AI Compute, Conflict Events, and Training Data Geography, 2023–2026

Country-level data compares three geographies of artificial intelligence from 2023 to 2026. It aggregates AI compute infrastructure location, organization headquarters for AI training datasets, and armed conflict event counts. The dataset was compiled by Nathan Coyle and deposited on Harvard Dataverse in July 2026.

Use Cases

• Analyze the geographic disparity in AI compute capacity based on Global North/South classification
• Correlate civilian-targeting conflict events with the absence of local AI infrastructure
• Map the headquarters of organizations compiling AI training datasets to assess data provenance bias

Strengths

• Integrates three independently maintained sources: Epoch AI, ACLED, and Data Provenance Initiative
• Includes a formatted Excel workbook, interactive HTML dashboard, methodology report, and summary infographics
• Provides specific findings: 92.9% of AI compute capacity is in the Global North

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Data may reflect geographic and temporal bias inherent to the source aggregation methods

Provenance

Source

Combines Epoch AI's compute tracker, ACLED conflict data, and Data Provenance Initiative audit

Collection Method

Aggregated from three sources using a hybrid Global North/South classification convention

Time Range

2023–2026

Freshness

Last updated 2026-07-15 04:29:23; freshness should be verified

Geography

Country-level global coverage