You write the inhaler refill without thinking, the way you have a hundred times, and only later do you notice the thread running from that prescription to the tool that wrote the note.
A child is wheezing in front of me. He is four, perhaps five, and his mother says the nights are the worst, that he wakes up fighting for air. No puede respirar bien en las noches. I listen to his chest, step up his controller, show her the spacer again, and book the follow-up. Ordinary asthma in an ordinary clinic. We are good at this part.
What I had not traced, until recently, is where the rest of the story lives.
The Colleague Who Could Not Sleep
A colleague could not sleep after reading what the data centers behind AI are doing to the communities built beside them. The water they drink. The air they breathe. The asthma in the children who live downwind. She told me she was done using AI tools, because she could not reconcile helping her patients with harming someone else’s.
I did not have an answer for her. But I could not stop thinking about her question.
Because the communities she was describing are the communities I serve.
What the Research Says
By 2030, air pollution from data centers in this country could cause up to 1,300 premature deaths a year and hundreds of thousands of asthma cases, with public-health costs approaching twenty billion dollars (Caltech and UC Riverside, “The Unpaid Toll,” 2024). The per-household burden of that pollution can run many times higher in low-income communities and communities of color (same study). Roughly two-thirds of new U.S. data centers have been built in water-stressed regions (World Resources Institute, 2026). And the facilities are sited, again and again, where land is cheap and resistance is thin, which in this country means the same neighborhoods that already carry the most.
Sources for the figures above, to cite and confirm before publishing: Caltech and UC Riverside, “The Unpaid Toll” (2024), Cornell in Nature Sustainability (2025), the IEA Energy and AI report (2025), and the World Resources Institute (2026).
The NAACP has a name for this. They call it the fight against dirty data centers, and they frame it as a new form of environmental racism.
Why This Reorganizes the Whole Conversation
We have been talking about AI’s environmental cost as a climate story, far away and slow. It is also a clinical story, close and now.
The energy and the water and the diesel exhaust do not stay abstract. They become a child’s inhaler. They become a missed shift, a hospital bill, a worse number on a pulmonary function test.
This is not the carbon footprint of a single prompt. It is the asthma downstream of an industry.
The carbon footprint of AI is not only a climate number. In my clinic, it can be a child who cannot sleep.
And the arithmetic is familiar. The benefit of the tool flows to the clinic and the company. The cost flows to a community that never got asked, often the same kind of community sitting in my waiting room.
Equity Lens
This is the part I cannot unsee.
The promise of AI in medicine is that it will help us care for the underserved. The infrastructure behind it may be harming the underserved to do it. The same families, the same zip codes, the same lungs.
That does not mean the tools are worthless, and it does not mean a clinician should quit them in guilt. It means we cannot keep treating the environmental cost as someone else’s department. When a system buys an AI tool, where its power comes from and where its pollution lands are clinical questions, because they arrive in our exam rooms as disease.
Why I Am Bringing This to the Podcast
Here is the honest truth. I am a family physician, not an environmental scientist. I can see the downstream. I cannot fully explain the upstream.
So I am bringing in two people who can.
On an upcoming episode I am sitting down with Betty Villantay, a first-year medical student whose work lives at the meeting point of technology and the environment, and Manijeh Berinji, a physician with a master’s in public health who has spent years bridging informatics and climate health. One is new to medicine. The other has been thinking about this intersection for a long time. Both care deeply about what this technology costs the planet, the body, and the communities we serve.
I wanted that range in the room on purpose. The person arriving with fresh eyes, and the person who has been carrying these questions for years.
I do not have this resolved. That is exactly why I am convening the conversation instead of pretending to conclude it.
Digital Health Pearls
The environmental cost of AI is a clinical issue. It arrives in primary care as asthma, heat illness, and energy poverty.
Data centers are sited disproportionately in low-income communities and communities of color. The burden tracks the same lines as every other environmental harm.
Ask of any AI tool not only whether it works, but what it costs to run and who lives next to that cost.
Use the smallest tool that does the job. Save the largest models for the problems that need them.
TL;DR
AI’s environmental footprint is not a distant climate abstraction. It is a clinical and equity problem that lands on the same underserved communities clinicians serve, as asthma, respiratory disease, water stress, and energy poverty. The research projects up to 1,300 premature deaths a year from data center pollution by 2030, with the heaviest burden on low-income communities of color. I do not have this resolved, which is why I am bringing two experts, Betty Villantay and Manijeh Berinji, into the conversation on the podcast.
Invitation to Compare Notes
Have you started to see the downstream of this infrastructure in your own panel, even without naming it.
And what would you most want two thoughtful experts to answer about the environmental cost of the tools we use every day. I am collecting questions before the recording.
Disclaimer: All views expressed are my own and do not represent my employer or any institution I am affiliated with. Any tools, products, or technologies mentioned are included for educational purposes only and are not sponsored or endorsed. Nothing in this piece should be interpreted as medical advice.