The Revolving Door Project, a Prospect partner, scrutinizes the executive branch and presidential power. Follow them at therevolvingdoorproject.org.
Even amid ever-intensifying division, there are some things that bind all of us together, chief among them our impermanence—and the experiences that make us face it.
Perhaps the greatest specter of mortality that haunts the public imagination is a disease that seems to touch everyone’s life in one way or another: cancer. It’s fitting that oncologist Siddhartha Mukherjee, in his biography of the disease, calls it the “emperor of all maladies.”
According to the Centers for Disease Control, there were over 1.8 million new cases of cancer in Americans in 2022, the most recent year with incidence data. In 2023, more than 600,000 Americans died from the disease. Those numbers are pretty typical; the National Cancer Institute projects that 2025 will see about two million new cases and 618,000 deaths. Only heart disease kills more Americans.
Given that, it’s no wonder that the hunt for a cure to cancer has animated public-health research for decades, highlighted by programs like the Cancer Moonshot announced by President Biden. Hundreds of new treatments have been approved by the Food and Drug Administration in the last decade, including 207 between 2016 and 2021.
But for all the incredible public-health achievements that continue to make cancer more and more survivable, cancer prevention all too often takes a back seat in policymaking. In particular, the artificial-intelligence industry is driving a gigantic increase in power consumption—while the Republican Party is doing all in its power to make sure that electricity is generated by pollution-spewing fossil fuels instead of clean renewables. The result will be a major increase in all kinds of illness, including cancer.
Read more from the Revolving Door Project
Data centers, of course, are used for things like cloud storage and computing, and they use a lot of electricity—about 176 terawatt-hours in 2023, or about 4.4 percent of total electricity production in the U.S. About 56 percent of that power comes from fossil fuels at present. The toxic air pollutants generated by fossil energy at all stages of production, transportation, and consumption are linked to an array of health risks, including cancers, asthma, and heart and respiratory diseases.
Now, tech companies have been promising for years that their investments will leverage clean energy. But Donald Trump and congressional Republicans just cut the legs out from under the American renewables industry by repealing President Biden’s subsidies for wind and solar—while also providing a fat subsidy for coal producers.
Indeed, AI infrastructure is already costing billions of dollars in added public-health costs, with one estimate putting the cost of increases in conditions like asthma and cancer due to data centers at $5.4 billion from 2019 to 2023. This cost is already ballooning, having grown by 20 percent from 2022 to 2023 and projected to be over $20 billion annually by 2030. That projection is based on data centers’ energy requirements increasing to between 7 and 12 percent of total electricity demand. If we were to come remotely close to the future dreamed of by some (admittedly unhinged) tech titans, one of whom told Congress that AI should be expected to eventually represent 99 percent of electricity demand, the public-health costs would be far higher.
But the cancer cost we’ll pay for continued artificial-intelligence build-out goes deeper than the billions it will cost society in added health expenditures. Just as a tumor can grow and spread far from its site of origin, having cancer metastasizes out into every aspect of life. Our current trajectory puts us on a course to inflict that fate on many Americans in the name of technology that is often as much boondoggle as breakthrough.
Meeting the Emperor
In February, I found out that I had a rare type of cancer called a carcinoid tumor (itself a subset of a rare category called neuroendocrine tumors).
I’d gotten pneumonia, and when I went to the clinic, a chest X-ray found a collapsed lung. The following day, a CT scan revealed the cause: a baseball-sized tumor that had overtaken the entirety of the airways into the upper half of the lung.
I was lucky; carcinoids generally don’t metastasize, and they grow very slowly. They can also usually be not just treated but outright cured via surgery. My luck held out pretty well through the entire process. But it was still agonizing in all the ways I knew to brace for, along with a thousand others that caught me completely off guard.
I was lucky to get sick when I did; my thoracic surgeon said that if the tumor had grown any more, it probably couldn’t have been removed and I would have had to get chemotherapy before attempting a surgical approach. Plus, I’m young, generally healthy, have good insurance, and have a reliable support network. I was extremely fortunate to have an employer who was endlessly understanding and supportive, even when there were weeks where I could barely get any work done at all.
I was lucky, aside from the giant tumor that shredded my lung beyond salvation.
