[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84821-en":3,"doc-seo-84821-105":29,"detail-sidebar-cat-0-en-105":90},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84821,2336464648322,"Aria","https://ap-avatar.wpscdn.com/avatar/2200025388227c56fec?_k=1778556882303663488",8,"Research & Report","Life Cycle Assessment of Pre-training the Lucie 7B Open-Source Large Language Model on the Jean Zay Supercomputer","Environmental impact estimates for large language model training are increasingly scrutinised, yet many reports focus narrowly on operational energy while providing limited visibility into manufacturing (embodied) emissions, water consumption, and the underlying high-performance computing infrastructure. This study performs a life cycle assessment of Lucie 7B pre-training, aligned with AFNOR SPEC 2314 “Frugal AI” and using Labos 1point5 for greenhouse gas accounting, covering data preparation through validation and the hardware life cycle from extraction to end-of-life.","arXiv :2607 .05408v 1 [ cs .CY] 9 Jun 2026  \nLife Cycle Assessment of Pre-training the Lucie 7B Open-Source Large Language Model on the Jean Zay  \nSupercomputer  \nMarc Léobet 1 Pierre-François Lavallée2  \nJean-Pierre Lorré3  \n1 Mens Data 2 CNRS / IDRIS 3 LINAGORA From OpenLLM-France Environmental Report – v1 – January 2026  \nAbstract  \nThe environmental impact of training large language models (LLMs) is increasingly scrutinised, yet most published estimates focus on operational energy and disclose little about manufacturing (embodied) emissions, water consumption, or the underlying highperformance computing (HPC) infrastructure. We present a life cycle assessment (LCA) of the pre-training of Lucie 7B, an open-source multilingual Foundation Model developed by the OpenLLM-France consortium and trained on the NVIDIA H100 partition of the Jean Zay supercomputer operated by IDRIS (CNRS) . The assessment is framed by the AFNOR SPEC 2314 “Frugal AI” reference and applies the Labos 1point5 methodology for greenhouse gas (GHG) accounting in computing. The study scope extends from data preparation to model validation, and integrates the full life cycle of the hardware infrastructure: manufacturing (including raw-material extraction), use (compute, temporary storage, system administration, cooling), and end-of-life.  \nWe report (i) an annual footprint of 417.5 t CO2 eq for the Jean Zay H100 partition, split almost equally between manufacturing and operation; (ii) an effective intensity of  \n36.7 g CO2 eq per H100 GPU-hour; (iii) a total training footprint of 21 t CO2 eq for Lucie 7B (574 564 H100 GPU-hours), inclusive of amortised hardware manufacturing; (iv) on-site water consumption of approximately 76 m3 for the training campaign and an annual Water Usage Effectiveness (WUE) of 0.07 L/kWh for IDRIS; (v) a heat-reuse factor (ERF) of 0.37 thanks to waste-heat recovery into the urban heating network. The study contributes one of the few publicly documented LCAs of an LLM training campaign that explicitly couples operational data with embodied emissions decomposed by subsystem (compute, storage, power chain, cooling), and discusses the implications for the design of frugal-by-construction AI systems in Europe.  \nKeywords: Life Cycle Assessment; Large Language Models; Pre-training; Embodied carbon; Water footprint; HPC; Frugal AI; AFNOR SPEC 2314; Jean Zay; Open-source AI.  \n1. Introduction  \nThe rapid scaling of deep learning models, and in particular of large language models (LLMs), has placed their environmental footprint at the center of academic, industrial, and regulatory debate [1–3] . Strubell et al. [1] first popularised the question for natural language processing,  \nmotivating a stream of work on operational energy use and CO 2eq emissions of training [2–4] . Subsequent contributions have extended the analytical perimeter to inference [5], data-centre water use [6, 7], embodied (manufacturing) emissions of computing hardware [8–10], and end-toend carbon modelling of model training and serving [11] . In parallel, methodological frameworks specifically targeting digital and AI systems have been proposed at the European level—most notably the AFNOR SPEC 2314 “General Reference Framework for Frugal AI” [12] — and tools such as the Labos 1point5 GHG estimator for computing [13] have provided reproducible per-CPU/per-GPU emission factors grounded in actual French HPC operations.  \nDespite this growing literature, publicly documented life cycle assessments (LCAs) of LLM training campaigns remain scarce. The most complete prior study is the work of Luccioni et al. on BLOOM-176B [14], which combines operational measurements with an explicit, if partial, accounting of hardware manufacturing. Patterson et al. [2, 3] disclose detailed operational figures for Google models (T5, GShard, GPT-3, Switch Transformer) but exclude embodied emissions. Touvron et al. [4] disclose energy and operational CO 2eq for the LLaMA family, again restricted to","cbCaidjs9JBRQQFy","https://ap.wps.com/l/cbCaidjs9JBRQQFy","pdf",541202,1,12,"English","en",105,"# Abstract\n# Introduction\n## Research gap and prior work comparison","[{\"question\":\"What life cycle stages does the study include for Lucie 7B pre-training?\",\"answer\":\"The scope covers data preparation through model validation, and integrates the full hardware life cycle: manufacturing, use (compute, storage, administration, cooling), and end-of-life.\"},{\"question\":\"How are greenhouse gases accounted for in the assessment?\",\"answer\":\"The study follows AFNOR SPEC 2314 “Frugal AI” as the framing reference and applies the Labos 1point5 methodology for GHG accounting in computing.\"},{\"question\":\"What quantified results does the report present for the Jean Zay H100 partition and the Lucie 7B training campaign?\",\"answer\":\"It reports an annual footprint of 417.5 t CO2 eq for the Jean Zay H100 partition, an effective intensity of 36.7 g CO2 eq per H100 GPU-hour, and a total training footprint of 21 t CO2 eq for Lucie 7B, including amortised hardware manufacturing.\"}]",1784198527,30,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"life-cycle-assessment-of-pre-training-the-lucie-7b-open-source-large-language-model-on-the-jean-zay-supercomputer","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/life-cycle-assessment-of-pre-training-the-lucie-7b-open-source-large-language-model-on-the-jean-zay-supercomputer/84821/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What life cycle stages does the study include for Lucie 7B pre-training?","Question",{"text":74,"@type":75},"The scope covers data preparation through model validation, and integrates the full hardware life cycle: manufacturing, use (compute, storage, administration, cooling), and end-of-life.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How are greenhouse gases accounted for in the assessment?",{"text":79,"@type":75},"The study follows AFNOR SPEC 2314 “Frugal AI” as the framing reference and applies the Labos 1point5 methodology for GHG accounting in computing.",{"name":81,"@type":72,"acceptedAnswer":82},"What quantified results does the report present for the Jean Zay H100 partition and the Lucie 7B training campaign?",{"text":83,"@type":75},"It reports an annual footprint of 417.5 t CO2 eq for the Jean Zay H100 partition, an effective intensity of 36.7 g CO2 eq per H100 GPU-hour, and a total training footprint of 21 t CO2 eq for Lucie 7B, including amortised hardware manufacturing.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,121,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":28,"slug":120},"research-report",{"id":122,"doc_module":4,"doc_module_name":45,"category_name":123,"show_sort_weight":124,"slug":125},9,"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]