{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,24]],"date-time":"2026-04-24T15:03:42Z","timestamp":1777043022075,"version":"3.51.4"},"reference-count":41,"publisher":"MDPI AG","issue":"4","license":[{"start":{"date-parts":[[2025,3,30]],"date-time":"2025-03-30T00:00:00Z","timestamp":1743292800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"Toshiba Europe Ltd."},{"name":"Bristol Research and Innovation Laboratory (BRIL)"},{"name":"Future Open Networks Research Challenge (FONRC) sponsored by the Department of Science Innovation and Technology (DSIT)"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Information"],"abstract":"<jats:p>This study presents an empirical investigation into the energy consumption of discriminative and generative AI models within real-world MLOps pipelines. For discriminative models, we examine various architectures and hyperparameters during training and inference and identify energy-efficient practices. For generative AI, large language models (LLMs) are assessed, with a focus primarily on energy consumption across different model sizes and varying service requests. Our study employs software-based power measurements, ensuring ease of replication across diverse configurations, models, and datasets. We analyse multiple models and hardware setups to uncover correlations among various metrics, identifying key contributors to energy consumption. The results indicate that, for discriminative models, optimising architectures, hyperparameters, and hardware can significantly reduce energy consumption without sacrificing performance. For LLMs, energy efficiency depends on balancing model size, reasoning complexity, and request-handling capacity, as larger models do not necessarily consume more energy when utilisation remains low. This analysis provides practical guidelines for designing green and sustainable ML operations, emphasising energy consumption and carbon-footprint reductions while maintaining performance. This paper can serve as a benchmark for accurately estimating total energy use across different types of AI models.<\/jats:p>","DOI":"10.3390\/info16040281","type":"journal-article","created":{"date-parts":[[2025,3,31]],"date-time":"2025-03-31T03:25:02Z","timestamp":1743391502000},"page":"281","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":7,"title":["Green MLOps to Green GenOps: An Empirical Study of Energy Consumption in Discriminative and Generative AI Operations"],"prefix":"10.3390","volume":"16","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-0821-4057","authenticated-orcid":false,"given":"Adri\u00e1n","family":"S\u00e1nchez-Momp\u00f3","sequence":"first","affiliation":[{"name":"Bristol Research and Innovation Laboratory, Toshiba Europe Ltd., Bristol BS1 4ND, UK"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3309-132X","authenticated-orcid":false,"given":"Ioannis","family":"Mavromatis","sequence":"additional","affiliation":[{"name":"Digital Catapult, London NW1 2RA, UK"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1516-1993","authenticated-orcid":false,"given":"Peizheng","family":"Li","sequence":"additional","affiliation":[{"name":"Bristol Research and Innovation Laboratory, Toshiba Europe Ltd., Bristol BS1 4ND, UK"}]},{"given":"Konstantinos","family":"Katsaros","sequence":"additional","affiliation":[{"name":"Digital Catapult, London NW1 2RA, UK"}]},{"given":"Aftab","family":"Khan","sequence":"additional","affiliation":[{"name":"Bristol Research and Innovation Laboratory, Toshiba Europe Ltd., Bristol BS1 4ND, UK"}]}],"member":"1968","published-online":{"date-parts":[[2025,3,30]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"48","DOI":"10.1109\/MTS.2020.2991496","article-title":"Estimating Carbon Emissions of Artificial Intelligence [Opinion]","volume":"39","author":"Luccioni","year":"2020","journal-title":"IEEE Technol. Soc. Mag"},{"key":"ref_2","doi-asserted-by":"crossref","unstructured":"Kathikeyan, T., Revathi, S., Supreeth, B.R., Sasidevi, J., Ahmed, M., and