2e/kWh when divided by the per-node annual energy. c_lo corresponds to 41 gCO2e/kWh as stated. Please reconcile the CI values used for c_hi.
  4. [Section IV-F] The phrase 'leaving assuming' should be corrected, and the paper says 'we have simulated' in the conclusion although the work is an analytical calculation; 'computed' or 'evaluated' would be more accurate.
  5. [Section II] The 'Pennsylvania-New Jersey-Maryland interconnection' is abbreviated as PMJ in the text; the standard abbreviation is PJM. Please correct.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the 5% scenario result is derived by substitution from explicit parameters, not fitted or self-referential.

full rationale

The paper's derivation chain is self-contained arithmetic from explicitly stated operational parameters. The baseline emissions (Eq. 11), shifted emissions (Eq. 16), and reduction (Eq. 18) are direct algebraic consequences of the definitions of c_emb, c_hi, c_lo, lambda, gamma, alpha_eff, beta, and eta. Equation (19) is a closed-form specialization of the same model, not an imported result. No parameter is fitted to reproduce the '~5%' conclusion; the scenario values in Tables I and II are author-chosen inputs, and the reported reductions are computed by substitution. The embodied-carbon values come from the author's LCA model [20], which is a self-citation, but it is code-reproduced and published, and the qualitative conclusion is not sensitive to it: removing embodied carbon from Table I changes the reduction from 4.6% to about 5.6%, so the central claim does not reduce to that citation. The 'without loss of generality' use of yearly averages in Sections IV-D and IV-F is a mathematical simplification whose bias is not quantified, especially because alpha_eff in Eq. (12) is a nonlinear capped function; however, that is a correctness/robustness concern about the model's approximation error, not circularity, since it does not define the predicted reduction in terms of itself or rename a fitted value as a prediction. The paper also explicitly states its model is optimistic and ignores grid capacity, demand, and curtailment, and it flags missing overhead data in the HPC scenarios, which further shows the direction of any correction rather than a concealed circular step. Overall, the derivation is self-contained and the reported reductions are scenario arithmetic, not the re-statement of an assumption.

Assumptions & free parameters 8 free parameters · 4 assumptions · 0 invented entities

The central result depends on numerous hand-set scenario parameters (alpha, beta, gamma, lambda, CI) drawn from literature or the author's LCA model. The closed-form upper bound (Eq 19) is parameter-free except CI values, but the ~5% figures are outputs of chosen parameters. No new physical entities are introduced.

free parameters (8)
  • gamma (idle power fraction) = 0.30
    Chosen from GPU idle power references [33]-[36]; directly scales operational emissions in all scenarios.
  • PUE = 1.16
    Assumed representative for a hyperscale data centre; multiplies all operational emissions.
  • lambda_hi, lambda_lo (loads) = 0.83 for AI scenarios; 0.80, 1.00, 0.50 in HPC scenarios
    AI load set by 250 MW nominal vs 300 MW max; HPC loads are assumptions from facility discussions.
  • alpha (movable workload fraction) = 0.20 for AI; 1.00 or 0.25 for HPC
    AI alpha corresponds to 50 MW of 250 MW; HPC alpha chosen as optimistic or realistic.
  • beta (time fraction) = 0.52 (solar), 0.54 (wind), 1.00 or 0.50 (HPC)
    Derived from sunshine hours, time zone overlap, wind load factors and correlation, then hand-adjusted.
  • node power = 4550 W (AI DGX-A100), 1200 W (HPC)
    From Nvidia and assumed HPC node configuration.
  • embodied carbon per node = 2066 kgCO2e/y in Table I; text says 5730
    From the author's LCA model [20]; the inconsistency between text and table is unexplained.
  • carbon intensity values = US 369, UK 211, DE 344, FR 44, SE 36, TW 642 etc.
    Taken from Ember [29]; external data, not fitted in this paper.
assumptions (4)
  • domain assumption Load lambda(t) and carbon intensity C(t) are independent random variables, so E(lambda*C)=E(lambda)E(C).
    Invoked in Section IV-D to replace time series with yearly averages.
  • domain assumption Yearly averages of lambda, alpha, beta and C can be used 'without loss of generality' despite the nonlinear cap on alpha_eff.
    Section IV-F uses this to simplify the model; the nonlinearity in Eqs 12/15 makes this approximation potentially biased.
  • domain assumption The target grid has sufficient capacity to accept the shifted load at full power.
    Section IV states the model is not grid-aware; the paper acknowledges real-world reductions are smaller.
  • domain assumption Embodied carbon is constant and end-of-life emissions can be ignored.
    Section IV and refs [17], [19]; used in all scenario calculations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Modelling Scenarios for Carbon-aware Geographic Load Shifting of Compute Workloads." pith.science (2026). https://pith.science/paper/OBCSTRER

