<h4>Objective</h4>Artificial intelligence (AI) is increasingly used in radiology, but its environmental implications have not been sufficiently studied, so far. This study aims to synthesize existing literature on the environmental sustainability of AI in radiology and highlights strategies proposed to mitigate its impact.<h4>Methods</h4>A scoping review was conducted following the Joanna Briggs Institute methodology. Searches across MEDLINE, Embase, CINAHL, and Web of Science focused on English and French publications from 2014 to 2024, targeting AI, environmental sustainability, and medical imaging. Eligible studies addressed environmental sustainability of AI in medical imaging. Conference abstracts, non-radiological or non-human studies, and unavailable full texts were excluded. Two independent reviewers assessed titles, abstracts, and full texts, while four reviewers conducted data extraction and analysis.<h4>Results</h4>The search identified 3,723 results, of which 13 met inclusion criteria: nine research articles and four reviews. Four themes emerged: energy consumption (n = 10), carbon footprint (n = 6), computational resources (n = 9), and water consumption (n = 2). Reported metrics included CO2-equivalent emissions, training time, power use effectiveness, equivalent distance travelled by car, energy demands, and water consumption. Strategies to enhance sustainability included lightweight model architectures, quantization and pruning, efficient optimizers, and early stopping. Broader recommendations encompassed integrating carbon and energy metrics into AI evaluation, transitioning to cloud computing, and developing an eco-label for radiology AI systems.<h4>Conclusions</h4>Research on sustainable AI in radiology remains scarce but is rapidly growing. This review highlights key metrics and strategies to guide future research and practice toward more transparent, consistent, and environmentally responsible AI development in radiology.<h4>Abbreviations</h4>AI, Artificial intelligence; CNN, Convolutional neural networks; CT, Computed tomography; CPU, Central Processing Unit; DL, Deep learning; FLOP, Floating-point operation; GHG, Greenhouses gas; GPU, Graphics Processing Unit; LCA, Life Cycle Assessment; LLM, Large Language Model; MeSH, Medical Subject Headings; ML, Machine learning; MRI, Magnetic resonance imaging; NLP, Natural language processing; PUE, Power Usage Effectiveness; TPU, Tensor Processing Unit; USA, United States of America; ViT, Vision Transformer; WUE, Water Usage Effectiveness.

Data centres support artificial intelligence (AI) development but place rapidly increasing demands on electricity and freshwater resources, with cooling representing a significant portion of their total energy consumption. Wastewater treatment plants (WWTPs) discharge large volumes of treated effluent with substantial cooling potential; however, their integration with data centre infrastructure has not been evaluated. Here we construct a global geodatabase of over 4775 data centres and 57,547 municipal WWTPs across 98 countries, integrating spatial analysis, engineering systems modelling, optimisation, and life-cycle assessment to quantify the benefits of combining treated water reuse with bidirectional thermal recovery. The analysis reveals a strong global spatial co-occurrence between data centres and WWTPs, enabling optimized national-scale pairings in which treated effluent is used for data centre cooling and the return heat is recovered to support sludge drying and anaerobic digestion. This symbiotic approach reduces greenhouse gas emissions by approximately 84 million tonnes of CO₂ equivalent annually, conserves approximately 1300 million m³ of freshwater, and provides net annual cost savings of approximately US$95.4 billion. The greatest mitigation and water-saving potential lies in the United States, Japan, China, the Netherlands, and the United Kingdom. These findings establish data-water symbiosis as a readily scalable infrastructure solution that decouples AI from its carbon and water footprints. WWTPs are poised to evolve from disposal facilities into critical energy-coupling hubs, enabling efficient thermal and water exchange across urban systems and accelerating progress towards multiple Sustainable Development Goals.

It is also remarkable that, while the conventional heating, ventilation and air conditioning systems (HVAC) for similar size rooms deal with heat fluxes in the order of 40-90 ,[10] cooling systems in data centers work with extremely higher heat loads: modern data center infrastructures can need to refresh apparatus that produces 6-10 as power density.[11] In the next few years, it is estimated that the power density consumption will still raise up to 15 .[12] Google for instance mentions the significance of taking serious action to reinforce energy efficiency in case of either running a small data center or a huge service. According to Google, many manufactures construct the suitable equipment to be used at temperatures higher than the standard 21°C we aforementioned. This may lead to next generation servers having the ability of performing at higher temperatures, fact that can make data center using less equipment and therefore save money.