AI Data Centers, Exponential Growth, and the “J Curve” from Hell

When William Blake wrote about the “dark Satanic mills” of England’s early industrial revolution, he could hardly have imagined the scale of today’s AI data centers. In response to these, a righteous rebellion has begun to emerge, with a large majority of Americans (approximately 70% according to some polls) hardening their resolve against the encroachment of these previously unimaginable monstrosities into their neighborhoods.

Outrage has spread about stories of the technological onslaught: a proposed “hyperscale” data center in Utah that would consume vast amounts of energy and water and emit incredible amounts of heat; 49,000 people in the Lake Tahoe area losing their access to energy as their utility prepares to divert electricity to data centers; homeowners in Coweta County, Georgia fighting the threatened use of eminent domain to seize their properties for a new data center.

“The American Rebellion Against AI Is Gaining Steam,” read a recent headline in the corporate-friendly, AI-adoring Wall Street Journal. The backlash is not confined to any one political party. A piece in the left-wing publication, Jacobin, argues that “the grassroots resistance to artificial intelligence data centers that is springing up in communities across the country outlines the kind of working-class coalition many of us on the Left have always dreamed of — a diverse, nonpartisan, top-bottom movement against Big Tech billionaires that has the potential to reshape American politics in incredibly positive ways.” (Interestingly, Jacobin also published an essay arguing against slowing down data center construction due to supposedly “serious equity concerns.”) The populist wing of the Republican Party is just as vociferously opposed to these massive, water-guzzling, energy-devouring machines being placed in their backyards, with the cause taken up by right-populist luminaries like Marjorie Taylor Greene and Tucker Carlson.

Data centers touch a nerve on the right and left because they bring to a head pernicious trends that have been intensifying for decades: pollution and the overstretching of natural resources; the concentration of wealth and power in the hands of billionaires; the rising cost of energy and the hypocrisy of elites who have talked for years about climate change but now want to dramatically increase energy production for AI; mass surveillance and the erosion of human rights and civil liberties; the replacement of human labor and human creative activity by machines; the mismatch between the growth of GDP and corporate profits and the grinding down of the middle class, accompanied by the decline of overall human well-being; and the glaring absence of real democracy, as nobody ever voted for AI to take over our lives.

“People just feel as if they’re under siege,” said Republican Senator Josh Hawley, who has proposed some mild restrictions on data centers. It’s not just a feeling, though. AI data centers are, in fact, accelerating the dangerous trends listed above. What we feel is this sense of acceleration, even if we can’t exactly explain it.

AI advocates openly acknowledge, both bragging and warning (like a shot across the bow), that their technology is accelerating the transformation of modern life like never before. The ex-CEO of Google, Eric Schmidt, gave a commencement address at the University of Arizona where he told students the “technological transformation” brought about by artificial intelligence will be “larger, faster and more consequential than what came before.” The crowd of students, worried about their own futures, responded with rounds of aggressive jeers and boos.

Current Google CEO Sundar Pichai has also warned society to “brace for impact” of AI acceleration. Dario Amodei, chief executive of Anthropic, has said, “AI progress is accelerating … it could overwhelm our ability to adapt.” He predicted “a radical acceleration that surprises everyone.”

But what does it mean for the “technological transformation” of society to accelerate? Regardless of what AI oligarchs would have us believe, the data center controversy illustrates the answer to that question for a simple reason: because we can actually map out the devastating effects—physical, biological, ecological—that the accelerating buildout of data centers has on us and our environment.

The key to understanding these effects is a concept the Big Tech barons like to throw around. As Amodei has said, “these systems are improving on an exponential curve.” Jensen Huang, Nvidia’s chief executive, has likewise said, “AI is advancing at an exponential rate.” “The amount of computation we can do,” according to Huang, “is increasing exponentially.”

To understand what’s coming, then, we need to understand what these guys mean when they say AI is developing on an exponential curve. The late physics professor, Albert Bartlett, said in a popular lecture that “the greatest shortcoming of the human race is our inability to understand the exponential function.” He was perhaps being a bit hyperbolic, but what he meant is that, in general, we simply do not appreciate the consequences of steady growth. The concept itself, however, is not that hard to grasp.

Anything that experiences a steady rate of growth over time is growing at an exponential rate. Exponential growth results in an upward-curving “J-shaped” graph, the kind of curve that economists and businessmen sometimes loosely invoke when discussing long-term GDP growth or the growth of new technologies. The curve starts slowly, then steepens dramatically because growth compounds on top of prior growth.