@misc{pith2026250907043,
  author       = {Pith},
  title        = {Pith review of: Modelling Scenarios for Carbon-aware Geographic Load Shifting of Compute Workloads},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OBCSTRER}},
  note         = {Machine review of arXiv:2509.07043}
}
read the original abstract

We present an analytical model to evaluate the reductions in emissions resulting from geographic load shifting. This model is optimistic as it ignores issues of grid capacity, demand and curtailment. In other words, real-world reductions will be smaller than the estimates. However, even with these assumptions, the presented scenarios show that the realistic reductions from carbon-aware geographic load shifting are small, of the order of 5\%. This is not enough to compensate the growth in emissions from global data centre expansion.

Figures

Figures reproduced from arXiv: 2509.07043 by the authors.

Figure 1
Figure 1. Years of growth as per the McKinsey scenarios compensated by [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Locations used in the scenarios. US, UK, Germany, France and Spain are used in the commercial data centre scenarios, the HPC sites (BNL, EPCC, [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Solar scenario, ideal assumptions (two sites are low-CI in any 8-hour period). 100 MW can be shifted. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Solar scenario, more realistic assumptions (two, three or one low-CI sites in an 8-hour period). Only (100+50+50)/3 MW can be shifted. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Wind scenario, ideal assumptions (wind load factor 50% for all sites, no correlation). 100 MW can be shifted. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Wind scenario, more realistic assumptions (load factor 35% and correlation 10% ). Only 63 MW can be shifted. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Emissions reduction as a function of load for the AI data centre scenarios. The figure illustrates the effect of load [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Emissions reduction as a function of load for HPC centre Scenario [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Emissions reduction as a function of load for the HPC centre scenarios. The figure illustrates the effect of load [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

56 extracted references · 49 canonical work pages

  1. [1]

    Our house is on fire: The climate emergency and computing’s responsibility

    B. Knowles, K. Widdicks, G. Blair, M. Berners-Lee, and A. Friday, “Our house is on fire: The climate emergency and computing’s responsibility.” Communications of the ACM, vol. 65, no. 6, pp. 38–40, 2022

    work page 2022

  2. [2]

    Assessing ict global emissions footprint: Trends to 2040 & recommendations,

    L. Belkhir and A. Elmeligi, “Assessing ict global emissions footprint: Trends to 2040 & recommendations,”Journal of cleaner production, vol. 177, pp. 448–463, 2018

    work page

  3. [3]

    AI power: Expanding data center capacity to meet growing demand,

    McKinsey & Company, “AI power: Expanding data center capacity to meet growing demand,” Oct

    work page

  4. [4]

    2024 United States Data Center Energy Usage Report,

    A. Shehabi, A. Hubbard, A. Newkirk, N. Lei, M. A. B. Siddik, B. Holecek, J. Koomey, E. Masanet, D. Sartoret al., “2024 United States Data Center Energy Usage Report,” Lawrence Berkeley National Laboratory, Tech. Rep. LBNL-2001637, dec 2024

    work page 2024

  5. [5]

    Electricity generation from other renewables, excluding bioenergy – Ember,

    Ember, “Electricity generation from other renewables, excluding bioenergy – Ember,” 2025, energy Institute - Statistical Review of World Energy (2024) - with major processing by Our World in Data. [Online]. Available: https://ourworldindata.org/grapher/ electricity-prod-source-stacked

    work page 2025

  6. [6]

    The real problem with the AI hype,

    W. Vanderbauwhede, “The real problem with the AI hype,” 01 2025. [Online]. Available: https://limited.systems/articles/ the-real-problem-with-AI/

    work page 2025

  7. [7]

    Carbon- aware computing for datacenters,

    A. Radovanovi ´c, R. Koningstein, I. Schneider, B. Chen, A. Duarte, B. Roy, D. Xiao, M. Haridasan, P. Hung, N. Careet al., “Carbon- aware computing for datacenters,”IEEE Transactions on Power Systems, vol. 38, no. 2, pp. 1270–1280, 2022

