Pith. sign in

REVIEW 4 major objections 4 minor 2 cited by

Exploring the sustainable scaling of AI dilemma: A projective study of corporations' AI environmental impacts

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper proposes a portfolio-level method to estimate AI environmental impacts, showing that large generative models use up to 4,600 times more energy per inference and that unconstrained adoption could raise AI electricity use…

desk verdict A genuinely integrative corporate AI footprint framework, but the headline 2030 numbers rest on an unvalidated 2024 model-size baseline and no released simulator. read the letter →

arxiv 2501.14334 v2 pith:R4AJQIUE submitted 2025-01-24 cs.AI cs.CYcs.LG

Clément Desroches , Martin Chauvin , Louis Ladan , Caroline Vateau , Simon Gosset , Philippe Cordier This is my paper · ORCID
classification cs.AIcs.CYcs.LG
keywords AIenvironmentalimpactgenerativelifecycleassessmentenergyconsumption2030scenariocorporateportfoliosustainabilityLLMinference
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 builds a practical, top-down method for estimating the environmental footprint of a company's entire AI portfolio without requiring deep LCA or AI expertise. It anchors the method on a typical large-company portfolio of 100 use cases and shows that generative AI already dominates the energy bill: large-model chat inference uses up to 4,600 times more electricity than a conventional NLP inference, and RAG/agent workflows push that to 25,000 times. It then projects the same portfolio to 2030 under four boundary scenarios and an intermediate scenario. The headline result is that an unconstrained high-adoption path raises AI electricity use by a factor of 24.4, while a frugal-efficiency path could cut it by 70%, and that no single efficiency lever reaches a 90% GHG reduction alone. The authors argue for standardized assessment, provider transparency, and a Return on Environment metric as the practical consequences.

What carries the argument

The carrying mechanism is a four-sub-model projection chain. First, an LCA model assigns embodied and operational impacts to three capacity types—compute, storage, and network—using a bill-of-materials for a cloud server with 8 A100-class GPUs. Second, a use-case clustering scheme splits AI into 192 clusters by AI type, task (chat, RAG, agents, tabular, computer vision, NLP), model-size bucket, user count, and usage frequency. Third, a representative company portfolio distributes use cases as 71% traditional AI and 29% GenAI, with the GenAI model-size mix anchored on open download counts scaled by a 6x closed-provider usage factor. Fourth, 2030 scenarios combine usage-growth CAGRs with systemic-efficiency factors for hardware FLOPs/W, PUE, grid decarbonization, quantization, model size, and output tokens. The per-inference impact formula sums contributions over the two lifecycle steps (fine-tuning and inference), the four component categories, and the two embodied/operational stages, which lets the same machinery compute GHG, water, primary energy, resource depletion, and final electricity use for any portfolio.

What would settle it

Run the three reference Llama models on a p4de.24xlarge instance and measure energy per chat inference; the model predicts roughly 0.093 Wh (8B), 1.55 Wh (70B), and 17.3 Wh (405B), and measured values more than about twice those numbers would break the 4,600x ratio and the portfolio projections built on it.

Watch

Extended reading notes

Core claim

The paper's central claim is that a company-level AI environmental footprint can be approximated from a small set of publicly available parameters—model size, use-case type, user count, usage frequency, and geographic distribution—and that doing so reveals a scale problem. Per-inference calculations show a high-size generative chat inference at about 17 Wh versus 0.0037 Wh for a traditional NLP inference, a 4,600-fold gap; high-size agentic workflows reach roughly 96 Wh, about 25,000 times the traditional baseline. Aggregated over a representative 100-use-case portfolio, generative AI accounts for 99.9% of inference energy despite being only 29% of use cases. Projecting to 2030, the high-adoption scenario lifts portfolio electricity use by factor 24.4 and GHG emissions by 18.6, whereas a scenario combining moderate adoption with ambitious hardware and grid improvements cuts energy by 70%. The authors further show that embodied impacts dominate water use (around 30%) and resource depletion (89%) while remaining small for GHG (5%), meaning carbon-only accounting misses the other environmental dimensions.

Load-bearing premise

The load-bearing premise is that the 2024 representative portfolio—71% traditional AI, 29% GenAI, with 85.8% of GenAI calls in the largest model-size bucket—and the hand-selected 2030 adoption rates (up to 47% GenAI CAGR and 55% agentic CAGR) reflect reality; if the true mix is smaller models or slower adoption, the 24.4x headline shrinks roughly proportionally.