Das, S. (2022, January 14\u201316). Artificial Intelligence and Mixed Reality Technology for Interactive Display of Images in Smart Area. Proceedings of the 2022 5th International Conference on Contemporary Computing and Informatics (IC3I), Uttar Pradesh, India.","DOI":"10.1109\/IC3I56241.2022.10072411"},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"5","DOI":"10.1007\/s41233-022-00052-1","article-title":"Immersive Media Experience: A Survey of Existing Methods and Tools for Human Influential Factors Assessment","volume":"7","author":"Moinnereau","year":"2022","journal-title":"Qual. User Exp."},{"key":"ref_4","doi-asserted-by":"crossref","first-page":"114820","DOI":"10.1016\/j.eswa.2021.114820","article-title":"Machine Learning for industrial applications: A comprehensive literature review","volume":"175","author":"Bertolini","year":"2021","journal-title":"Expert Syst. Appl."},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"280","DOI":"10.1109\/OJCS.2023.3300321","article-title":"A Survey on ChatGPT: AI\u2013Generated Contents, Challenges, and Solutions","volume":"4","author":"Wang","year":"2023","journal-title":"IEEE Open J. Comput. Soc"},{"key":"ref_6","unstructured":"Li, P., S\u00e1nchez-Momp\u00f3, A., Farnham, T., Khan, A., and Aijaz, A. (2024). Large Generative AI Models meet Open Networks for 6G: Integration, Platform, and Monetization. arXiv."},{"key":"ref_7","doi-asserted-by":"crossref","first-page":"178586","DOI":"10.1109\/ACCESS.2024.3507186","article-title":"AI-Native Multi-Access Future Networks\u2014The REASON Architecture","volume":"12","author":"Katsaros","year":"2024","journal-title":"IEEE Access"},{"key":"ref_8","doi-asserted-by":"crossref","first-page":"18","DOI":"10.1109\/MC.2022.3148714","article-title":"The Carbon Footprint of Machine Learning Training Will Plateau, Then Shrink","volume":"55","author":"Patterson","year":"2022","journal-title":"Computer"},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"54","DOI":"10.1145\/3381831","article-title":"Green AI","volume":"63","author":"Schwartz","year":"2020","journal-title":"Commun. ACM"},{"key":"ref_10","doi-asserted-by":"crossref","first-page":"e1507","DOI":"10.1002\/widm.1507","article-title":"A Systematic Review of Green AI","volume":"13","author":"Verdecchia","year":"2023","journal-title":"WIREs Data Min. Knowl. Discov."},{"key":"ref_11","doi-asserted-by":"crossref","unstructured":"Singh, A., Patel, N.P., Ehtesham, A., Kumar, S., and Khoei, T.T. (2024). A Survey of Sustainability in Large Language Models: Applications, Economics, and Challenges. arXiv.","DOI":"10.1109\/CCWC62904.2025.10903774"},{"key":"ref_12","doi-asserted-by":"crossref","unstructured":"Yang, T.J., Chen, Y.H., and Sze, V. (2017, January 21\u201326). Designing Energy-Efficient Convolutional Neural Networks Using Energy-Aware Pruning. Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA.","DOI":"10.1109\/CVPR.2017.643"},{"key":"ref_13","doi-asserted-by":"crossref","unstructured":"Eliezer, N.S., Banner, R., Ben-Yaakov, H., Hoffer, E., and Michaeli, T. (2022, January 23\u201327). Power Awareness In Low Precision Neural Networks. Proceedings of the Computer Vision\u2014ECCV 2022 Workshops, Tel Aviv, Israel.","DOI":"10.1007\/978-3-031-25082-8_5"},{"key":"ref_14","unstructured":"de Reus, P., Oprescu, A., and Zuidema, J. (2024). An Exploration of the Effect of Quantisation on Energy Consumption and Inference Time of StarCoder2. arXiv."},{"key":"ref_15","unstructured":"Cottier, B., Rahman, R., Fattorini, L., Maslej, N., and Owen, D. (2024). The rising costs of training frontier AI models. arXiv."},{"key":"ref_16","unstructured":"Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.A., Lacroix, T., Rozi\u00e8re, B., Goyal, N., Hambro, E., and Azhar, F. (2023). LLaMA: Open and Efficient Foundation Language Models. arXiv."