Healthcare organizations are rapidly adopting artificial intelligence to improve care delivery, streamline operations, and manage patient populations, but the technology carries a growing environmental cost that the sector is only beginning to measure. Seema Wadhwa, founder and CEO of the nonprofit Health in Climate and former executive director of environmental stewardship at Kaiser Permanente, says AI “has an important role to play in care delivery, care management and healthcare operations,” while stressing that its energy demands must be weighed against the sector’s climate commitments.
Background: Healthcare’s Climate Footprint
The World Health Organization has described climate change as the single greatest threat to human health. Health systems are themselves significant contributors to that threat, accounting for a substantial share of emissions in many countries through energy use, supply chains, pharmaceuticals, and facilities.
Many large providers have set emissions reduction targets in recent years, often aiming for net-zero operations within the coming decades. AI adoption now introduces a new variable into those plans, because the computing infrastructure behind machine learning models draws significant electricity.
The Energy Cost of Computation
Training and running large AI models requires dense clusters of graphics processors housed in data centers that operate around the clock. These facilities consume electricity for computation and for cooling, and their demand has grown quickly as models become larger and more widely deployed.
Industry analyses from energy research organizations have projected that data center electricity consumption could rise sharply over the rest of the decade as AI workloads expand. Healthcare applications, from diagnostic imaging analysis to predictive staffing tools, contribute to that load, even if individually they represent a small fraction of the total.
Weighing Benefits Against Impact
Proponents argue that AI can reduce emissions elsewhere in healthcare. Diagnostic tools can cut unnecessary imaging and testing, predictive models can optimize energy use in hospital buildings, and remote monitoring can reduce travel for patients and clinicians. Measuring those net effects, however, requires consistent data that many organizations do not yet collect.
Wadhwa’s framing points to this tension: the question is not whether AI should be used in healthcare, but how its deployment can be structured so that operational gains do not come at the expense of climate goals. Procurement decisions, choice of cloud providers, and the efficiency of models themselves all influence the outcome.
Some health systems are beginning to request emissions disclosures from technology vendors and to ask whether AI tools run on renewable power. Such requests reflect a broader shift toward treating digital infrastructure as part of a health organization’s environmental footprint rather than as a separate category.
Implications for the Industry
For hospital executives and procurement teams, the emerging challenge is to incorporate AI energy use into sustainability reporting and vendor evaluation. Regulators and accreditation bodies may eventually expect the same level of environmental accounting for digital systems that they now expect for physical operations.
For clinicians and patients, the trade-offs are less visible but still relevant. Decisions about which AI tools to adopt will shape both the quality of care and the sector’s contribution to emissions, making transparency about these costs an increasingly important part of responsible innovation.
What to Watch Next
Observers will be watching whether health systems publish standardized measurements of AI-related energy consumption and whether vendors begin disclosing efficiency metrics for their models. Policy developments on data center emissions and clean-energy procurement could also shape the economics of AI deployment in healthcare. How quickly these measurement frameworks mature will likely determine whether the sector can scale AI while meeting its climate commitments.
Frequently Asked Questions
Does using AI in healthcare actually increase a hospital's carbon footprint?
It can, because machine learning models run on data center servers that consume electricity for computation and cooling. However, the net effect depends on whether AI reduces emissions elsewhere, such as by cutting unnecessary imaging, optimizing building energy use, or reducing patient and clinician travel. Measuring that balance requires consistent data that many organizations do not yet collect.
Why is AI energy use considered part of a health organization's environmental footprint?
Digital infrastructure is increasingly seen as part of operations rather than a separate category. Cloud computing and AI workloads draw real electricity, so their emissions count toward a provider's climate commitments, including net-zero targets. Some health systems now treat software and data center energy use with the same accounting rigor as physical facilities and supply chains.
How can hospitals reduce the climate impact of the AI tools they adopt?
Procurement decisions matter most. Hospitals can ask vendors for emissions disclosures, check whether tools run on renewable power, and choose cloud providers with clean-energy commitments. The efficiency of the models themselves also influences energy use, so selecting leaner models suited to the task can reduce consumption without sacrificing the clinical or operational benefit.
Are AI data centers really expected to grow their electricity demand significantly?
Industry analyses from energy research organizations have projected that data center electricity consumption could rise sharply over the rest of the decade as AI workloads expand. Healthcare contributes to that load through applications such as diagnostic imaging analysis and predictive staffing, though individually these represent a small share of total demand. Growth depends on how quickly models scale and spread.
Will regulators require healthcare providers to report AI-related emissions?
This is not yet mandated, but the article notes that regulators and accreditation bodies may eventually expect environmental accounting for digital systems similar to what is already expected for physical operations. Policy developments on data center emissions and clean-energy procurement could also shape how AI deployment is economically structured in the sector.