Editorial extensions

If this is right

  • If the methodology is adopted, companies can estimate AI footprint from public parameters without waiting for providers to disclose internal data, lowering the barrier to net-zero accounting.
  • The per-inference ratios imply that shifting a workload from traditional NLP to a large generative or agentic system changes the energy profile by orders of magnitude, making use-case selection a first-order sustainability lever.
  • If the high-adoption 2030 scenario holds, corporate AI electricity scales by 24.4x; even the intermediate path (about 7.6x) strains net-zero targets without additional efficiency gains.
  • The 175x to 565x hardware efficiency improvement required for a 90% GHG reduction under high adoption indicates that hardware progress alone cannot offset usage growth, so adoption limits, frugal model design, grid decarbonization, and transparency are structural necessities.
  • Because embodied impacts dominate resource depletion and a substantial share of water use, focusing only on operational GHG reductions can leave water and minerals problems unaddressed, supporting the paper's call for multi-criteria assessment and an eco-score.

Reading between the lines

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

  • Beyond the paper's own proposals, the same 192-cluster taxonomy could be turned into a public task-level eco-score for any model, making energy per task the unit of comparison rather than a model-level rating.
  • The projection is conditional on the 85.8% high-model-size share, itself derived from open download counts times a 6x closed-provider usage factor; direct telemetry from closed providers would tighten this number more than any other single measurement.
  • The Return on Environment metric is only outlined; a natural completion is to pair it with consequential LCA so that AI's indirect energy savings are weighed against its direct footprint, a step the paper explicitly leaves out.
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

4 major / 4 minor

Summary. This manuscript proposes a four-layer methodology for estimating the environmental impacts of a company's AI portfolio: a life-cycle assessment model for hardware, an AI use-case clustering model, a representative corporate portfolio model, and a set of 2030 projection scenarios. The authors calibrate per-inference energy figures against external measurements from Luccioni et al. and Artificial Analysis, report per-inference impacts across GHG, water, primary energy, and resource depletion, and present scenario indices relative to a fictional 2024 portfolio. Headline results are that large generative AI models consume up to about 4,600x more energy per inference than traditional NLP models and that the "High adoption without boundaries" 2030 scenario yields a 24.4x increase in AI electricity use relative to 2024. The paper also contains a mitigation thought experiment suggesting that 175x-565x hardware efficiency improvements would be needed to offset 90% GHG reductions under high adoption scenarios.

Significance. The paper's main value is practical: it offers a parameterized, multi-criteria framework that companies could use without deep LCA expertise, and it is anchored by a genuine external calibration to Luccioni et al. The scenario logic is transparent, and the attention to water, resource depletion, and embodied impacts goes beyond the usual carbon-only analyses. If the assumptions are made fully testable, the framework could be a useful decision-support tool for sustainability practitioners. However, the headline 24.4x factor is not an empirical forecast but an index built on hand-set growth assumptions and a fragile 2024 portfolio baseline; the current sensitivity analysis does not stress-test the most assumption-heavy input. With corrections, released data, and clearly conditional framing, the contribution would be sound; as it stands, several central quantitative claims need further support.

major comments (4)
  1. [Results, Table 5] Table 5 contains internally inconsistent row labels. The row labeled "High adoption without boundaries" reports 755/576/650 for energy/GHG/water, which are exactly the Intermediate scenario values in Table 2, while the row labeled "Intermediate scenario" reports 2440/1862/2102, the high-adoption values. The two "Offset scenario" rows then associate the 565x and 175x hardware-efficiency factors in an order that contradicts the Discussion's statement that the high-adoption scenario requires 565x efficiency. Please relabel the rows and reassign the 565x/175x factors so the thought experiment is internally consistent.
  2. [Supplementary, Company Portfolio Model, Model Size (Table 25)] The 2024 baseline places 85.8% of GenAI usage in the High model-size bucket. This share is derived by adding an estimated GPT-4 "downloads" count of about 31.4M on the assumption that OpenAI has 6x Ollama's usage, assigning all OpenAI traffic to the High bucket, and comparing with Llama-3.1 Hugging Face downloads. Because per-inference energy for High chat is about 186x Low chat (Table 1), this single assumption dominates both the 2024 denominator and the 2030 numerator of every scenario index, including the 24.4x headline. The sensitivity analysis in Table 3 perturbs only the 2030 model-size evolution and never varies the 2024 size-mix baseline. A direct sensitivity test over plausible 2024 High/Low/Medium shares is needed before the headline factor can be considered robust.
  3. [Abstract and Results, Table 2] The abstract says the model "forecasts AI electricity use up to 2030" and that AI electricity use "is projected to rise by a factor of 24.4," but the 24.4x is the output of a deliberately extreme scenario with hand-set CAGRs of 47% for GenAI and 55% for agentic use cases, a threefold model-size increase, and a threefold output-token increase (Tables 2, 28-31). No uncertainty intervals or alternative-parameter ranges are reported for this scenario index. Please state consistently that these are conditional "if-then" projections, and report a sensitivity of the 24.4x factor to the main CAGR and token assumptions, as is already done for the Intermediate scenario.
  4. [Methods, Model 3 and Supplementary, Use Case usage] The statistical portfolio that anchors all results is constructed from non-public Capgemini data: a list of 350+ client use cases labeled with an LLM, and "typical company" usage frequencies calibrated to Capgemini experience. The "excel simulator" mentioned in Methods is not provided, so an independent reader cannot recompute the 2024 portfolio or the scenario indices. Please release the simulator or a parameterized, anonymized version with all distributions explicitly tabulated, or clearly mark the relevant sections as illustrative rather than reproducible.
minor comments (4)
  1. [Results, before Table 2] The text contains the placeholder "Error! Reference source not found." in place of a citation to Table 2; this must be fixed.
  2. [Supplementary, 2030 Systemic projections, Model Efficiency] The text refers to "Paccou et al. study6," but reference 6 is Wijnhoven and Paccou; please correct the citation style.
  3. [Figure 2 caption] The caption states twice that "CAGR, by definition, represents exponential growth over time"; please remove the redundant clause.
  4. [Discussion, Return on Environment] The "Return on Environment" metric is introduced as a recommendation but is not defined quantitatively; a short definition or reference would help readers understand what is being proposed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the per-inference ratios and 2030 scenario indices are computed from stated external benchmarks and explicit scenario assumptions, not fitted to their own targets.

