The data center debate is currently raging. Many argue that data centers are important to remaining competitive internationally in the global AI technology push. Those opposed say data centers are bad for the environment, overuse resources, create excessive noise, mar the landscape and do not create many jobs—among other reasons.
On April 11, local residents and activists gathered at the University of Michigan Diag in Ann Arbor for a rally opposing the unchecked development of massive data centers across rural and suburban communities. Among the speakers was State Representative James DeSana, a Republican representing Michigan’s 29th district, whose constituents are increasingly crossing traditional party lines to fight back against these data centers. DeSana said he was invited to speak at the Diag by “MOP Up Michigan”, which stands for Money Out of Politics.
Representative DeSana detailed why he joined forces with advocacy groups and why he believes the current trajectory of data center development poses a severe threat to Michigan’s environment, taxpayers and agricultural heritage.
“Data centers decrease the value of every resident around them within a mile around them, and that the people building the data, the companies, these multi-billion dollar companies that are building the data centers, have no intention of compensating people for their loss in property value,” DeSana said. “And that the low-frequency sound that comes out of the data centers is damaging to human beings.”
Local resistance in the Heartland
The pushback against data centers is no longer an isolated issue, but a coordinated effort spanning multiple townships in Michigan. While some local governments have already succumbed to corporate pressure, other communities are actively resisting the rezoning of local land. DeSana gave a sobering status report on where these various projects currently stand across the region.
“The Stargate in Saline, unfortunately, is proceeding,” DeSana said. “It’s under construction. It’s basically going to get built. It’s a terrible thing for Saline Township. It will change the entire character of the township, and it’s going to put this massive industrial building right out in the middle of the country, which is not good, and the other data centers like Ypsi Township, they’re fighting them. The data center in Van Buren Township, I think, is pretty much a done deal. I think they’ve approved the rezoning…”
Legislative pushback and repealing corporate tax breaks
To counter the influx of these facilities, bipartisan efforts are beginning to take shape in Lansing. Lawmakers from both sides of the aisle are starting to question the wisdom of using taxpayer-funded incentives to attract multi-billion-dollar tech conglomerates, especially when local residents bear the brunt of the negative externalities. DeSana is currently backing aggressive legislative measures to halt new builds and strip away previous financial incentives.
“Yes, there are Republicans and Democrats that are on the one-year moratorium, so we’ve got a bill that has a one-year moratorium, that we would have no data centers for one year in Michigan,” DeSana said. “I’m a co-sponsor on reversing the tax breaks that we gave the sales tax breaks that we gave these data centers over a year ago. I want to reverse those tax breaks. This is crazy that we’re giving multi-billion dollar corporations tax breaks on data centers that most of the people don’t want. 80% of the people don’t want these.”
Threatening farmland and a strained energy grid
One of the most glaring contradictions of the data center boom, according to DeSana, is the choice of location. Instead of utilizing abandoned industrial spaces, developers are targeting pristine agricultural land. According to many, this land-use choice, combined with the massive energy footprint required to power and cool thousands of servers, threatens to compromise both food security and the stability of Michigan’s electrical grid.
“Why are they going into virgin farmland and tearing up our agricultural communities and taking up thousands of acres of farmland and also taking up a lot of our capacity to produce electricity?” DeSana said. “Those are my biggest questions. Why, when we already have a strained grid, are we putting in data centers that are going to strain the grid even further, and why are we putting them out in the country when obviously, farmland preservation is a concern of all of ours? Why are we chewing up this much farmland when you consider the fact that it takes, you know, 28,000 acres of solar panels to run one data center.”
DeSana has a strong, environmentally helpful idea to put the data centers in areas that need to be cleaned up.
“Well, one of my biggest points I make with everybody is, why are these data centers not going into brownfield redevelopment? Why are they not going into contaminated sites that are now available for development?” DeSana said.
Protecting the taxpayer from environmental fallout
Looking ahead, DeSana warns that the long-term environmental and financial costs of these facilities will ultimately land on the public if proactive measures aren’t taken. He urges citizens to pressure their local governments to implement strict financial safeguards, ensuring that multi-billion-dollar tech companies—not Michigan families—are held accountable for the eventual decommissioning and cleanup of these industrial sites.
“They should demand that their township require these data centers to put up a bond to clean up their site at the end of life, when these buildings are no longer needed, and we’re going to have contaminated industrial sites, and guess who will pay for that…the taxpayers, of course, 100% every single time, and it isn’t even debatable that the US taxpayer, the Michigan taxpayer, ends up cleaning up contaminated industrial sites, so we should require them to put up money in advance to pay for the cleanup of these buildings, and that will definitely cause them to have to think twice about whether they can do this or not…”
Researcher Mosharaf Chowdhury discusses infrastructure, energy efficiency, and the growing demands of artificial intelligence
As artificial intelligence rapidly reshapes industries, education, and daily life, the demand for massive computing infrastructure is growing at an unprecedented pace. Behind every AI chatbot response, image generator, or large language model lies a vast network of data centers powering the technology.
According to Mosharaf Chowdhury of the University of Michigan, the pressure on computing infrastructure is only intensifying as AI systems become larger and more widely used.
“AI workloads (both training and inference) require enormous computational resources, and demand is growing rapidly,” Chowdhury explained. “Training a single large AI model can require thousands of GPUs working together for weeks or months. And inference, which is what happens every time someone sends a query to an AI service like ChatGPT or generates an image, adds up to an even larger footprint due to the sheer volume.”
