2026-10-06T06:32:34Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34068456 2026-10-05T09:13:09Z category_25975 item_type_3 month_year_10_2026

Data and code for “Unequal geographies of large language models across labour, compute and energy” Haoxiang Zhao (25310646) Industry economics and industrial organisation Large language models Labour exposure AI infrastructure Facility power capacity Operational carbon emissions Industrial energy Spatial inequality <p dir="ltr">This repository contains data and code supporting the study “Unequal geographies of large language models across labour, compute and energy”. The study connects potential occupational exposure to large language models with labour and energy responses and the geographic distribution of documented advanced-compute facilities.</p><p dir="ltr">The deposit includes numerical source data for four main figures and six supplementary figures, supporting tables, data-matching and sample-selection records, archived analysis code, and selected World Bank and Epoch AI data snapshots. Documentation provides variable definitions, data sources, dependencies and execution instructions.</p><p dir="ltr">Facility capacity refers to documented electrical power capacity in megawatts, rather than computational throughput. Operational carbon emissions are estimated under stated utilisation assumptions using national grid carbon intensity. Final reported intervals use 90% coverage.</p><p dir="ltr">Restricted IEA source data and original Chinese microdata are excluded. Source-access instructions are provided. The package includes tested validation and Figure 4 regeneration scripts; the complete pipeline from original inputs to regression results has not yet been independently rerun.</p> 2026-10-05T09:13:09Z Dataset Dataset 10.6084/m9.figshare.34068456.v2 https://figshare.com/articles/dataset/Data_and_code_for_Unequal_geographies_of_large_language_models_across_labour_compute_and_energy_/34068456 CC BY 4.0