full rationale

The paper's claimed results are not equivalent to its inputs by construction. Per-inference energies for Traditional AI are taken from Luccioni et al.'s external measurements, and GenAI latencies and throughputs are taken from Artificial Analysis; the hardware LCA uses a bill-of-materials model with NegaOctet data. The 4600x ratio is the ratio of the authors' computed high-size chat inference energy (1.73e-2 kWh) to their computed traditional NLP inference energy (3.70e-6 kWh), so it is a model output, not a fitted parameter renamed as a prediction. The 24.4x 2030 index is a conditional scenario calculation: the paper explicitly states the assumed CAGRs (47% GenAI, 55% agentic in the high-adoption case) and the efficiency and electricity-mix factors, and it reports sensitivity analyses for 2030 model size and agentic penetration. The 2024 portfolio baseline, including the 85.8% High model-size share derived from HuggingFace downloads plus a 6x OpenAI usage proxy, is an input assumption. That assumption is fragile and not stress-tested at the 2024 baseline, but fragility of an input is an uncertainty and robustness concern, not circularity. The only self-reference burden is the use of Capgemini internal use-case lists and Capgemini Research Institute reports for portfolio construction; these are not the source of the central per-inference or scenario results, and the paper's own limitations section acknowledges the uncertainty of its projection factors. No load-bearing step reduces to a self-citation or to the definition of the quantity being predicted.

Assumptions & free parameters 10 free parameters · 6 assumptions · 0 invented entities

The central results rest on many unmeasured, hand-set, or proprietary inputs. None of these parameters is fitted to reproduce the target conclusion; they are assumptions about deployment scale, usage frequency, model size, and efficiency that propagate through the LCA calculations. The axioms list the main external datasets and modeling premises that the paper relies on without independent verification.