He added that the current infrastructure may not be enough to sustain the pace of AI expansion.
“So from a purely technical standpoint, the existing infrastructure is not sufficient to meet current and projected demand,” he said.
Universities facing growing AI resource challenges
While much public discussion surrounding data centers focuses on economics, land use and energy consumption, Chowdhury emphasized the impact computing access has on higher education and academic research.
“Access to large-scale computing is essential for universities to remain competitive in AI research,” he said. “Right now, many top computer scientists are opting to work in industry instead of academia because only in industry can they access the computational resources they need.”
Chowdhury noted that limited computing resources affect both research and classroom instruction.
“For instance, our work on reducing AI energy consumption requires running large AI models, and we’re severely constrained by limited access to computing,” he said. “Even in my graduate course on AI systems, students don’t have access to the GPUs needed to work with the large-scale models we study.”
He warned that the long-term consequences could reshape the educational mission of universities.
“In the long run, this trend of leading AI researchers choosing industry over academia might also impact the education of our graduate and undergraduate students, which is the core mission of any university,” he said.
Searching for more energy-efficient AI
One of the central concerns surrounding new data centers is their growing energy consumption. Chowdhury’s research focuses on improving the efficiency of AI systems, and he says meaningful progress is already underway.
“Through our Zeus project at the University of Michigan, we’ve developed open-source tools that directly measure AI energy consumption from hardware, not envelope calculations based on worst-case assumptions,” he explained.
Using those measurements, researchers have already achieved substantial reductions in power usage.
“We’ve developed optimization techniques that can reduce training energy by 40% to 50% for large models,” Chowdhury said.
RELATED: So, Who is the Ann Arbor Responsible Energy Coalition?
He also highlighted the importance of transparency in measuring AI energy use.
“On the inference side, our ML Energy Leaderboard tracks the energy consumption of top open-source models, giving the community concrete data to work with when deciding which models to use,” he said.
Still, Chowdhury cautioned that software improvements alone will not solve the problem.
“Software-level efficiency improvements alone won’t eradicate the problem; it needs concerted efforts throughout the entire AI stack and in the physical data center infrastructures,” he said.
Efficiency improvements already exist
Chowdhury believes many efficiency solutions could be implemented immediately if companies choose to adopt them.
“Some of them could have an impact today,” he said. “Our measurement and optimization tools have been open-source for years, and we’ve seen strong interest in adoption from companies like Nvidia, who we’ve been collaborating with for almost a year on AI energy optimization.”
He also pointed to growing engagement from other technology leaders and academic fields.
“We’ve also seen adoption across academic research in fields like natural language processing, databases and computer architecture,” he said.
According to Chowdhury, the issue is less about inventing new technologies and more about encouraging widespread adoption.
“The challenge isn’t the technology; it’s adoption and incentives,” he said. “AI data centers increasingly become power-gated, meaning they can’t get more power from the grid, energy efficiency and performance are no longer tradeoffs.”
He said reducing energy per workload ultimately benefits operators as well.
“Using less energy per workload means you can run more workloads within the same power budget,” he explained. “The tools and techniques to make a meaningful difference already exist.”
Universities as public-interest research leaders
Chowdhury said universities play a unique role in developing open-source tools and independent research that may not emerge from the private sector alone.
“Universities play a critical role in developing the foundational research and open-source tools that the entire ecosystem benefits from,” he said. “Our Zeus project is a good example: it came out of academic research and is now being validated by industry.”
He added that universities often have incentives different from corporations.
“This kind of work is hard to do in industry because companies often don’t have the incentive to open-source energy optimization tools,” he said.
However, he warned that universities can only continue this work if they gain access to adequate computing resources.
“Investing in AI infrastructure resources at universities is an investment in the independent, public-interest research that can shape how this technology develops responsibly,” Chowdhury said.
Calls for transparency and better data
As communities across the country debate whether to approve large-scale data center developments, Chowdhury believes transparency should be central to the conversation.
“Too often, energy estimates for data centers are based on envelope calculations, multiplying maximum power draw by the number of servers, which overstates or misrepresents actual consumption patterns,” he said.
Instead, he argues that communities should demand measured, verifiable information from developers.
“Communities should ask for transparent, measured energy data and concrete plans for efficiency optimization, not just projections based on worst-case numbers,” Chowdhury said. “That kind of transparency benefits everyone.”
Chowdhury also noted that he and colleagues are working to create the Center for Informed Voices in AI Infrastructure Choices (CIVIC), an initiative aimed at bringing together expertise from policy, infrastructure, and environmental fields to guide public conversations on data center and AI infrastructure decisions.
Donna Marie Iadipaolo is a writer, journalist, and State of Michigan certified teacher, since 1990. She has written for national publications like The Village Voice, Ear Magazine of New Music, Insurance & Technology, and TheStreet.
She is now writing locally for many publications, including Current Magazine, Ann Arbor Family, and the Ann Arbor Independent. Her undergraduate degree is from the University of Michigan, Ann Arbor, where she graduated with an honors bachelor’s degree and three teacher certificate majors: mathematics, social sciences, English. She also earned three graduate degrees in Master of Science, Master of Arts, and Education Specialist Degree.
