Dr. Dennis Morrison is the Chief Clinical Officer at Eleos and the owner of Morrison Consulting. Previously, he has served as Chief Clinical Officer at Netsmart Technologies, CEO of Centerstone Research Institute (CRI), and CEO of Center for Behavioral Health (CBH).
The artificial intelligence (AI) revolution is reshaping virtually all aspects of our world. Yet as adoption grows in both our personal and professional lives, we are confronted with an uncomfortable truth:
Every time we use an AI application, it takes a significant amount of power.
Whether it’s accessing route guidance via GPS, receiving a purchase suggestion from Amazon, entering a ChatGPT query, using augmented intelligence to accelerate clinical documentation, or relying on AI to help interpret MRI results, every AI algorithm uses power—and that power comes from massive data centers with high environmental and societal costs.
That said, the overwhelming benefits of AI cannot be overlooked. So, the question becomes, “Which use cases are worth the associated costs—and how can we start to offset those costs when warranted?”
I don’t have all the answers—and I’m not the “AI referee”—but I do think it’s a topic that deserves more dialogue. My intent with this piece is to spark that conversation by laying out:
- The data,
- The many sides of this evolving argument, and
- Where I think we should go from here.
Unpacking the Consequences of AI
Healthcare professionals are well-versed in ethics, so their collective concern over the ethical ramifications of AI use isn’t surprising—especially with regard to the environmental impact of data centers. And that concern is well-founded.
The Rapid Expansion of Data Centers
Data center infrastructure has grown significantly in recent years. According to Business Insider, there were approximately 311 data centers worldwide in 2010. By 2024, that number had exploded to 1,240.
Although we have historically used data centers for web-hosted apps, streaming services, and other online tools, this fourfold increase is mostly attributable to an increased demand for AI computing.
Growing Energy Demand
A single large data center can consume as much power as a small city, with federal estimates predicting the demand for data center power to triple over the next three years.
To put this into perspective, US data centers are projected to soon consume more electricity than the entire nation of Poland (population 36.6 million).
This energy demand has implications for other sectors as well. For example, it is forcing utility companies to abandon their renewable energy goals and revert back to reliance on fossil fuels.
Water Use in a Warming World
Data centers also require large quantities of water for their cooling systems—often several million gallons per day, per facility. Unfortunately, approximately 40% of these facilities are located in areas already experiencing high water scarcity.
As climate change intensifies drought conditions in many regions, this competition for water resources creates direct conflicts with agricultural and residential needs.
Public Health and Air Quality Costs
The backup generators that ensure uninterrupted power also contribute to air pollution.
Research estimates these facilities create air pollution-related public health costs to the tune of $5.7 billion–$9.2 billion annually. Another analysis suggests these costs could reach $20 billion over a six-year period when accounting for the full range of health impacts—from respiratory diseases to cardiovascular complications.
Community Burdens
While data centers often receive substantial tax breaks—sometimes amounting to millions of dollars—they create relatively few full-time jobs.
Meanwhile, local residents find themselves living next to industrial complexes that operate around the clock—creating unwanted noise, traffic, and infrastructure impacts. Many nearby residents report a constant low-grade hum coming from these facilities. It’s less “cloud computing” and more “industrial neighbor.”
Comparing Environmental Footprints
Clearly, there’s no hiding from the fact that data centers bring about a host of negative side effects. But in order to understand whether data centers represent a significant ethical concern, it might be helpful to compare the negative impact of these facilities to that associated with other human endeavors.
With that in mind, I’ll pose this question: How does the environmental impact of AI data centers compare to the environmental impact of other industries—particularly healthcare itself?
The Rise of Disposable Healthcare
Some of us are old enough to remember when medical devices were sterilized and reused multiple times. Syringes and IV bottles were glass, not plastic. Needles were resharpened and reused. The first disposable syringes weren’t invented until about 1955, only gaining widespread adoption in the mid-1960s.