I bring this up not to solicit sympathy, but to point out that the personal cost of having cancer, even in lucky cases like mine, is excruciating. And it’s deeply irresponsible to proceed with the exponential expansion of an industry that will afflict many thousands of people and families without discussing these statistically inevitable debilitating and demoralizing consequences. So my hope is that my experience can help to make clear the intangible suffering that can never be captured in topline cost-benefit analyses.
The Cost of Cancer
It’s now been more than two months since I had my lung removed. That’s more than two months cancer-free (my margins were entirely clear!). But I’m still very much living in the shadow of that experience and will be for some time.
For the next two years, I have to get CT scans every six months to make sure the tumor doesn’t come back, followed by annual screenings for a decade. I get winded much more easily than I’m used to. Exhaustion is almost a daily occurrence for me. And I still get intermittent stinging pain from where the surgery damaged the bundles of nerves that run along my ribs. And those were the things that I was prepared for.
I wasn’t prepared for my vocal cord to be damaged; after a few days of my voice approximating an asthmatic goose, I’ve now spent months sounding mostly somewhere between Marge Simpson and Robin Williams in Mrs. Doubtfire, which has been incredibly frustrating, to put it mildly. Now I shy away from striking up conversations in public so I don’t sound like a screeching witch from a cartoon in front of strangers. I’m physically incapable of singing or humming.
I wasn’t prepared for just how hard getting through the day could be and how difficult being able to get any work done has been. Even with an understanding boss, it feels bad on a psychological level to not be able to get things done. There’s internalized shame around the fact that something that I feel should take me a week winds up taking three.
There are so many parts of the recovery that I didn’t even know to brace myself for. No one told me that one of the drugs I was on when they sent me home had severe depression and suicidality as common side effects; I didn’t know that was why I couldn’t stop sobbing for the first day out of the hospital. No one told me that I wouldn’t be able to wear a bra for months because it aggravates the nerves that surgery had damaged. No one told me that, as the space my lung had occupied gradually filled with liquid, it would make unsettling gurgling noises.
Each of these, on its own, isn’t that big of a deal. But all of this, all at once, coalesces into a severe constriction of the way you experience the world. Cancer and recovery have limited what I could do, what I could wear, and what I could say. It made me go through life navigating a body that felt foreign (mostly from disconcerting chest-cavity sloshing), with a voice that wasn’t mine.
Such experiences are not unique to—or even especially pronounced in—my case. Survivors of breast cancer often live through losing one or both breasts. People losing their hair as they go through chemotherapy has become a pop culture staple (and gets at both limiting self-expression and your body feeling unfamiliar). Millions of people who survive cancer go on to experience long-term disability that impairs their movement and/or ability to care for themselves.
Little of this gets captured in topline policy analysis, but all of it is painful. All of it is a part of the cost of cancer.
AI: Cure or Cause?
Among the many proposed boons from advancing artificial intelligence is the acceleration of cutting-edge medical research, including toward the elusive cure for cancer. I hope that vision becomes reality, but the health of people in an indeterminate hypothetical cannot be permitted to override the health and well-being of real people’s lives. Even better than a cure would be preventing people from getting cancer and other chronic illnesses in the first place.
AI can be a powerful technology, but it eats up a lot of power. Despite the fact that Chinese firm DeepSeek demonstrated that you can create meaningful advancements without dramatic increases in infrastructure, the American tech industry is deeply committed to a single-minded emphasis on advancement through scalability.
According to Goldman Sachs, AI only accounts for about 14 percent of data center usage at present. But that is predicted to increase to 27 percent by 2027, at which point the data center industry will be 150 percent larger. Even before Republicans repealed Biden’s climate policy, much AI power demand was being met by natural gas plants, incentivizing the construction of new gas power plants and pipelines.
It’ll only get worse. Indeed, because wind and especially solar are the only possible way to satisfy AI’s skyrocketing energy demand with new capacity, in practice data centers will largely end up bidding up the price of existing fossil fuel power—jacking up Americans’ electricity bills while also poisoning their air.
Cancer and other health issues are well-known consequences of heavy fossil fuel infrastructure, to the point that there’s even a term for the areas whose inhabitants bear the brunt of the public-health costs: sacrifice zones. The most prominent sacrifice zone is probably Cancer Alley in Louisiana, where communities have been exposed to so many carcinogens that virtually everyone has lost family to cancer. According to chilling reporting from Human Rights Watch, some people have lost dozens of loved ones to the disease.
