The Banking Sector's Own AI Footprint: The energy and carbon impacts of artificial intelligence applications in banking operations

13456-os tétel angol nyelvű utánközlése.

Abstract

Over the past decade, the banking sector has become a key actor in the sustainability transition through the measurement of financed emissions (Scope 3) and the decarbonisation of portfolios. At the same time, however, artificial intelligence (AI) solutions have been proliferating rapidly in the operations of credit institutions – from customer-service chatbots and credit-risk models to agentic systems that automate multiple steps of business workflows. The energy and resource requirements of these applications constitute a dynamically growing component of the sector’s own environmental footprint that has so far received little attention. This study has three objectives. First, it reviews the international literature and measurement methodologies concerning the lifecycle environmental impacts of AI systems, including training, inference, embodied emissions, and water and resource consumption. Second, it examines how current trends in the adoption of AI in banking – particularly the growing use of generative and agentic AI – affect credit institutions’ Scope 2 and Scope 3 emissions categories and, through an exploratory document analysis of the 2025 disclosures of nine credit institutions, provides empirical evidence that AI-related energy consumption cannot currently be identified in prevailing reporting practices. Third, it proposes a methodological framework for quantifying the “banking AI footprint” and integrating it into sustainability reporting (CSRD/ESRS). The study argues that, alongside the narrative of “greening by AI”, the sector must also institutionalise a “greening of AI” perspective: without energy-related criteria being incorporated into AI procurement decisions, model sizing, and cloud-service agreements, digital transformation may undermine the credibility of banks’ own climate targets.