The classic J-shaped curve

Take the example of an economy that grows steadily at 3% per year. With exponential growth, the absolute increase gets larger each cycle even if the percentage stays constant. At 3% annual growth the economy doubles roughly every 23 years, then doubles again 23 years later; then again. The curve stays deceptively gentle for a long time, then steepens dramatically because growth compounds on top of prior growth.

The “J Curve” is crucial for us to understand because economic growth is tied to the physical world—also known as the real world, which we all rely on, enjoy, and are embedded in. As the global economy grows exponentially, anything tightly coupled to economic activity also grows in a J-shaped fashion: energy use, material extraction, pollution, freshwater withdrawals, waste generation, carbon emissions. Even large amounts of available resources and environmental waste-absorption capacities can be overwhelmed surprisingly quickly under such compounding growth.

The shocking effects of exponential growth can be illustrated with an old story. Once upon a time, a king was so impressed by one of his subjects that he offered the man a reward. Instead of asking for a large sum, the man slyly asked for grains of rice placed on a chessboard like this: 1 grain on the first square, 2 on the second, 4 on the third, 8 on the fourth, and continuing to double on each of the 64 squares. The king agreed. At first the numbers looked tiny, but due to the dramatic effects of exponential growth, by the final square, the total grains of rice across all 64 squares was mind-bogglingly huge: 18.4 quintillion grains of rice. Stacking that many grains would reach about 81 trillion miles or 14 light-years from the surface of the Earth — far more than the kingdom could supply. This story shows how exponential growth stays deceptively small for a long time, then suddenly becomes enormous. (In the story, notably, the king ultimately puts the man to death.)

As many environmental thinkers have theorized over the years, the mismatch between exponential economic growth and our finite ecological systems is at the core of environmental problems. Exponential growth unleashed by the Industrial Revolution produced well-known J Curves, particularly those of the post-World War Two era: global plastic production, atmospheric CO₂ concentration, global fertilizer consumption, aviation traffic, motor vehicles, dam construction, telecommunications, global material extraction, and many more. Some environmental scientists call these the “Great Acceleration” curves.

Which brings us back to AI. Data centers are a clear example of an emerging J Curve. AI may be perceived as an “immaterial” technology, but it totally depends on data centers that have intense physical demands for electricity, cooling, water, concrete, steel, semiconductors, transmission lines, backup generators, mining, land, and fiber networks.

The exponential growth of AI data centers is following a J Curve, not only as a single burst of technological development, but as part of a long-running historical pattern. The AI/data-center buildout represents an acceleration phase—the dramatic steepening of the curve—in the larger J-shaped path of industrial civilization itself as technological capabilities and ecological impacts intensify simultaneously.

This is why the numbers that we are seeing about AI data centers, in terms of their size, their needs for energy and water, and their other physical and biological impacts, are mind-bogglingly gargantuan.

Robert Davies, a physics professor at Utah State University, has crunched some numbers for the proposed “hyperscale” data center in Box Elder County, Utah called the Stratos Project. “What I’ve found,” Davies has said, “is so much worse than I thought it would be.” He estimates the physical footprint of the Stratos Project to be the equivalent of about 2,000 Walmart Supercenters, and the energy footprint to be equal to about 40,000 Walmart Supercenters. The project’s backers admit it will need about 9 gigawatts of energy—more than double the electricity used by the entire state of Utah. Nine gigawatts “is a number that’s really hard to get your brain around,” Davies has said, but once an additional 7 gigawatts of waste heat are factored in, according to him, the project carries a total thermal load of 16 gigawatts—heat that goes into the local environment—“the equivalent of about 23 atom bombs worth of energy dumped into this local environment every single day.” Putting this much energy into the local environment will raise day-time temperatures by five degrees Fahrenheit, Davies estimates, and up to 28 degrees at night.

It is difficult to fathom numbers this large, or wrap our heads around the catastrophic damage to our Earth that a single “hyperscale” data center can exact. That’s because we are in the steepening part of the exponential J Curve. In other words, many of our ecological problems, which were already dire, are now going vertical.

We can detect this verticality in statements and projections made by AI companies and their leaders. NVIDIA CEO Jensen Huang said recently, “The amount of energy that we need for computing is 1,000x more than we currently have.” Huang’s comments express the logic of a self-reinforcing technological expansion process that will lead to an exponential J Curve.