    2022

  8. [8]

    Let’s wait awhile: How temporal workload shifting can reduce carbon emissions in the cloud,

    P. Wiesner, I. Behnke, D. Scheinert, K. Gontarska, and L. Thamsen, “Let’s wait awhile: How temporal workload shifting can reduce carbon emissions in the cloud,” inProceedings of the 22nd International Middleware Conference, 2021, pp. 260–272

    work page 2021

Show all 56 references
  1. [9]

    Using geographic load shifting to reduce carbon emissions,

    J. Lindberg, B. C. Lesieutre, and L. A. Roald, “Using geographic load shifting to reduce carbon emissions,”Electric Power Systems Research, vol. 212, p. 108586, 2022. [Online]. Available: https: //www.sciencedirect.com/science/article/pii/S0378779622006757

    2022

  2. [10]

    The ieee reliability test system: A proposed 2019 update,

    C. Barrows, A. Bloom, A. Ehlen, J. Ik ¨aheimo, J. Jorgenson, D. Krish- namurthy, J. Lau, B. McBennett, M. O’Connell, E. Preston, A. Staid, G. Stephen, and J.-P. Watson, “The ieee reliability test system: A proposed 2019 update,”IEEE Transactions on Power Systems, vol. 35, no. ...

    2019

  3. [11]

    Carbon explorer: A holistic framework for designing carbon aware datacenters,

    B. Acun, B. Lee, F. Kazhamiaka, K. Maeng, U. Gupta, M. Chakkaravarthy, D. Brooks, and C.-J. Wu, “Carbon explorer: A holistic framework for designing carbon aware datacenters,” in Proc. ASPLOS 2023, ser. ASPLOS 2023. New York, NY , USA: Association for Computing Machinery, 2023...

    2023

  4. [12]

    Carbon-aware load balance control of data centers with renewable generations,

    W.-T. Lin, G. Chen, and H. Li, “Carbon-aware load balance control of data centers with renewable generations,”IEEE Transactions on Cloud Computing, vol. 11, no. 2, pp. 1111–1121, 2023

    2023

  5. [13]

    CAFE: Carbon-Aware Federated Learning in Geographically Distributed Data Centers,

    J. Bian, L. Wang, S. Ren, and J. Xu, “CAFE: Carbon-Aware Federated Learning in Geographically Distributed Data Centers,” inProceedings of the 15th ACM International Conference on Future and Sustainable Energy Systems, 2024, pp. 347–360

    2024

  6. [14]

    Spatio-temporal load shifting for truly clean computing,

    I. Riepin, T. Brown, and V . M. Zavala, “Spatio-temporal load shifting for truly clean computing,”Advances in Applied Energy, vol. 17, p. 100202, 2025

    2025

  7. [15]

    Conductor: A col- laboration framework for multi-data-center demand response,

    F. Acun, I. C. Paschalidis, and A. K. Coskun, “Conductor: A col- laboration framework for multi-data-center demand response,” in2024 IEEE 15th International Green and Sustainable Computing Conference (IGSC), 2024, pp. 22–28. IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING, VOL. XX,...

    2024

  8. [16]

    Mitigating curtailment and carbon emissions through load migration between data centers,

    J. Zheng, A. A. Chien, and S. Suh, “Mitigating curtailment and carbon emissions through load migration between data centers,”Joule, vol. 4, no. 10, pp. 2208–2222, 2020

    2020

  9. [17]

    The life cycle assessment of a uk data centre,

    B. Whitehead, D. Andrews, and A. Shah, “The life cycle assessment of a uk data centre,”The International Journal of Life Cycle Assessment, vol. 20, pp. 332–349, 2015

    2015

  10. [18]

    The dirty secret of ssds: Embodied carbon,

    S. Tannu and P. J. Nair, “The dirty secret of ssds: Embodied carbon,” SIGENERGY Energy Inform. Rev., vol. 3, no. 3, p. 4–9, Oct. 2023. [Online]. Available: https://doi.org/10.1145/3630614.3630616

    2023

  11. [19]