},{"key":"ref_17","first-page":"795","article-title":"Sustainable AI: Environmental Implications, Challenges and Opportunities","volume":"4","author":"Wu","year":"2022","journal-title":"Proc. Mach. Learn. Syst."},{"key":"ref_18","doi-asserted-by":"crossref","unstructured":"Islam, M.S., Zisad, S.N., Kor, A.L., and Hasan, M.H. (2023, January 23\u201325). Sustainability of Machine Learning Models: An Energy Consumption Centric Evaluation. Proceedings of the 2023 International Conference on Electrical, Computer and Communication Engineering (ECCE), Chittagong, Bangladesh.","DOI":"10.1109\/ECCE57851.2023.10101532"},{"key":"ref_19","first-page":"13693","article-title":"Energy and Policy Considerations for Modern Deep Learning Research","volume":"34","author":"Strubell","year":"2020","journal-title":"Proc. AAAI Conf. Artif. Intell."},{"key":"ref_20","doi-asserted-by":"crossref","unstructured":"Samsi, S., Zhao, D., McDonald, J., Li, B., Michaleas, A., Jones, M., Bergeron, W., Kepner, J., Tiwari, D., and Gadepally, V. (2023, January 25\u201329). From Words to Watts: Benchmarking the Energy Costs of Large Language Model Inference. Proceedings of the 2023 IEEE High Performance Extreme Computing Conference (HPEC), Boston, MA, USA.","DOI":"10.1109\/HPEC58863.2023.10363447"},{"key":"ref_21","unstructured":"Husom, E.J., Goknil, A., Shar, L.K., and Sen, S. (2024). The Price of Prompting: Profiling Energy Use in Large Language Models Inference. arXiv."},{"key":"ref_22","unstructured":"Li, P., Mavromatis, I., Farnham, T., Aijaz, A., and Khan, A. (2024). Adapting MLOps for Diverse In-Network Intelligence in 6G Era: Challenges and Solutions. arXiv."},{"key":"ref_23","doi-asserted-by":"crossref","first-page":"63606","DOI":"10.1109\/ACCESS.2022.3181730","article-title":"MLOps: A Taxonomy and a Methodology","volume":"10","author":"Testi","year":"2022","journal-title":"IEEE Access"},{"key":"ref_24","doi-asserted-by":"crossref","unstructured":"Teo, T.W., Chua, H.N., Jasser, M.B., and Wong, R.T. (2024, January 1\u20132). Integrating Large Language Models and Machine Learning for Fake News Detection. Proceedings of the 2024 20th IEEE International Colloquium on Signal Processing and Its Applications, CSPA 2024\u2014Conference Proceedings, Langkawi, Malaysia.","DOI":"10.1109\/CSPA60979.2024.10525308"},{"key":"ref_25","unstructured":"Satorras, V.G., Akata, Z., and Welling, M. (2019). Combining Generative and Discriminative Models for Hybrid Inference. arXiv."},{"key":"ref_26","doi-asserted-by":"crossref","first-page":"3581","DOI":"10.1109\/JSAC.2024.3459037","article-title":"Generative AI Agents With Large Language Model for Satellite Networks via a Mixture of Experts Transmission","volume":"42","author":"Zhang","year":"2024","journal-title":"IEEE J. Sel. Areas Commun."},{"key":"ref_27","doi-asserted-by":"crossref","unstructured":"Mavromatis, I., Katsaros, K., and Khan, A. (2024, January 1\u20133). Computing Within Limits: An Empirical Study of Energy Consumption in ML Training and Inference. Proceedings of the International Scientific Conference on Information, Communication and Energy Systems and Technologies (ICEST 2024)\u2014Workshop on Artificial Intelligence for Sustainable Development (ARISDE 2024), Sozopol, Bulgaria.","DOI":"10.1109\/ICEST62335.2024.10639701"},{"key":"ref_28","doi-asserted-by":"crossref","unstructured":"Conti, G., Jimenez, D., del Rio, A., Castano-Solis, S., Serrano, J., and Fraile-Ardanuy, J. (2023). A Multi-Port Hardware Energy Meter System for Data Centers and Server Farms Monitoring. Sensors, 23.","DOI":"10.3390\/s23010119"},{"key":"ref_29","doi-asserted-by":"crossref","unstructured":"Rinaldi, S., Bonafini, F., Ferrari, P., Flammini, A., Pasetti, M., and Sisinni, E. (2019, January 22\u201327). Software-based Time Synchronization for Integrating Power Hardware in the Loop Emulation in IEEE1588 Power Profile Testbed. Proceedings of the 2019 IEEE International Symposium on Precision Clock Synchronization for Measurement, Control, and Communication (ISPCS),  Portland, OR, USA.","DOI":"10.1109\/ISPCS.2019.8886644"},{"key":"ref_30","doi-asserted-by":"crossref","first-page":"1045","DOI":"10.1016\/j.ins.2020.09.033","article-title":"A Hardware-aware CPU Power Measurement Based on the Power-exponent Function model for Cloud Servers","volume":"547","author":"Lin","year":"2021","journal-title":"Inf. Sci."