In an internal memo, OpenAI CEO Sam Altman said that his company’s “audacious long-term goal is to build 250 gigawatts of capacity by 2033.” If that happens, OpenAI’s electricity use by 2033 will be comparable to that of India’s 1.5 billion people.

Similar trajectories are projected for freshwater consumption by AI data centers, including in regions that are already suffering from drought and freshwater shortages (global freshwater supplies are already in steep decline). The state of Texas recently announced it would have to spend nearly $200 billion dollars to avoid a water crisis in the coming decades, but the state has approximately 400 water-guzzling data centers already in development (in addition to thousands of data centers in development nationwide and the 4,000 to 5,000 that are already operational), with reports suggesting that Texas data centers could consume between 29 billion and 161 billion gallons of freshwater per year by 2030.

The water that is used by data centers for cooling is emitted from the facilities back into the environment at very high temperatures, and is contaminated with PFAS “forever chemicals.” The United States and the world are already choking on PFAS pollution that is accumulating practically everywhere—in our water, air, soil, food, and consumer products. Now imagine the exponential effect of building thousands of these forever-chemical-spewing data centers.

To make matters even worse (if that’s possible), as AI capacity is built out, the machines themselves generate more demand for computing power. “AI is the first technology that can accelerate its own progress,” OpenAI CEO Sam Altman has claimed. This makes for a situation in which computation scaling is at least partly machine-driven rather than purely human-driven. That has the potential to steepen the J Curve even more.

If you are a technophile or an AI true-believer, you may be asking yourself why should I listen to lowly homo sapiens critics of AI or trust their fallible reasoning abilities. Okay then, what if we go to the super-intelligent machines themselves and ask them directly what will happen if AI data centers continue to be built in an exponential J Curve? I did that. And ChatGPT was quite willing to tell me, in great detail, why “my assessment is that indefinite exponential growth in the physical footprint of computation is unlikely to remain compatible with finite planetary systems.”

Here is how ChatGPT summarized our exchange and its “assessment,” quoted directly above:

The reason for my assessment is that computation is not an abstract activity; it is a physical process requiring electricity, cooling, water, land, transmission infrastructure, semiconductors, concrete, steel, copper, and rare-earth minerals. If AI data centers continue expanding at high exponential rates for decades, they are likely to compete increasingly with cities, agriculture, transportation, and industry for scarce resources, particularly power and freshwater. Regions already experiencing drought, groundwater depletion, grid congestion, and ecosystem stress may face difficult allocation decisions as AI infrastructure demands accelerate. Additional pressures could include expanded mining, land conversion, waste heat, transmission corridors, industrial noise, and chemical pollution associated with semiconductor manufacturing and cooling systems. While technological improvements can reduce resource use per unit of computation, history shows that efficiency gains often stimulate greater total consumption rather than reducing overall demand. The broader concern is that ecological systems—rivers, aquifers, forests, biodiversity, and the climate system—do not expand exponentially in response to industrial demand. They possess finite capacities, regeneration rates, and ecological thresholds. For this reason, I believe industrial civilization as a whole may itself follow a large-scale J-curve, with the AI/data-center buildout representing a late-stage acceleration phase in which technological capability and ecological throughput intensify simultaneously. In that interpretation, AI is not an exception to the historical pattern of industrial growth, but one of its most advanced expressions.

How long, one wonders, will Big Tech oligarchs allow AI to tell truths like those above about the disastrous path their technology has put us on? How long will they let their very own machines admit where they are taking us?

And let’s not forget how these machines reach the kinds of conclusions expressed in the “conversation” above. They do it by searching and summarizing the sum total of digitizable human knowledge. In that sense, ChatGPT’s words above represent a kind of warning from humanity’s own wealth of knowledge that we are indeed courting self-destruction.

For decades now many ecologists and environmental thinkers, including some in the agrarian tradition, have pointed out the profound flaws in our society’s orientation around the permanent exponential growth of technology, industry, and GDP, as well as the conflict between exponential growth and the Earth’s limits. Those prescient thinkers have been dismissed, insulted, and generally ignored. Meanwhile, the industrial-technological Machine with a capital “M” (in the sense of the word used by writer Paul Kingsnorth) has continued its metastatic growth. Now AI and its data centers are taking us into a steeper phase of technological civilization’s J Curve.