    Quantifying Data Center Scope 3 GHG Emissions to Prioritize Reduction Efforts,

    P. Lin, R. Bunger, and V . Avelar, “Quantifying Data Center Scope 3 GHG Emissions to Prioritize Reduction Efforts,” may 2023. [Online]. Available: https://www.se.com/ww/en/download/document/SPD WP99 EN/

    2023

  12. [20]

    Life cycle analysis for emis- sions of scientific computing centres,

    M. Wadenstein and W. Vanderbauwhede, “Life cycle analysis for emis- sions of scientific computing centres,”The European Physical Journal C, vol. 85, no. 8, p. 913, Aug. 2025

    2025

  13. [21]

    Model for life cycle analysis for emissions of scientific computing centres,

    ——, “Model for life cycle analysis for emissions of scientific computing centres,” Aug. 2025. [Online]. Available: https://doi.org/10. 5281/zenodo.16894210

    2025

  14. [22]

    Digital & environment: How to evaluate server manufacturing footprint, beyond greenhouse gas emis- sions?

    R. Lorenzini, “Digital & environment: How to evaluate server manufacturing footprint, beyond greenhouse gas emis- sions?” Nov 2021. [Online]. Available: https://boavizta.org/en/blog/ empreinte-de-la-fabrication-d-un-serveur

    2021

  15. [23]

    Act: designing sustainable computer systems with an architectural carbon modeling tool,

    U. Gupta, M. Elgamal, G. Hills, G.-Y . Wei, H.-H. S. Lee, D. Brooks, and C.-J. Wu, “Act: designing sustainable computer systems with an architectural carbon modeling tool,” inProceedings of the 49th Annual International Symposium on Computer Architecture, ser. ISCA ’22. New Yo...

    2022

  16. [24]

    Dtco including sustainability: Power-performance-area-cost- environmental score (ppace) analysis for logic technologies,

    M. Garcia Bardon, P. Wuytens, L.-A. Ragnarsson, G. Mirabelli, D. Jang, G. Willems, A. Mallik, A. Spessot, J. Ryckaert, and B. Parvais, “Dtco including sustainability: Power-performance-area-cost- environmental score (ppace) analysis for logic technologies,” in2020 IEEE Interna...

    2020

  17. [25]

    SK Hynix Details DDR5-6400,

    A. Shilov, “SK Hynix Details DDR5-6400,” Feb

  18. [26]

    An introduction to the algebra of random variables,

    J. Kermond, “An introduction to the algebra of random variables,” inProceedings of the 47th Annual Conference of the Mathematical Association of Victoria, 2010, pp. 1–16

    2010

  19. [27]

    The effect of the network in cutting carbon for geo-shifted workloads,

    Y . Guo, A. Tomlinson, R. Su, and G. Porter, “The effect of the network in cutting carbon for geo-shifted workloads,”arXiv preprint arXiv:2504.14022, 2025

    2025 arXiv

  20. [28]

    Causal allocation of fixed impacts in product systems: Assessing the effect of data demand on network energy consumption,

    D. Schien, P. Shabajee, L. Krug, G. McSorley, and C. Preist, “Causal allocation of fixed impacts in product systems: Assessing the effect of data demand on network energy consumption,”Journal of Industrial Ecology, vol. n/a, no. n/a, Jul 2025. [Online]. Available: https://onli...

    2025 doi

  21. [29]

    Carbon intensity of electricity generation - Ember and Energy Institute,

    Ember, “Carbon intensity of electricity generation - Ember and Energy Institute,” 2025, energy Institute - Statistical Review of World Energy (2024) - with major processing by Our World in Data. [Online]. Avail- able: https://ourworldindata.org/grapher/carbon-intensity-electricity

    2025

  22. [30]

    A model for the achievable reduction in emissions from carbon-aware geographic load shifting ,

    W. Vanderbauwhede, “ A model for the achievable reduction in emissions from carbon-aware geographic load shifting ,”

  23. [31]

    Scotland’s Ravenscraig pro- posed for £4bn hyperscale data centre,

    Mark Allen Group, “Scotland’s Ravenscraig pro- posed for £4bn hyperscale data centre,” Jun 2025. [Online]. Available: https://www.theengineer.co.uk/content/news/ ravenscraig-site-proposed-for-4bn-hyperscale-data-centre

    2025

  24. [32]

