Green AI in Healthcare Could Cut Environmental Costs

Estimated reading time: 5 minutes · Last updated: 11 October 2026

A paper published on 11 October 2026 in the journal Artificial Intelligence in Medicine, authored by researchers at the Norwegian University of Science and Technology (NTNU), argues that healthcare must adopt "Green AI" to avoid large environmental costs. Alok Mishra of NTNU says medical AI consumes substantial energy and water via large data centres, while the International Energy Agency (IEA) warns that data-centre power demand could double by 2030. The research reviewed 47 healthcare AI studies that considered sustainability and found those efforts fragmented; the authors call for lifecycle assessments and greater transparency so the clinical benefits of AI do not transfer hidden environmental costs to public health.

AI models require lots of energy and huge data centres. These data centres consume large amounts of electricity and use significant quantities of fresh water.

Alok Mishra, Professor in Data Management and Software Engineering at NTNU

Why environmental cost matters for medical AI

Healthcare has been an early and enthusiastic adopter of advanced AI workflows, from image interpretation to treatment planning. The NTNU authors argue that those clinical gains come with hidden resource use: models need training, servers and storage, and the infrastructure that supports them consumes energy and water.

The International Energy Agency projects that global data-centre power demand could double by 2030, a scale of growth the NTNU paper and the IEA say will raise emissions unless the sector changes how it measures impact. Mishra links these infrastructure trends to health outcomes, noting recent climate-driven disasters that increase pressure on health systems and make the environmental footprint of medical technology a public-health concern.

Where healthcare AI is used — and the trade-offs clinicians face

Clinical teams deploy AI across diagnostics, therapy planning, drug research and administrative work. Åsmund Flobak, oncologist at St. Olav’s Hospital and professor at NTNU, describes uses that include automated image analysis for lab-grown tumour assays, AI-assisted contouring in radiotherapy and auto-generated consultation notes; these applications speed decisions and can improve patient care.

Flobak told the NTNU authors that environmental impact rarely enters day-to-day clinical choice because clinicians prioritise direct patient benefit within existing guidelines. That practical tension—clear clinical advantage versus diffuse environmental harm—is a central reason Mishra and colleagues call for system-level rules rather than leaving sustainability to individual departments.

What 'Green AI' would require in healthcare

Mishra’s Green AI concept asks development teams and providers to assess the full lifecycle of AI systems: manufacturing, model training, deployment, operation and end-of-life disposal or recycling. The authors compare this to household energy ratings or airline carbon information: a standardised metric would allow purchasers and regulators to compare tools by environmental footprint, not just clinical performance.

The paper finds existing sustainability efforts in healthcare AI are often isolated pilots or vendor claims rather than comprehensive, comparable assessments. To change that, the authors recommend transparent reporting of energy and water use, independent lifecycle studies and regulators or funders requiring environmental accounting alongside accuracy and bias metrics.

How AI can both worsen and reduce energy use

The NTNU authors and IEA are explicit that AI is a two-way street: large models and sprawling storage raise demand, but AI can optimise energy systems and reduce waste elsewhere. Fatih Birol, Executive Director of the IEA, framed this in the agency’s 2025 AI & Energy report under the phrase “Energy for AI, and AI for Energy,” noting practical opportunities to save transmission capacity.

The IEA quantified one such upside at up to 175 gigawatts of transmission capacity saved through smarter grid operation—an amount the agency likened to powering Oslo for a year. The paper stresses that realising those savings in health systems requires deliberate design choices and policies that reward net lifecycle benefits, not only computational performance.

Outlook: two paths for healthcare AI

The bottom line

NTNU’s paper places environmental impact alongside privacy, bias and transparency as a core consideration when healthcare adopts AI. The review of 47 studies shows current sustainability work is uneven and often isolated; the authors recommend standard lifecycle reporting, independent verification and regulatory or funder requirements. At the same time, the IEA’s findings underscore that AI can deliver system-level energy savings if designed and governed with emissions in mind. For clinicians and hospital procurement teams, the immediate task is to demand comparable environmental data so clinical benefits do not come at the expense of public health and resilience.

What to watch

• Watch for follow-up publications or guidance from NTNU or further papers in Artificial Intelligence in Medicine expanding on lifecycle metrics; no date has been set.
• Watch for policy-level responses from health ministries or regional health authorities proposing standard reporting for energy and resource use in clinical AI; no date has been set.
• Watch for IEA updates or technical guidance linking AI-driven grid optimisation to healthcare facility resilience; no date has been set.