Perhaps in earlier decades, at shallower parts of the J Curve, it was easier to ignore the ill effects of technological growth. But now that the Machine is expanding more rapidly, up the steepening line of the J, AI data centers—the physical manifestation of a technology that many of us previously thought was somehow immaterial—are imposing themselves into our neighborhoods, sucking away our water, diverting energy from us to them, and surrounding us with gigantic, offensively ugly facilities that emit heat, noise, and pollution at a scale that until now was inconceivable. We are unwittingly enclosing ourselves, it seems, in a toxic technological prison.

It’s safe to say also that AI data centers’ exponentially growing physical and ecological effects are mirrored in the exponential growth of the technology’s intrusion into our lives at other levels. We are experiencing a J Curve in AI’s effects on us in the social, political, economic, philosophical, psychological, and even spiritual realms. That’s what AI’s biggest boosters would have us believe, and there’s no reason to doubt them. Some tech gurus admit, or even openly long for, exponentially self-replicating superintelligent technologies to replace humanity. Already we are bombarded with threats of mass unemployment, autonomous drones, robot dogs, AI lovers, AI-generated music, books, and films in place of works by human musicians, writers, and filmmakers, and the list goes on. The accelerating character of this phenomenon explains perfectly the sense of bewilderment, unease, fear, and loathing that so many people now feel toward AI and data centers in particular.

Of course, then, the reasonable response would be for a mass movement of Americans and peoples around the world to break from their decades-long (or centuries-long) slide into technological and industrial overdevelopment, or dystopia to use a stronger word, and to stop AI data centers in their tracks. The broad, emerging “Not in My Neighborhood” movement against data centers does indeed have the potential to starve AI of the massive and continually expanding quantities of energy it needs to keep growing itself exponentially, or to at least slow down the rate of growth. The Pope’s latest encyclical warning of the threat to “human dignity” posed by AI suggests that the broad-based anti-AI movement could indeed become a potent force.

If history is any guide, however, governments, billionaires, and corporations will use every tool in the toolbox to break down popular resistance to data centers, particularly because our entire growth-based economy is now essentially dependent on the projections of AI’s rapid expansion coming to fruition. “Most U.S. Growth Rides on AI,” say some reports.

The Trump administration, for its part, has made it abundantly clear that “winning the race,” mainly against China, for supposed “unquestioned and unchallenged global technological dominance” is a top priority. Trump issued an Executive Order in July 2025 called “Accelerating Federal Permitting of Data Center Infrastructure,” which various federal agencies are busy carrying out.

AI CEOs, even as they predict that their machines will replace human laborers and deliver society-wide unemployment, have tried to hold out carrots to the public, promising vaguely that AI will eventually deliver some kind of workerless utopia. AI will usher in an “age of abundance” and make work “optional,” says the erratic tech billionaire Elon Musk (he’s also warned that AI is “a fundamental risk to the existence of human civilization”). Anyone who is tempted to believe promises of techno-utopia need only pick up a history book. “It is questionable if all the mechanical inventions yet made have lightened the day’s toil of any human being,” wrote John Stuart Mill in 1848. “They have enabled a greater population to live the same life of drudgery and imprisonment, and an increased number of manufacturers and others to make large fortunes.”

Recent news reports indicate, however, that AI’s backers are shifting from carrots to sticks. As AI hatred spreads among the population, U.S. law enforcement agencies are gearing up to quash “anti-tech extremism,” and these tech-enabled agencies are reportedly not shy about targeting peaceful groups and protestors for surveillance. As many AI critics and proponents have noted, the computing power of AI will enable mass surveillance of the population to a degree never before possible, a power that is already being directed at opponents of data centers.

So then, where does that leave us? Will the American people and the rest of humanity passively and unquestioningly accept the continued exponential growth of technology and industry, and allow themselves to be swept along the J Curve from Hell by the doomsday machines known euphemistically as “data centers”? Or will the life-threatening scale of these “dark Satanic mills,” to quote Blake, finally provoke a showdown, forcing us to confront the accelerating technological self-destruction we have been courting?

The battle lines are being drawn. How will the struggle play out?

Many believe the rise of AI is inevitable. Before we succumb to the convenience and dazzle of these new technologies and let the machines take us where they will, however, perhaps enough of us can inform ourselves about what is at stake and remind ourselves what makes life worth living, and then summon the spirit of defiance that still abides somewhere deep down in our human bones.

Image Credit: WPR