    Energy and Power Efficiency for Applications on the Latest NVIDIA Technology,

    A. Gray, “Energy and Power Efficiency for Applications on the Latest NVIDIA Technology,” Mar 2024. [Online]. Available: https://developer-blogs.nvidia.com/wp-content/uploads/2024/ 10/Energy-Efficiency-GTC.pdf

    2024

  25. [33]

    AMD EPYC 7742 Benchmarks and Review Simply Peerless,

    P. Kennedy, “AMD EPYC 7742 Benchmarks and Review Simply Peerless,” dec 2019. [Online]. Available: https://www.servethehome. com/amd-epyc-7742-benchmarks-and-review-simply-peerless/

    2019

  26. [34]

    SPEC CPU 2017 Integer Speed Result,

    Standard Performance Evaluation Corporation, “SPEC CPU 2017 Integer Speed Result,” 2019. [Online]. Available: http://spec.org/ cpu2017/results/res2020q2/cpu2017-20200413-21925.pdf

    2017

  27. [35]

    Accurate and convenient energy measurements for gpus: A detailed study of nvidia gpu’s built-in power sensor,

    Z. Yang, K. Adamek, and W. Armour, “Accurate and convenient energy measurements for gpus: A detailed study of nvidia gpu’s built-in power sensor,” inSC24: International Conference for High Performance Computing, Networking, Storage and Analysis. IEEE, Nov. 2024, pp. 1–

    2024

  28. [36]

    Improving GPU energy efficiency through an application-transparent frequency scaling policy with performance assurance,

    Y . Zhang, Q. Wang, Z. Lin, P. Xu, and B. Wang, “Improving GPU energy efficiency through an application-transparent frequency scaling policy with performance assurance,” inProceedings of the Nineteenth European Conference on Computer Systems, 2024, pp. 769–785

    2024

  29. [37]

    Annex III: Technology-specific cost and performance parameters

    S. S., T. Bruckner, L. Fulton, E. Hertwich, A. McKinnon, D. Perczyk, J. Roy, R. Schaeffer, R. Sims, P. Smith, and R. Wiser, “Annex III: Technology-specific cost and performance parameters.” 2014, Eden- hofer, O., R. Pichs-Madruga, Y . Sokona, E. Farahani, S. Kadner, K. Seyboth...

    2014

  30. [38]

    Application of lca to determine environmental impact of concentrated photovoltaic solar panels - state-of-the-art,

    A. Ziemi ´nska-Stolarska, M. Pietrzak, and I. Zbici ´nski, “Application of lca to determine environmental impact of concentrated photovoltaic solar panels - state-of-the-art,”Energies, vol. 14, no. 11, 2021. [Online]. Available: https://www.mdpi.com/1996-1073/14/11/3143

    2021

  31. [39]

    Available: http://dx.doi.org/10.1109/SC41406.2024.00028

    [Online]. Available: http://dx.doi.org/10.1109/SC41406.2024.00028

    2024 arXiv

  32. [40]

    WMO Climate Normals,

    NOAA National Centers for Environmental Information, “WMO Climate Normals,” 2025. [Online]. Available: https://www.ncei.noaa. gov/products/wmo-climate-normals

    2025

  33. [41]

    Deutschlandwetter im Jahr 2022,

    Deutsche Wetterdienst, “Deutschlandwetter im Jahr 2022,” 2022. [Online]. Available: https://www.dwd.de/DE/presse/pressemitteilungen/ DE/2022/20221230 deutschlandwetter jahr2022 news.html

    2022

  34. [42]

    Comprehensive life cycle assessment of large wind turbines in the us,

    A. Alsaleh and M. Sattler, “Comprehensive life cycle assessment of large wind turbines in the us,”Clean Technologies and Environmental Policy, vol. 21, pp. 887–903, 2019

    2019

  35. [43]

    UK temperature, rainfall and sunshine time series,

    UK Met Office, “UK temperature, rainfall and sunshine time series,” 2025. [Online]. Avail- able: https://www.metoffice.gov.uk/research/climate/maps-and-data/ uk-temperature-rainfall-and-sunshine-time-series

    2025

  36. [44]

    Life cycle assessment and life cycle cost analysis of a 40 mw wind farm with consideration of the infrastructure,

    Q. Li, H. Duan, M. Xie, P. Kang, Y . Ma, R. Zhong, T. Gao, W. Zhong, B. Wen, F. Bai, and A. K. Vuppaladadiyam, “Life cycle assessment and life cycle cost analysis of a 40 mw wind farm with consideration of the infrastructure,”Renewable and Sustainable Energy Reviews, vol. 138,...

