The Environmental Impact of AI in Medicine
Written by Dr. Megan Hill, Climate Action Committee
As many of us begin to incorporate large language model (LLM) technology through medical Artificial Intelligence (AI) scribes into our practices, it is important to balance their use against environmental costs. There are significant benefits to medical practice from these programs, including greater efficiency, improvements in record quality and communication. This should be weighted against the impact of AI on the expanding environmental footprint of our health care system, which is already significant.
AI tools are generated from brick and mortar data centers which require significant electricity to operate. AI software consumes significantly more energy than more basic programs such as a web search, both in its training and ongoing use.¹ New AI model development and learning requires even higher levels of electricity, and with the drive to continue to innovate, additional energy used to train prior versions goes to waste, while overall energy expenditure remains high, rather than becoming more efficient once a model has been developed.² With the growth of AI, these data centers are expanding exponentially, drawing on larger proportions of their regions’ electricity.¹ For example, data centers in Ireland currently consume more than 20% of the entire national electrical supply.¹ Over 10% of AI data centers are located in regions of the US heavily reliant on coal and natural gas, as opposed to more sustainable energy sources,¹ and most regions have insufficient infrastructure to support these energy needs though sustainable sources alone.²
Further costs for AI data centers include significant water needs to cool hardware, further exacerbating global water stress;¹,² materials for hardware;¹,² “dirty” mining processes;¹ transportation emissions;² and e-waste.¹
Osmanlliu et al encourage physicians to have agency in their AI use, encouraging prioritization of AI use for tasks where it can provide real value, in some cases at lower cost than with human operators.¹ Their Table 1 outlines a framework of environmental costs versus clinical value to consider with various type of AI programming.
Further steps we as physicians can take include raising the importance of these concerns with both legislators and service providers. Inquiring with AI scribe providers about their environmental policies, and expressing that it is a priority to me in choosing a provider can be a simple yet impactful step. Discussing the judicious use of AI with family, friends and community outside of the medical system can also encourage reduced use of these programs outside of their utility as a value-adding tool.
Megan Hill MD CCFP(AM) FRCP
Osmanlliu et al. The Urgency of Environmentally Sustainable and Socially Just Deployment of Artificial Intelligence in Health Care July 9, 2025; NEJM Catal Innov Care Deliv 2025;6(8); DOI: 10.1056/
Zewe, A. Rapid development and deployment of powerful generative AI models comes with environmental consequences, including increased electricity demand and water consumption. MIT News; January 17, 2025