Data

AI compute and data intensity indicators
Annual firm-level measures of how intensively firms use AI compute and data, covering 12,038 US listed firms from 2000 to 2023. The paper uses them to test whether compute, access to data, or the interaction between the two drives firm dominance and industry dynamism, using two shocks: the 2006 launch of Amazon Web Services and the 2017 introduction of transformer-based models.
PAPER  Is It AI or Data That Drives Firm Market Power? with Orlando Gomes and Roxana Mihet, Journal of Monetary Economics, 2026
DATASET (HARVARD DATAVERSE) ↗

AI intensity indicators and replication package
A firm-level measure of AI intensity for US public firms, built from patents and 10-K business descriptions. The paper uses it to show that cyberrisk suppresses innovation at non-AI firms, while frontier AI firms with internally developed capabilities are protected. The full replication package is released with the indicators.
PAPER  Cyberrisk and AI Firms with Roxana Mihet and Julian Jang-Jaccard, Review of Corporate Finance Studies, 2026
DATASET (XLSX) ↗   REPLICATION PACKAGE ↗

AI firms: country-level data and investment relationships
Two releases on AI production. The country-level data map AI-producing firms into the five layers of the AI supply chain (compute, cloud and related infrastructure, data tools, AI models and AI applications) and show where AI production is located. The chart data cover investment deals among AI firms between 2021 and 2025, including the share of deals where the investor and the target also have a supplier–customer relationship.
PAPER  The Geography of AI Firms with Vatsala Shreeti, BIS Working Papers No 1343
PAPER  Circular Relationships among AI Firms with Jon Frost, Rudraksh Kansal, Vatsala Shreeti and Leanne Si Ying Zhang, BIS Bulletin No 137, 2026
COUNTRY DATA (XLSX) ↗   CHART DATA (XLSX) ↗