Replication Data for: The Infrastructure Paradox of AI Development in LMICs

Purpose of dataset: This dataset underpins the findings reported in this study and is deposited to enable independent verification, replication, and secondary analysis of AI infrastructure constraints across LMICs. Nature of dataset: The dataset comprises four components: (i) raw survey responses as exported from Google Forms; (ii) the cleaned and recoded dataset used for all reported analyses; (iii) the complete survey questionnaire including the informed consent disclosure; and (iv) a codebook documenting all variable definitions, response categories, and recoding decisions. All data are fully anonymous with no personally identifiable information collected at any stage. Scope of dataset: The data capture responses from 91 AI practitioners across 27 countries and four geopolitical regions, covering self-reported assessments of electricity, internet, hardware, and data infrastructure alongside indicators of national AI policy, intervention activity, innovation strength, and investment sources. The data correspond to all quantitative findings reported in the manuscript.