    2021

  37. [45]

    Comparative lca of technology improvement opportunities for a 1.5-mw wind turbine in the context of an onshore wind farm,

    M. Ozoemena, W. M. Cheung, and R. Hasan, “Comparative lca of technology improvement opportunities for a 1.5-mw wind turbine in the context of an onshore wind farm,”Clean Technologies and Environ- mental Policy, vol. 20, no. 1, pp. 173–190, 2018

    2018

  38. [46]

    Wind farm capacity factor forecasting: An australian case study,

    A. Albatayneh, R. AbuAlRous, M. Kay, R. Abdallah, A. Juaidi, A. Garc ´ıa-Cruz, and F. Manzano-Agugliaro, “Wind farm capacity factor forecasting: An australian case study,”Energy Nexus, vol. 18, p. 100422, 2025. [Online]. Available: https://www.sciencedirect.com/ science/articl...

    2025

  39. [47]

    Life cycle assessment of floating offshore wind farms: An evaluation of operation and maintenance,

    A. Garcia-Teruel, G. Rinaldi, P. R. Thies, L. Johanning, and H. Jeffrey, “Life cycle assessment of floating offshore wind farms: An evaluation of operation and maintenance,”Applied Energy, vol. 307, p. 118067, 2022. [Online]. Available: https: //www.sciencedirect.com/science/a...

    2022

  40. [48]

    Carbon Aware Computing: Next Green Breakthrough or New Greenwashing?

    I. Velasco, “Carbon Aware Computing: Next Green Breakthrough or New Greenwashing?” 01

  41. [49]

    Renewable Energy Market Update –Outlook for 2023 and 2024,

    International Energy Agency, “Renewable Energy Market Update –Outlook for 2023 and 2024,” Jun 2023. [Online]. Available: https: //www.iea.org/reports/renewable-energy-market-update-june-2023/ will-more-wind-and-solar-pv-capacity-lead-to-more-generation-curtailment

    2023

  42. [50]

    Emissions and energy efficiency on large-scale high performance computing facilities: Archer2 uk national supercomputing service case study,

    A. Jackson, A. Simpson, and A. Turner, “Emissions and energy efficiency on large-scale high performance computing facilities: Archer2 uk national supercomputing service case study,” inProceedings of the SC23 Workshops of the International Conference on High Performance Computi...

    2023

  43. [51]

    How does wind farm performance decline with age?

    I. Staffell and R. Green, “How does wind farm performance decline with age?”Renewable Energy, vol. 66, pp. 775–786, 2014. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ S0960148113005727

    2014

  44. [53]

    Available: https://hackernoon.com/ carbon-aware-computing-next-green-breakthrough-or-new-greenwashing

    [Online]. Available: https://hackernoon.com/ carbon-aware-computing-next-green-breakthrough-or-new-greenwashing

  45. [56]

    A decision support model for waste heat recovery systems design in data center and high-performance computing clusters utilizing liquid cooling and phase change materials,

    V . Ljungdahl, M. Jradi, and C. Veje, “A decision support model for waste heat recovery systems design in data center and high-performance computing clusters utilizing liquid cooling and phase change materials,”Applied Thermal Engineering, vol. 201, p. 117671, 2022. [Online]. ...

    2022

  46. [2019]

    Available: https://www.anandtech.com/show/13999/ sk-hynix-details-its-ddr56400-dram-chip

    [Online]. Available: https://www.anandtech.com/show/13999/ sk-hynix-details-its-ddr56400-dram-chip

  47. [2024]

    Available: https://www.mckinsey.com/ industries/technology-media-and-telecommunications/our-insights/ ai-power-expanding-data-center-capacity-to-meet-growing-demand

    [Online]. Available: https://www.mckinsey.com/ industries/technology-media-and-telecommunications/our-insights/ ai-power-expanding-data-center-capacity-to-meet-growing-demand

  48. [2025]

    Available: https://codeberg.org/wimvanderbauwhede/ carbon-aware-geographic-load-shifting-model

    [Online]. Available: https://codeberg.org/wimvanderbauwhede/ carbon-aware-geographic-load-shifting-model

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.