},{"key":"ref_31","unstructured":"NVIDIA Corporation (2016). nvidia-smi.txt."},{"key":"ref_32","doi-asserted-by":"crossref","unstructured":"Katsenou, A., Mao, J., and Mavromatis, I. (2022, January 7\u20139). Energy-Rate-Quality Tradeoffs of State-of-the-Art Video Codecs. Proceedings of the 2022 Picture Coding Symposium (PCS), San Jose, CA, USA.","DOI":"10.1109\/PCS56426.2022.10017999"},{"key":"ref_33","doi-asserted-by":"crossref","unstructured":"Vogelsang, T. (2010, January 4\u20138). Understanding the Energy Consumption of Dynamic Random Access Memories. Proceedings of the Annual IEEE\/ACM International Symposium on Microarchitecture, Atlanta, GA, USA.","DOI":"10.1109\/MICRO.2010.42"},{"key":"ref_34","doi-asserted-by":"crossref","unstructured":"Teo, J., and Chia, J.T. (2018, January 11\u201312). Deep Neural Classifiers For Eeg-Based Emotion Recognition In Immersive Environments. Proceedings of the 2018 International Conference on Smart Computing and Electronic Enterprise (ICSCEE), Shah Alam, Malaysia.","DOI":"10.1109\/ICSCEE.2018.8538382"},{"key":"ref_35","first-page":"151","article-title":"Image Classification Methods Applied in Immersive Environments for Fine Motor Skills Training in Early Education","volume":"5","author":"Riano","year":"2019","journal-title":"Int. J. Interact. Multimed. Artif. Intell."},{"key":"ref_36","unstructured":"Krizhevsky, A. (2009). Learning Multiple Layers of Features from Tiny Images, University of Toronto."},{"key":"ref_37","doi-asserted-by":"crossref","unstructured":"Mavromatis, I., De Feo, S., Carnelli, P., Piechocki, R.J., and Khan, A. (2023, January 6\u20138). FROST: Towards Energy-efficient AI-on-5G Platforms\u2014A GPU Power Capping Evaluation. Proceedings of the 2023 IEEE Conference on Standards for Communications and Networking (CSCN), Munich, Germany.","DOI":"10.1109\/CSCN60443.2023.10453214"},{"key":"ref_38","doi-asserted-by":"crossref","unstructured":"Aldin, N.B., and Aldin, S.S.A.B. (2022, January 29\u201331). Accuracy Comparison of Different Batch Size for a Supervised Machine Learning Task with Image Classification. Proceedings of the 2022 9th International Conference on Electrical and Electronics Engineering (ICEEE), Alanya, Turkey.","DOI":"10.1109\/ICEEE55327.2022.9772551"},{"key":"ref_39","doi-asserted-by":"crossref","unstructured":"Kwon, W., Li, Z., Zhuang, S., Sheng, Y., Zheng, L., Yu, C.H., Gonzalez, J.E., Zhang, H., and Stoica, I. (2023, January 23\u201326). Efficient Memory Management for Large Language Model Serving with PagedAttention. Proceedings of the 29th Symposium on Operating Systems Principles, Koblenz, Germany.","DOI":"10.1145\/3600006.3613165"},{"key":"ref_40","unstructured":"Zheng, L., Chiang, W.L., Sheng, Y., Zhuang, S., Wu, Z., Zhuang, Y., Lin, Z., Li, Z., Li, D., and Xing, E.P. (2023). Judging LLM-as-a-judge with MT-Bench and Chatbot Arena. arXiv."},{"key":"ref_41","unstructured":"Ye, X. (2023). calflops: A FLOPs and Params calculate tool for neural networks in pytorch framework."}],"container-title":["Information"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2078-2489\/16\/4\/281\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,9]],"date-time":"2025-10-09T17:06:01Z","timestamp":1760029561000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2078-2489\/16\/4\/281"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,3,30]]},"references-count":41,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2025,4]]}},"alternative-id":["info16040281"],"URL":"https:\/\/doi.org\/10.3390\/info16040281","relation":{},"ISSN":["2078-2489"],"issn-type":[{"value":"2078-2489","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,3,30]]}}}