free parameters (10)
  • 2024 GenAI model-size distribution = low 13.1%, medium 1.1%, high 85.8%
    Derived from Llama-3.1 Hugging Face downloads and the assumption that OpenAI GPT-4 usage is 6x Llama usage and maps to high size; drives weighted portfolio energy.
  • Use-case type distribution = Classic 71% (Tabular 79%, CV 11%, NLP 10%); GenAI 29% (Agents 33%, Chatbot 28%, RAG 39%)
    From a Capgemini internal list of 350+ use cases labelled by an LLM; determines the composition of the 2024 portfolio and all scenario extrapolations.
  • Usage and frequency distribution = Table 24 values, e.g. Classic AI: 80% low users; GenAI: 40% medium, 20% very high
    Hand-set from Capgemini experience; multiplies per-inference energy into annual portfolio energy.
  • GenAI and Agentic adoption CAGRs 2024-2030 = Intermediate 40%/45%; High 47%/55%; Low 32%/35%
    Averaged from market analyses; exponential drivers of the 2030 scenario results.
  • Model size evolution to 2030 = x3 in limited-efficiency scenarios; stabilization in high-efficiency scenarios
    Based on lifearchitect.ai trend analysis and small language model adoption; sensitivity analysis shows energy scales about 1:1 with model size.
  • Output token evolution to 2030 = x1.33 limited efficiency, x2 intermediate, x3 high adoption
    Assumed reasoning-model adoption; energy per inference scales about 1:1 with output tokens.
  • Hardware efficiency improvement = x4.4 continuous, x4.8 breakthrough
    From Epoch AI 1.28x/year historical trend and an assumed 20% share of Cerebras-like inference; central to the Limited growth scenario.
  • Grid emission reduction = -24% stated policies, -45% IPCC 1.5C pathways
    Converts electricity consumption to GHG emissions in scenarios; directly controls the GHG indices.
  • RAG token counts = 5333 input tokens, 363 output tokens
    From three unnamed production RAG projects; sets RAG per-inference energy and storage/network impact.
  • vGPU counts for reference LLMs = Llama 8B: 3 vGPUs; 70B: 19 vGPUs; 405B: 106 vGPUs
    Computed from FP16 memory with 1.3 overhead divided into 10GB MIG slices; the 405B count drives the 4600x energy ratio.
assumptions (6)
  • domain assumption The NegaOctet database and Ecoinvent emission factors correctly represent the embodied and operational impacts of cloud servers, storage, and networks across US/EU/China grids.
    Used for all LCA impact factors in Table 8; not validated against measured datacenter data in this paper.
  • domain assumption Artificial Analysis TTFT and throughput averages across AWS, GCP, and Azure are representative of enterprise GenAI inference performance.
    These speeds determine compute hours for every GenAI inference; no sensitivity analysis is run on throughput.
  • domain assumption Luccioni et al.'s Hugging Face model energy measurements represent traditional AI inference workloads, with a synthetic Random Forest simulation used for tabular data.
    Forms the baseline for the 25x and 4600x comparisons between GenAI and traditional AI.
  • domain assumption Global 2000 companies follow the same AI use-case distribution as the Capgemini-derived fictive portfolio.
    Used to extrapolate 7.8 TWh global AI electricity consumption in 2024; no empirical global portfolio data is provided.
  • domain assumption The historical 1.28x/year hardware efficiency improvement continues linearly through 2030.
    Underpins the x4.4 hardware efficiency factor in all non-breakthrough scenarios; slower efficiency growth would change all scenario indices.
  • domain assumption The FP16 memory model with 1.3 overhead and 10GB MIG vGPU slices gives correct vGPU counts for serving LLMs.
    Sets vGPU hours and energy per inference for GenAI; the paper defers the 1:2 FP16 sensitivity refinement to future work.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exploring the sustainable scaling of AI dilemma: A projective study of corporations' AI environmental impacts." pith.science (2026). https://pith.science/paper/R4AJQIUE

@misc{pith2026250114334,
  author       = {Pith},
  title        = {Pith review of: Exploring the sustainable scaling of AI dilemma: A projective study of corporations' AI environmental impacts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R4AJQIUE}},
  note         = {Machine review of arXiv:2501.14334}
}
read the original abstract

The rapid growth of artificial intelligence (AI), particularly Large Language Models (LLMs), has raised concerns regarding its global environmental impact that extends beyond greenhouse gas emissions to include consideration of hardware fabrication and end-of-life processes. The opacity from major providers hinders companies' abilities to evaluate their AI-related environmental impacts and achieve net-zero targets. In this paper, we propose a methodology to estimate the environmental impact of a company's AI portfolio, providing actionable insights without necessitating extensive AI and Life-Cycle Assessment (LCA) expertise. Results confirm that large generative AI models consume up to 4600x more energy than traditional models. Our modelling approach, which accounts for increased AI usage, hardware computing efficiency, and changes in electricity mix in line with IPCC scenarios, forecasts AI electricity use up to 2030. Under a high adoption scenario, driven by widespread Generative AI and agents adoption associated to increasingly complex models and frameworks, AI electricity use is projected to rise by a factor of 24.4. Mitigating the environmental impact of Generative AI by 2030 requires coordinated efforts across the AI value chain. Isolated measures in hardware efficiency, model efficiency, or grid improvements alone are insufficient. We advocate for standardized environmental assessment frameworks, greater transparency from the all actors of the value chain and the introduction of a "Return on Environment" metric to align AI development with net-zero goals.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Not All Water Consumption Is Equal: A Water Stress Weighted Metric for Sustainable Computing

    cs.DC 2025-06 conditional novelty 5.0 of 10

    SCARF is a water-stress-weighted impact metric that multiplies raw computing water use by a basin-level, time-discounted water stress factor, revealing large location and seasonal differences in the true water cost of...

  2. Irresponsible AI: big tech's influence on AI research and associated impacts

    cs.CY 2025-11 conditional novelty 2.0 of 10

    A review/position paper arguing that big tech's outsized influence on AI research channels the field toward scaling and general-purpose systems, producing environmental and social harms that technical fixes alone cann...

