The machines are learning. And they are thirsty.
Somewhere in the Arizona desert, a data center hums with the low-frequency drone of a billion transistors firing in lockstep. The sound is the heartbeat of the artificial intelligence revolution — a revolution that promises to remake everything from the way we work to the way we wage war. But that heartbeat comes at a cost the tech barons have been slow to tally, and the bill is coming due in water, watts, and carbon.
New Jersey is also experiencing a massive boom in data center and AI infrastructure development, with over 80 facilities already operating or in development across the state.
CoreWeave is building a $1.8 billion, 140-megawatt AI data center spanning 400,000 square feet at the former Shering Plough site in Kenilworth,
DataOne and Nebius have secured local tax incentives to construct a hyperscale 300- to 400-megawatt facility in Vineland, one of the largest such developments in the region, because it only makes sense to allow profiteers to destroy the economy and environment if taxpayers are funding it.
The numbers, now laid bare by researchers at Cornell University, are staggering enough to give a Texan pause in a drought. Due to massive energy loads—with some facilities using as much power as 200,000 homes—local electric bills spiked, prompting swift political action.
By 2030, if the current rate of AI expansion holds, the data centers powering America’s digital future will pump somewhere between 24 million and 44 million metric tons of carbon dioxide into the atmosphere each year.
That is the equivalent of adding 5 million to 10 million new automobiles to the nation’s roadways — a traffic jam in the sky. At the same time, these server farms will drain 731 million to 1.125 billion cubic meters of water annually, enough to supply the household needs of 6 million to 10 million Americans.
The findings, published Nov. 10 in the journal Nature Sustainability, represent the first state-by-state accounting of AI’s environmental footprint. And the verdict, as they say in the newsroom, is not good.
“Artificial intelligence is changing every sector of society, but its rapid growth comes with a real footprint in energy, water and carbon,” said Fengqi You, the Roxanne E. and Michael J. Zak Professor in Energy Systems Engineering at Cornell, who led the three-year study. “Our study is built to answer a simple question: Given the magnitude of the AI computing boom, what environmental trajectory will it take? And more importantly, what choices steer it toward sustainability?”
The answers, as it turns out, depend a great deal on where you plant your servers and how fast you can decarbonize the juice that powers them.
Professor You and his team — which included researchers from Sweden, Canada and Italy — spent three years compiling what he called “multiple dimensions” of financial, marketing and manufacturing data.
They mapped location-specific information on power systems and water resources against the projected expansion of AI infrastructure. They even used AI itself to fill the gaps in corporate reporting, which, as Professor You noted, is about as transparent as a Vegas poker game.
“Sustainability information, like energy, water, and climate, tends to be open and public,” Professor You said. “But industrial data is hard, because not every company is reporting everything.”
Location, Location, Location
The study’s most striking finding is also its most practical: Where you build matters more than almost anything else.
Many of the nation’s current data clusters are rising in water-scarce regions like Nevada and Arizona. Northern Virginia, which hosts about one-eighth of the world’s data center capacity, is rapidly clustering so tightly that local infrastructure and water resources are buckling under the strain.
The solution, the researchers found, is to build smarter. Locating facilities in regions with lower water stress and improving cooling efficiency could slash water demands by about 52%. When combined with grid decarbonization and operational best practices, total water reductions could reach 86%. Carbon emissions, under the same approach, could fall by roughly 73%.
The sweet spots? The Midwest and the so-called “windbelt” states — particularly Texas, Montana, Nebraska and South Dakota — offer the best combined carbon-and-water profile. New York, Professor You noted, remains a low-carbon option thanks to its mix of nuclear, hydropower and growing renewables, though water-efficient cooling and additional clean power are essential.
“There isn’t a silver bullet,” Professor You said. “Siting, grid decarbonization and efficient operations work together — that’s how you get reductions on the order of roughly 73% for carbon and 86% for water.”
The Net-Zero Mirage
The big tech companies that have pledged to reach net-zero emissions by 2030 — names like Google, Microsoft and Meta — may find those promises harder to keep than they imagined.
Even if the grid decarbonizes, the researchers warn, total emissions could still rise roughly 20% if AI demand grows faster than the clean-energy transition.
In the most ambitious high-renewables scenario, carbon dioxide would drop only about 15% by 2030, leaving approximately 11 million tons of residual emissions. Reaching net-zero would require roughly 28 gigawatts of wind or 43 gigawatts of solar capacity — numbers that make the current build-out look like a drop in the bucket.
“Even if each kilowatt-hour gets cleaner, total emissions can rise if AI demand grows faster than the grid decarbonizes,” Professor You said. “The solution is to accelerate the clean-energy transition in the same places where AI computing is expanding.”
Deploying advanced liquid cooling and improved server utilization could shave off another 7% of carbon dioxide and lower water use by 29%. But the researchers caution that these are incremental gains in what amounts to a sprint.
A Sovereign Solution?
If the tech industry won’t police itself, Sen. Bernie Sanders has a proposal that would give the American people a seat at the table — and a piece of the action.
In June, the Vermont independent introduced the American AI Sovereign Wealth Fund Act, legislation that would give the public a 50% ownership stake in the largest AI companies through a one-time tax on stock.
At current valuations, the resulting fund would be worth an estimated $7 trillion. A 5% annual dividend, Sanders said, could provide more than $1,000 to every American.
“Left unchecked, artificial intelligence and robotics threaten the jobs, privacy rights and mental health of every man, woman and child in America,” Sanders said in a statement. “As a society, we can no longer sit back and allow a handful of Big Tech oligarchs to determine the future of this revolutionary technology with no democratic input.”
The bill would create an Independent Commission for Democratic AI — seven members nominated by the president and confirmed by the Senate — to manage the fund and use voting shares to block decisions that harm the public interest. It would also require large companies to split their AI and non-AI businesses.
“The future of AI and the fate of humanity must not be decided behind closed doors in Silicon Valley by billionaires seeking to maximize their power and profit,” Sanders said. “It must be decided by workers, parents, teachers, artists, scientists, communities and the American people.”
The Broader Burden
The environmental cost of AI, as Mahmut Kandemir of Penn State has documented, extends far beyond the data center walls.
In 2023, data centers consumed 4.4% of U.S. electricity — a number that could triple by 2028. By 2030 to 2035, Kandemir warned, data centers could account for 20% of global electricity use, putting immense strain on power grids. The short lifespan of GPUs and other high-performance computing components creates a growing mountain of electronic waste, and manufacturing these components requires the extraction of rare earth minerals — a process that depletes natural resources and degrades the environment.
Kandemir, a distinguished professor of computer science and engineering at Penn State, has spent his career optimizing computer systems for speed and efficiency. Now, he said, he sees an unprecedented connection between his research and its environmental impact.
“Only a handful of organizations such as Google, Microsoft and Amazon can afford to train large-scale models due to the immense costs associated with hardware, electricity, cooling and maintenance,” Kandemir said.
The Build-Out Moment
The Cornell study, supported by the National Science Foundation and the Eric and Wendy Schmidt AI in Science program, arrives at a pivotal moment. Companies like OpenAI and Google are funneling billions into rapidly building AI data centers to keep up with demand. The choices made this decade, according to Professor You, will echo for generations.
“This is the build-out moment,” he said. “The AI infrastructure choices we make this decade will decide whether AI accelerates climate progress or becomes a new environmental burden.”
The machines are learning. The question is whether we are learning with them — or learning the hard way.
Discover more from NJTODAY.NET
Subscribe to get the latest posts sent to your email.