If you feel uneasy about the ethics of a prepackaged syringe, you could always ask your nurse for a gently used, freshly resharpened one. (I’m kidding. Mostly. 😉)
The same pattern was repeated for other reusable products in healthcare, creating a “disposable revolution.” It is difficult to calculate exactly how much medical waste was generated in the 1950s, because quantities weren’t recorded; the waste was simply discarded like any other refuse.
It wasn’t until the 1980s that medical waste was tracked as an environmental concern.
The Scale and Risk of Modern Medical Waste
Today, US healthcare facilities generate approximately 5.9 to 6 million tons of waste annually, 15% of which is classified as hazardous. This hazardous waste includes infectious materials, pharmaceutical byproducts, radioactive material, and sharps that require specialized handling and disposal.
The incineration of medical waste—particularly when done incorrectly—releases harmful pollutants, including human carcinogens like dioxins and furans, mercury, lead, and particulate matter. These pollutants can cause adverse health effects, including respiratory diseases and cancer.
Medical Waste vs. Data Center Impact
Still, the environmental impact of medical waste versus that associated with data centers is not a perfect comparison. That’s because:
- Data centers support both essential and frivolous functions, but virtually all medical waste results from direct patient care. Every bandage, syringe, and contaminated material represents treatment delivered to an individual patient. Ethical concerns about medical waste are more easily assuaged because the human benefit is more obvious.
- Medical waste is distributed across thousands of facilities serving local communities. The environmental impact of data centers, on the other hand, is concentrated in specific locations—often in communities that receive little direct benefit from the facility’s presence.
- Medical waste is an established problem with known impacts and existing regulatory frameworks. Data center growth represents a rapidly accelerating challenge with impacts that are still scaling upward (and not entirely understood).
While healthcare waste is substantial, its volume has remained relatively stable over recent decades, especially with improved disposal practices. By contrast, data center energy consumption is on an exponential growth curve.
In fact, consumer electricity bills in some areas are projected to double by 2039 due to data center expansion.
This creates a situation where everyday citizens bear an increasing financial burden for infrastructure that primarily serves corporate interests and, increasingly, AI applications of varying social value.
Considering the Ethical Arguments
These environmental realities raise deeper ethical questions about responsibility, benefit, and harm.
The Case Against AI’s Environmental Cost
Environmentally centered arguments against the use of AI include:
- The exponential growth in energy demand coincides with urgent climate crisis warnings. As utilities abandon renewable energy efforts and revert to relying mainly on fossil fuels, data centers may actively undermine broader societal efforts to address climate change.
- Technology companies are not consistently transparent about their true environmental footprint. This makes it difficult for consumers and policymakers to make informed decisions. The costs—including higher electricity bills, health impacts from air pollution, and water scarcity—are being externalized to communities while profits remain privatized.
- The environmental damage caused by data centers is more immediate and measurable. The long-term benefits of AI applications, while apparently attractive now, remain largely speculative—particularly for consumer-facing applications like entertainment chatbots. This creates an ethical imbalance where certain harm is traded for uncertain gain.
- Using AI systems makes individuals complicit in environmental degradation. In an era of climate crisis, this argument holds particular weight when considering luxury technologies versus essential services.
- Marginalized communities often bear a disproportionate burden of data center impacts because of their proximity to data facilities—but they arguably benefit the least from the technologies they support.
The Compelling Case for AI’s Benefits
Despite the drawbacks cited above, defenders of AI cite important counterarguments:
- Artificial intelligence enables pro-environment activities requiring complex calculations, such as climate modeling, optimizing renewable energy distribution, and improving efficiency across industries. In fact, AI applications in transportation, agriculture, and manufacturing could create environmental impact reductions that offset—or even outweigh—data center costs.
- Not all AI applications carry equal moral weight. Using AI for medical diagnostics, disease research, or disaster prediction may justify environmental costs in ways that entertainment chatbots do not.
- Individual users of AI tools represent a minuscule fraction of total data center consumption compared to the training of large models by corporations. The environmental impact of personal AI use may be less than that tied to many routine activities, like driving or air travel.