Reference graph

Works this paper leans on

81 extracted references · 60 canonical work pages · cited by 2 Pith papers

  1. [1]

    Capgemini https://www.capgemini.com/insights/research-library/sustainable-gen- ai/ (2024)

    Sustainable Gen AI. Capgemini https://www.capgemini.com/insights/research-library/sustainable-gen- ai/ (2024)

    2024

  2. [2]

    Pricing and Competition for Generative AI

    Mahmood, R. Pricing and Competition for Generative AI. Preprint at https://doi.org/10.48550/arXiv.2411.02661 (2024)

    work page Pith review arXiv doi:10.48550/arxiv.2411.02661 2024

  3. [3]

    & Lefèvre, L

    Berthelot, A., Caron, E., Jay, M. & Lefèvre, L. Estimating the environmental impact of Generative-AI services using an LCA-based methodology. Procedia CIRP 122, 707–712 (2024)

    work page 2024

  4. [4]

    Electricity 2024 – Analysis

    IEA. Electricity 2024 – Analysis. https://www.iea.org/reports/electricity-2024 (2024)

    work page 2024

  5. [5]

    https://www.deloitte.com/global/en/issues/climate/powering-ai.html

    Powering AI | Deloitte Global. https://www.deloitte.com/global/en/issues/climate/powering-ai.html

    work page

  6. [6]

    & Paccou, R

    Wijnhoven, F. & Paccou, R. Artificial Intelligence and Electricity: A System Dynamics Approach. in Schneider Electric/insights/sustainability 1–96 (Schneider Electric Sustainability Research Institute, France, 2024)

    work page 2024

  7. [7]

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

    AI data center growth: Meeting the demand | McKinsey. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/ai- power-expanding-data-center-capacity-to-meet-growing-demand

    work page

  8. [8]

    Int’l Legal Materials 55, 743 (2016)

    Paris Agreement. Int’l Legal Materials 55, 743 (2016)

    work page 2016

Show all 81 references
  1. [9]

    S., Jernite, Y

    Luccioni, A. S., Jernite, Y. & Strubell, E. Power Hungry Processing: Watts Driving the Cost of AI Deployment? in The 2024 ACM Conference on Fairness, Accountability, and Transparency 85–99 (2024). doi:10.1145/3630106.3658542

    2024

  2. [10]

    & Lannelongue, L

    Bouza, L., Bugeau, A. & Lannelongue, L. How to estimate carbon footprint when training deep learning models? A guide and review. Environ. Res. Commun. 5, 115014 (2023)

    2023

  3. [11]

    & Sze, V

    Yang, T.-J., Chen, Y.-H., Emer, J. & Sze, V. A method to estimate the energy consumption of deep neural networks. in 2017 51st Asilomar Conference on Signals, Systems, and Computers 1916–1920 (IEEE, Pacific Grove, CA, USA, 2017). doi:10.1109/ACSSC.2017.8335698

    2017

  4. [12]

    & Zong, Z

    Li, D., Chen, X., Becchi, M. & Zong, Z. Evaluating the Energy Efficiency of Deep Convolutional Neural Networks on CPUs and GPUs. in 2016 IEEE International Conferences on Big Data and Cloud Computing (BDCloud), Social Computing and Networking (SocialCom), Sustainable Computing...

    2016 doi

  5. [13]

    GPU-Based Deep Learning Inference: A Performance and Power Analysis

    Nvidia. GPU-Based Deep Learning Inference: A Performance and Power Analysis. https://www.nvidia.com/content/tegra/embedded-systems/pdf/jetson_tx1_whitepaper.pdf (2015)

    2015

  6. [14]

    & Culurciello, E

    Canziani, A., Paszke, A. & Culurciello, E. An Analysis of Deep Neural Network Models for Practical Applications. Preprint at https://doi.org/10.48550/arXiv.1605.07678 (2017)

    2017 doi arXiv

  7. [15]

    Wu, C.-J. et al. Sustainable AI: Environmental Implications, Challenges and Opportunities. Preprint at https://doi.org/10.48550/arXiv.2111.00364 (2022)

    2022 doi arXiv

  8. [16]

    & Inouye, M

    Lannelongue, L., Grealey, J. & Inouye, M. Green Algorithms: Quantifying the carbon footprint of computation. Preprint at https://doi.org/10.48550/arXiv.2007.07610 (2020)

    2020 doi arXiv

  9. [17]

    Henderson, P. et al. Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning. Preprint at https://doi.org/10.48550/arXiv.2002.05651 (2022)

    2022 doi arXiv

  10. [18]

    Grattafiori, A. et al. The Llama 3 Herd of Models. Preprint at https://doi.org/10.48550/arXiv.2407.21783 (2024)

    2024 doi arXiv

  11. [19]

    Luccioni, A. S. & Hernandez-Garcia, A. Counting Carbon: A Survey of Factors Influencing the Emissions of Machine Learning. Preprint at https://doi.org/10.48550/arXiv.2302.08476 (2023)

    2023 doi arXiv

  12. [20]

    & McCallum, A

    Strubell, E., Ganesh, A. & McCallum, A. Energy and Policy Considerations for Deep Learning in NLP. Preprint at https://doi.org/10.48550/arXiv.1906.02243 (2019)

    2019 doi arXiv

  13. [21]

    & Dandres, T

    Lacoste, A., Luccioni, A., Schmidt, V. & Dandres, T. Quantifying the Carbon Emissions of Machine Learning. Preprint at https://doi.org/10.48550/arXiv.1910.09700 (2019)