The energy infrastructure powering AI data centers exposes the public to multiple dangerous pollutants, including nitrogen oxides (NOx), formaldehyde, and tiny particulates measuring less than 2.5 micrometers (PM 2.5). In particular, PM 2.5 is a “nonthreshold pollutant,” meaning exposure at any level is unsafe. The polluting nature of data centers is made worse by the backup diesel generators used to keep the facilities running during power outages, which emit hundreds of times more NOx than a traditional gas power plant.
These risks are exacerbated by the fact that data centers are being sited more and more in densely populated areas so they can access existing power grids. This will increase the number of people facing health risks from the emissions. Data centers also disproportionately get built in poor, majority-minority areas (like Cancer Alley).
Smothering South Memphis
In the summer of 2024, Elon Musk’s xAI converted an abandoned factory in South Memphis into what Musk touts as the “most powerful AI training system in the world.” Christened Colossus, the enormous data center exemplifies the downsides of a pell-mell race to build out fossil fuel–powered AI systems.
Billed to Memphis as a crucial economic development that would provide hundreds of jobs, the details of Colossus were obscured from the public view. As state Rep. Justin Pearson pointed out in an op-ed for MSNBC, the data center’s operations have been shrouded from the public and sold on false promises. The site was converted to a massive data center without approval from the city council or community input. As part of the negotiations to build the facility, 14 city officials signed nondisclosure agreements that prevented them from publicly discussing details of the project. Other officials, including city council members and county commissioners, were reportedly unaware of the project until it became public knowledge. An xAI representative who was supposed to speak with the county board of commissioners did not show up to the meeting.
The facility began operating 35 methane-powered turbines onsite despite having no permit, then arguing that they were justified because of an exception for temporary turbines. However, the Southern Environmental Law Center has argued that, based on the specifications, xAI’s turbines are not subject to that exemption. The mayor of Memphis also defended xAI by asserting that only 15 of the turbines were actually being operated. However, thermal imaging of the facility showed that at least 33 were. That level of electricity generation is roughly equivalent to a medium-sized power plant.
In January, several months after the facility began operating, xAI applied for a permit to operate 15 of their turbines. Last week, the Shelby County Health Department granted the permit, over the objections of community members, who went as far as organizing a march in protest. The fight against greater pollution from the facility has now broadened to include officials in Mississippi as well. The Southern Environmental Law Center and the NAACP are moving to sue xAI over alleged Clean Air Act violations (this litigation would be, for context, the exact type of lawsuit that the “abundance” movement derides).
For its part, xAI is already planning a second data center in Memphis. Keshaun Pearson, co-founder of Memphis Community Against Pollution, made the case in Jacobin that Musk’s playbook for siting dangerous polluting data centers in poor areas over community objections is a road map that tech oligarchs may well look to apply across the country.
What Is It All For?
Despite AI’s promise, much of its current use ranges from useless to destructive. Social media feeds are now stuffed with “AI slop” that has had real-world downsides, like making it harder for emergency management personnel to use social media to help target aid during disasters.
Tech companies are making entire ecosystems that wind up making people lonelier and exacerbate mental health concerns. Recent evidence from MIT and OpenAI found that high usage of ChatGPT correlated to significantly more feelings of loneliness. More research is needed to make any definitive causal conclusions, but there’s enough there to pour some cold water on the idea that AI is actually the solution to our epoch of isolation.
Another recent study found that AI usage is impairing our intellectual abilities as well. That research builds on a rapidly expanding literature with similar findings. It would be a bleak future indeed if future generations have had their critical thinking abilities so atrophied by heavy AI use that they can’t function without it.
There is also the fact that generative AI is undermining our ability to comprehend the world around us in two distinct ways. First, many people, particularly those who didn’t grow up in our online age, can fall prey to the unending stream of AI-produced misinformation present in places like Facebook.
Add in the better-discussed harms of job destruction, energy price increases, AI-generated scams and frauds, and environmental damage, and it seems like the bill we’re footing for Silicon Valley’s latest dubious mania is awfully steep.