- Some data centers can and are transitioning to renewable energy. Improved cooling technologies and efficiency gains in computer chip design continue to advance. Companies like Microsoft have announced comprehensive plans to reduce community impacts.
- Many essential technologies carry environmental costs—healthcare, telecommunications infrastructure, and transportation systems, just to name a few. Singling out AI may be arbitrary when these other systems also consume substantial resources.
- AI is already deeply embedded in infrastructure for healthcare, finance, and security systems. Abrupt cessation could cause significant societal disruption.
Pursuing Innovation Responsibly
The tension between data center impacts and AI benefits reflects a broader challenge that has emerged as society has grown more technologically dependent:
How do we pursue innovation while being environmentally and socially responsible?
There are some positive signs of progress on this front. As noted above, Microsoft recently announced a five-point plan in which the company commits to:
- Ask utility companies to set higher rates to cover electricity costs.
- Minimize water use and replenish more water than it withdraws.
- Create data center jobs for area residents.
- Pay its “full share” of local property taxes for data centers.
- Invest in AI training for locals.
This hopefully becomes a blueprint for other large corporations to continue offsetting the environmental and ethical costs of AI. But when it comes to healthcare, costs and benefits aren’t as clear.
The Unique Healthcare Conundrum
The delicate ethical balance between resource use and patient benefit is nothing new for healthcare professionals.
We accept that MRI machines consume significant electricity, surgical procedures generate substantial waste, and pharmaceutical manufacturing negatively impacts the environment.
We do so because the benefit to patients justifies these costs.
Similarly, the question for healthcare AI becomes, “Does this application provide sufficient benefit to justify its environmental cost?”
For instance:
- A neural network that improves breast cancer detection rates may clearly meet this threshold.
- An ambient listening tool that generates a better, more compliant note while decreasing clinician burnout also seems worth the tradeoff.
- An AI chatbot that provides companionship but potentially displaces human connection occupies murkier ethical territory.
Furthermore, most medical applications of AI—from analyzing imaging to identifying drug interactions to using ambient listening to improve clinical documentation—use a tiny fraction of the computing power required to train massive language models.
Shaping AI With Intention
Regardless of where you stand on the ethics of data center impacts, the overarching reality is clear: Artificial intelligence is not going away.
The question is not whether AI will continue to develop, but how we will shape its development to minimize harm while maximizing benefit.
For individuals concerned about data center impacts, personal choices to limit AI use will be largely symbolic. The vast majority of environmental impact comes from corporate model training, not consumer use.
Of course, one might also choose the “AI Amish” route and reject all forms of AI outright. If that’s their conviction, I respect it. But it’s harder than it sounds in a world where AI is embedded in everyday tools—like GPS, Facebook, Amazon, and other non-healthcare, AI-intensive applications. They should also recognize that doing so will have minimal impact on the AI industry at large.
And of course, there is the old saying: “AI won’t take your job—someone using AI will.”
If you really want to have an impact on this problem, more effective approaches include:
- Supporting policies that mandate environmental standards for data centers;
- Advocating for corporate transparency about environmental impacts;
- Choosing to use AI selectively for applications with clear benefits;
- Demanding that technology companies invest in renewable energy and efficiency; and
- Engaging with elected representatives on data center regulation.
Like most ethical quandaries, this one cannot be reduced to simple binaries. Context, usage, scale, and a myriad of other factors matter.
The immediate, measurable harm of environmental degradation must be weighed against both the speculative future benefits of AI and the potential for technology to evolve toward sustainability.
The rapid expansion of AI capabilities and corresponding data center infrastructure is here today, and it deserves the attention of all healthcare professionals.
And because this topic can get heavy fast, I’ll leave you with this quote from Woody Allen that I think instills hope in all of us:
“More than any other time in history, mankind faces a crossroads.
One path leads to despair and utter hopelessness.
The other, to total extinction.
Let us pray we have the wisdom to choose correctly.”
Want to learn how Eleos designs healthcare AI tools with ethics in mind? Request a demo of our purpose-built AI platform.
