Green Foreign Direct Investment is flowing far beyond renewables: a taxonomy-guided LLM analysis

Foreign direct investment (FDI) finances and diffuses the capital, skills and know-how needed for the low-carbon transition, yet ‘green FDI’ remains systematically under-measured. Sector proxies – typically renewable energy and waste remediation – miss green activities embedded within other industries, yielding a partial and geographically-biased picture. Here we combine Large Language Models with the EU Taxonomy for Sustainable Activities to classify investment projects against sector-specific green criteria. Applying this taxonomy-guided framework to 109,084 inward greenfield FDI projects into the European Union and the United Kingdom (2013– 2024), we identify 15.7% of FDI value as green – around twice the share captured by traditional measures. We show that sector metrics miss large volumes in manufacturing and services, and that beyond-energy green FDI is more strongly linked to extra-European investors, implying distinct geopolitical dependencies. Blinded human coding and robustness tests confirm high accuracy and reproducibility. Together, these results enable scalable monitoring of investment alignment with climate objectives.