    2019 doi arXiv

  14. [22]

    Faiz, A. et al. LLMCarbon: Modeling the end-to-end Carbon Footprint of Large Language Models. Preprint at https://doi.org/10.48550/arXiv.2309.14393 (2024)

    2024 doi arXiv

  15. [23]

    Courty, B. et al. mlco2/codecarbon: v2.4.1. Zenodo https://doi.org/10.5281/zenodo.11171501 (2024)

    2024 doi

  16. [24]

    & Jiang, L

    Fu, Z., Chen, F., Zhou, S., Li, H. & Jiang, L. LLMCO2: Advancing Accurate Carbon Footprint Prediction for LLM Inferences. Preprint at https://doi.org/10.48550/arXiv.2410.02950 (2024)

    2024 doi arXiv

  17. [25]

    https://ecologits.ai/latest/methodology/llm_inference/ (2024)

    EcoLogits. https://ecologits.ai/latest/methodology/llm_inference/ (2024)

    2024

  18. [26]

    S., Viguier, S

    Luccioni, A. S., Viguier, S. & Ligozat, A.-L. Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language Model. Preprint at https://doi.org/10.48550/arXiv.2211.02001 (2022)

    2022 doi arXiv

  19. [27]

    Lozhkov, A. et al. StarCoder 2 and The Stack v2: The Next Generation. Preprint at https://doi.org/10.48550/arXiv.2402.19173 (2024)

    2024 doi arXiv

  20. [28]

    Cloud Carbon Footprint

    Thoughtworks, Inc. Cloud Carbon Footprint. Thoughtworks, Inc. (2020)

    2020

  21. [29]

    https://huggingface.co/spaces/ml- energy/leaderboard

    ML.ENERGY Leaderboard - a Hugging Face Space by ml-energy. https://huggingface.co/spaces/ml- energy/leaderboard

  22. [30]

    https://huggingface.co/spaces/optimum/llm-perf-leaderboard

    LLM-Perf Leaderboard - a Hugging Face Space by optimum. https://huggingface.co/spaces/optimum/llm-perf-leaderboard

  23. [31]

    Artificial Analysis https://artificialanalysis.ai

    AI Model & API Providers Analysis. Artificial Analysis https://artificialanalysis.ai

  24. [32]

    LangChain State of AI 2024 Report

    2024

  25. [33]

    The AI Index 2024 Annual Report

    Nestor Maslej et al. The AI Index 2024 Annual Report. https://aiindex.stanford.edu/wp- content/uploads/2024/05/HAI_AI-Index-Report-2024.pdf (2024)

    2024

  26. [34]

    Centres de données Google https://www.google.com/intl/fr/about/datacenters/efficiency/

    Efficacité – Centres de données – Google. Centres de données Google https://www.google.com/intl/fr/about/datacenters/efficiency/

  27. [35]

    IEA https://www.iea.org/reports/world-energy-outlook-2024 (2024)

    World Energy Outlook 2024 – Analysis. IEA https://www.iea.org/reports/world-energy-outlook-2024 (2024)

    2024

  28. [36]

    Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 Laying down Harmonised Rules on Artificial Intelligence and Amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and D...

    2024

  29. [37]

    Vogginger, B. et al. Neuromorphic hardware for sustainable AI data centers. Preprint at https://doi.org/10.48550/arXiv.2402.02521 (2024)

    2024 doi arXiv

  30. [38]

    Will businesses or laws and regulations ever prioritise environmental sustainability for AI systems? https://oecd.ai/en/wonk/businesses-regulations-environmental-sustainability

  31. [39]

    Luccioni, S. et al. Light bulbs have energy ratings — so why can’t AI chatbots? Nature 632, 736–738 (2024)

    2024

  32. [40]

    Score environnemental du véhicule https://score- environnemental-bonus.ademe.fr

    Score environnemental du véhicule. Score environnemental du véhicule https://score- environnemental-bonus.ademe.fr

  33. [41]

    What are carbon budgets and how can they help us reach net zero? World Economic Forum https://www.weforum.org/stories/2024/03/carbon-budget-climate-change-net-zero/ (2024)

    2024

  34. [42]

    Tschand, A. et al. MLPerf Power: Benchmarking the Energy Efficiency of Machine Learning Systems from Microwatts to Megawatts for Sustainable AI. Preprint at https://doi.org/10.48550/arXiv.2410.12032 (2024)

    2024 doi arXiv

  35. [43]

    Change, I. C. The physical science basis. (No Title) (2013)

    2013

  36. [44]

    Van Oers, L., De Koning, A., Guinée, J. B. & Huppes, G. Abiotic resource depletion in LCA. Road and Hydraulic Engineering Institute, Ministry of Transport and Water, Amsterdam (2002)

    2002

  37. [45]

