Sign in Subscribe
Health care industry

Balancing the Promise of Health AI with Its Carbon Costs

Strategic Insights

— 1 min read

As the healthcare industry increasingly relies on artificial intelligence (AI) to respond to patient queries and alleviate the stress on overtaxed workforces, a new Cornell University study highlights the urgent need to consider the environmental footprint of AI in health care settings. The researchers propose a framework, Sustainably Advancing Health AI (SAHAI), to optimize AI-related energy consumption and emissions in health care settings. The study suggests that a significant portion of the carbon emissions generated by AI-powered tools can be mitigated through careful planning and design.

Key Takeaways:

  • The healthcare industry is expected to grow to a $187 billion industry in the next five years, with a significant portion of this growth attributed to the increasing use of AI in automated patient responses.
  • A Cornell University study has developed a framework, Sustainably Advancing Health AI (SAHAI), to optimize AI-related energy consumption and emissions in health care settings.
  • The study estimates that a year of running an AI-powered messaging tool would produce around 48,000 kilograms of carbon dioxide (CO2), equivalent to 2,300 "tree-years."
  • Providers must consider multiple factors, including energy usage, water consumption for cooling, and model accuracy, when deciding how and when to deploy AI.
  • It is more cost-effective and efficient to consider sustainability when designing AI systems rather than retrofitting them after the fact.
  • Hospitals can prioritize data centers that operate on renewable energy to reduce the operational emissions of AI workloads.

Statistics:

  • The healthcare industry is expected to grow to a $187 billion industry in the next five years.
  • A year of running an AI-powered messaging tool would produce around 48,000 kilograms of carbon dioxide (CO2), equivalent to 2,300 "tree-years."
  • 3,000 physicians answered 50 messages per day in a hypothetical AI-generated messaging application analysis.
  • A lightweight generative pretrained transformer (GPT) model consumes less computing power than a larger model.
  • 21 kg of CO2 per tree per year is a commonly cited estimate for the amount of CO2 a tree pulls out of the atmosphere in a year.

Sources:

  • "Sustainably Advancing Health AI: A Decision Framework to Mitigate the Energy, Emissions, and Cost of AI Implementation" by Chethan Sarabu, Udit Gupta, Anu Ramachandran, Shomit Ghose, and Vivian Lee, published in NEJM Catalyst on September 12, 2025.

Read more