AI's hidden carbon and water footprint (vu.nl)

A recent study published in the scientific journal Patterns reveals the substantial carbon and water footprints associated with AI systems, particularly as the demand for AI applications surges. Data scientist Alex de Vries-Gao estimates that by 2025, AI systems could generate between 32.6 and 79.7 million tonnes of CO₂ emissions annually—comparable to the annual emissions of a major city like New York. This projection stems from an analysis combining sustainability reports from leading tech companies and global energy statistics, highlighting how the growing electricity demand for AI could translate into significant environmental impacts. Additionally, the study uncovers that AI systems' water consumption could reach between 312.5 and 764.6 billion liters annually, which is three to four times higher than current estimates for data center water use. As companies often underreport their water consumption linked to electricity generation, De Vries-Gao emphasizes the urgent need for greater transparency in energy and water use reporting. He advocates for stricter regulations that differentiate AI-related environmental impacts from other applications, asserting that understanding these costs is essential for ensuring that AI contributes positively to sustainability.