    Boulay, A.-M. et al. The WULCA consensus characterization model for water scarcity footprints: assessing impacts of water consumption based on available water remaining (AWARE). Int J Life Cycle Assess 23, 368–378 (2018)

    2018

  38. [47]

    Harnessing the Value of Generative AI

    Capgemini Research Institute. Harnessing the Value of Generative AI. https://www.capgemini.com/wp-content/uploads/2024/05/Final-Web-Version-Report-Gen-AI-in- Organization-Refresh.pdf (2024)

    2024

  39. [48]

    in Climate Change 2022 - Mitigation of Climate Change (ed

    Energy Systems. in Climate Change 2022 - Mitigation of Climate Change (ed. Intergovernmental Panel On Climate Change (Ipcc)) 613–746 (Cambridge University Press, 2023). doi:10.1017/9781009157926.008

    2022 doi

  40. [49]

    2030 climate ambition

    up Europe’s, S. 2030 climate ambition. Investing in a climate-neutral future for the benefit of our people (2020)

    2020

  41. [50]

    Earles, J. M. & Halog, A. Consequential life cycle assessment: a review. The international journal of life cycle assessment 16, 445–453 (2011)

    2011

  42. [51]

    Attributional and consequential life cycle assessment

    Ekvall, T. Attributional and consequential life cycle assessment. in Sustainability Assessment at the 21st century (IntechOpen, 2019)

    2019

  43. [52]

    & Combaz, J

    Ligozat, A.-L., Lefevre, J., Bugeau, A. & Combaz, J. Unraveling the Hidden Environmental Impacts of AI Solutions for Environment Life Cycle Assessment of AI Solutions. Sustainability 14, (2022)

    2022

  44. [53]

    https://eplca.jrc.ec.europa.eu/EnvironmentalFootprint.html

    European Platform on LCA | EPLCA. https://eplca.jrc.ec.europa.eu/EnvironmentalFootprint.html

  45. [54]

    TechPowerUp https://www.techpowerup.com/gpu-specs/a100-pcie- 80-gb.c3821 (2025)

    NVIDIA A100 PCIe 80 GB Specs. TechPowerUp https://www.techpowerup.com/gpu-specs/a100-pcie- 80-gb.c3821 (2025)

    2025

  46. [55]

    https://costcalc.cloudoptimo.com/aws-pricing- calculator/ec2/p4de.24xlarge

    p4de.24xlarge Pricing and Specs: AWS EC2. https://costcalc.cloudoptimo.com/aws-pricing- calculator/ec2/p4de.24xlarge

  47. [56]

    https://www.chaynikam.info/en/Xeon_Platinum_8275CL.html

    Intel Xeon Platinum 8275CL. https://www.chaynikam.info/en/Xeon_Platinum_8275CL.html

  48. [57]

    https://www.supermicro.com/en/products/motherboard/X12SPA-TF

    X12SPA-TF | Motherboards | Products | Supermicro. https://www.supermicro.com/en/products/motherboard/X12SPA-TF

  49. [58]

    TSMC se lasse de jouer des rectangles dans des cercles ; le fondeur veut de nouveaux wafers ! Le Comptoir du Hardware //www.comptoir-hardware.com/actus/processeurs/47586-tsmc-se-lasse-de- jouer-des-rectangles-dans-des-ronds-le-fondeur-veut-de-nouveaux-wafers-.html (2024)

    2024

  50. [59]

    Die Per Wafer (free) Calculator - Trusted by Amkor and GF

    anysilicon. Die Per Wafer (free) Calculator - Trusted by Amkor and GF. AnySilicon https://anysilicon.com/die-per-wafer-formula-free-calculators/ (2013)

    2013

  51. [60]

    Analyse et optimisation du procédé de découpe de plaques de silicium

    Rouault de Coligny, P. Analyse et optimisation du procédé de découpe de plaques de silicium. (École Nationale Supérieure des Mines de Paris, 2002)

    2002

  52. [61]

    How Foundries Calculate Die Yield

    Sekar, V. How Foundries Calculate Die Yield. https://www.viksnewsletter.com/p/how-foundries- calculate-die-yield (2024)

    2024

  53. [62]

    https://blog.langchain.dev/langchain-state-of-ai-2024/ (2024)

    LangChain State of AI 2024 Report. https://blog.langchain.dev/langchain-state-of-ai-2024/ (2024)

    2024

  54. [63]

    https://www.cbre.com/insights/reports/global-data-center-trends- 2024

    Global Data Center Trends 2024. https://www.cbre.com/insights/reports/global-data-center-trends- 2024

    2024

  55. [64]

    https://support.ecoinvent.org/energy

    Energy. https://support.ecoinvent.org/energy

  56. [65]

    Uptime Institute https://uptimeinstitute.com/resources/research-and-reports/uptime-institute-global-data-center-survey- results-2024

