Michigan Study: AI Energy Use May Drop 50%, Cutting Costs

Michigan Researchers Say AI Could Soon Use Half the Power

AI is everywhere these days, from voice assistants in your home to the algorithms that decide what you see online. But all that computing power comes at a cost: massive energy bills and a growing environmental footprint. Now, a new study from researchers in Michigan claims that AI systems could soon use 50% less energy than they do today. For anyone living near a data center, worried about rising utility bills, or concerned about climate change, this could be a big deal.

What Happened

The study, published by a team at the University of Michigan, looked at how improvements in AI hardware and smarter software could cut energy use in half. This matters because big tech companies like Google, Amazon, and Microsoft run huge data centers packed with servers that power AI applications. These facilities can use as much electricity as a small city. In 2023, for example, Google's data centers consumed over 18 terawatt-hours of electricity—roughly as much as the city of San Francisco uses in a year.

Data centers are also major sources of greenhouse gas emissions and can strain local water supplies, especially in areas already facing drought. The Michigan study suggests that by switching to more efficient chips and optimizing how AI models are trained and run, the tech industry could cut energy use by 50% within the next few years. This could mean fewer emissions and less stress on local grids.

How This Affects Everyday People

So what does a 50% drop in AI energy use mean for regular people? First, it could help lower utility bills in towns near big data centers. These facilities often drive up local electricity prices because of their huge demand. If data centers use less power, it could take some pressure off the grid and keep rates more stable for everyone else.

Second, less energy use means less pollution. Data centers are often powered by fossil fuels, especially in states like Michigan where coal and natural gas are still common. Cutting their energy use in half could reduce local air pollution and help fight climate change. For families with kids who have asthma or for communities worried about air quality, this is no small thing.

Third, there’s the issue of water. Many people don’t realize that AI data centers also use massive amounts of water to keep their servers cool. Lower energy use could mean less water is needed, which matters in places where every drop counts. For more on this, see our explainer on how much water AI uses.

However, it’s important to note what this change doesn’t address. Lower energy use won’t bring back jobs lost to automation, nor will it solve concerns about privacy or AI’s influence on children. The study is good news for the environment, but it doesn’t erase every worry people have about AI’s growing role in daily life.

The Bigger Picture

This Michigan study is part of a larger trend. As AI becomes more common, its environmental impact is coming under greater scrutiny. In 2024, the European Union passed the Energy Efficiency Directive, which requires large data centers to report their energy use and take steps to reduce it. In the U.S., several states are considering similar rules, though none have passed yet at the federal level.

Tech companies are also feeling pressure from local communities. In 2025, residents in Mesa, Arizona, protested a planned data center over concerns about water use and rising electric bills. Stories like these have led to more public pushback, as covered in our report on why people are pushing back on AI. The Michigan study gives tech companies a possible answer to these criticisms: if they can cut energy use in half, they may be able to expand without as much local opposition.

But there’s another side to the story. Even with a 50% reduction, AI’s energy use is still enormous and growing fast. As more schools, hospitals, and local governments adopt AI tools, the total demand keeps rising. And while the Michigan study points to technical solutions, it doesn’t address who pays for upgrades, or whether savings will be passed on to regular people. For a closer look at where these facilities are popping up, check our AI data centers near you resource.

What You Can Do

• Ask your local officials about data centers. Find out if any are planned or operating nearby, and ask how they affect local energy and water use.
• Check your energy bill for surcharges. Some utility companies add fees to cover the extra demand from data centers. If you see new charges, ask your provider to explain them.
• Support clean energy policies. Vote for local or state measures that require data centers to use renewable energy or report their environmental impact.
• Educate your family about AI’s environmental footprint. Talk with your kids about how their favorite apps and games use energy, and why it matters.
• Stay informed. Follow updates on AI’s water and energy footprint with our plain-English explainer.

The Bottom Line

The Michigan study offers a rare bit of good news in the debate over AI’s impact: smarter technology could soon mean less pollution and lower costs for local communities. But energy savings alone won’t solve every problem. As AI grows, it’s up to all of us to ask tough questions about who benefits, who pays, and how these changes shape our daily lives. Staying informed and involved is the best way to make sure technology works for people—not the other way around.