    Uptime Institute Global Data Center Survey Results 2024. Uptime Institute https://uptimeinstitute.com/resources/research-and-reports/uptime-institute-global-data-center-survey- results-2024

    2024

  57. [66]

    Amazon goes nuclear, to invest more than $500 million to develop small modular reactors

    Olick, D. Amazon goes nuclear, to invest more than $500 million to develop small modular reactors. CNBC https://www.cnbc.com/2024/10/16/amazon-goes-nuclear-investing-more-than-500-million-to- develop-small-module-reactors.html (2024)

    2024

  58. [67]

    & Neelis, M

    Schlömer, S., Hänsel, G., de Jager, D. & Neelis, M. Technology-specific Cost and Performance Parameters

  59. [68]

    & Névéol, A

    Morand, C., Ligozat, A.-L. & Névéol, A. How Green Can AI Be? A Study of Trends in Machine Learning Environmental Impacts. Preprint at https://doi.org/10.48550/arXiv.2412.17376 (2024)

    2024 doi arXiv

  60. [69]

    Can AI Scaling Continue Through 2030? Epoch AI https://epoch.ai/blog/can-ai-scaling- continue-through-2030 (2024)

    Sevilla, J. Can AI Scaling Continue Through 2030? Epoch AI https://epoch.ai/blog/can-ai-scaling- continue-through-2030 (2024)

    2024

  61. [70]

    Villalobos, P. et al. Will we run out of data? Limits of LLM scaling based on human-generated data. Preprint at https://doi.org/10.48550/arXiv.2211.04325 (2024)

    2024 doi arXiv

  62. [71]

    These AI Models Are Pretty Mid

    Bousquette, I. These AI Models Are Pretty Mid. That’s Why Companies Love Them. Wall Street Journal (2024)

    2024

  63. [72]

    Small is the new big: The rise of small language models

    elishavalekar. Small is the new big: The rise of small language models. Capgemini https://www.capgemini.com/insights/expert-perspectives/small-is-the-new-big-the-rise-of-small- language-models/ (2024)

    2024

  64. [73]

    Wu, S. et al. A Comparative Study on Reasoning Patterns of OpenAI’s o1 Model. Preprint at https://doi.org/10.48550/arXiv.2410.13639 (2024)

    2024 doi arXiv

  65. [74]

    Statista https://www.statista.com/forecasts/1474143/global-ai-market- size

    Global AI market size 2030. Statista https://www.statista.com/forecasts/1474143/global-ai-market- size

  66. [75]

    https://www.grandviewresearch.com/industry-analysis/artificial-intelligence-ai-market

    Artificial Intelligence Market Size, Share, Growth Report 2030. https://www.grandviewresearch.com/industry-analysis/artificial-intelligence-ai-market

  67. [76]

    MarketsandMarkets https://www.marketsandmarkets.com/Market-Reports/ai-agents-market-15761548.html

    AI Agents Market Size, Share and Global Forecast to 2030. MarketsandMarkets https://www.marketsandmarkets.com/Market-Reports/ai-agents-market-15761548.html

  68. [77]

    https://www.mckinsey.com/capabilities/quantumblack/our- insights/the-state-of-ai

    The state of AI in early 2024 | McKinsey. https://www.mckinsey.com/capabilities/quantumblack/our- insights/the-state-of-ai

    2024

  69. [78]

    https://www.grandviewresearch.com/press-release/global-generative-ai-market

    Generative AI Market Size To Reach $109.37 Billion By 2030. https://www.grandviewresearch.com/press-release/global-generative-ai-market

  70. [79]

    Statista https://www.statista.com/forecasts/1449838/generative-ai-market-size-worldwide

    Generative AI market size worldwide 2030. Statista https://www.statista.com/forecasts/1449838/generative-ai-market-size-worldwide

  71. [80]

    Motherboard; mix of equipment, without processor or RAM

    Generative AI to Become a $1.3 Trillion Market by 2032, Research Finds | Press | Bloomberg LP. Bloomberg L.P. Acknowledgements We would like to express our deepest gratitude to the direction of Capgemini Invent for their sponsorship and insightful reviews of the paper, with pa...

    2013

  72. [2030]

    75% of organizations say that large-scale deployment of generative AI PoCs is a significant challenge

    This outlook anticipates the widespread adoption of next-generation architectures, such as Cerebras inference that runs models like Llama 3.1 20x faster than GPU according to their report31. Moreover, we see new appearing technological breakthrough with photonic-powered chips,...

    2022

  73. [2050]

    Kairos expects to deploy the first SMR by 2030, followed by further deployments through 2035

    In 2030, in scenarios limiting warming to 1.5°C (>50%) with no or limited overshoot, net CO2 and GHG emissions fall by 35–51% and 38–52% respectively.” Consequently, we used efficiency factor of 45% reflecting Paris agreement targets. Furthermore, additional grid efficiency ca...

    2010

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.