By Amy Myers Jaffe

We all know AI's dirty secret: It gobbles up a huge amount of electricity -- and spits out a large volume of greenhouse gases in the process.

But what if using AI can also save us energy?

Artificial intelligence has the potential to drastically slash energy demand across a swath of industries and cut down on their carbon emissions. And it may be so effective that it will easily balance out its own power demands and carbon emissions -- and more.

In my research, I have found many ways that AI technology will streamline industries and make them greener. It can help discover and develop more eco-friendly materials for manufacturing. It will manage systems in big buildings -- everything from the HVAC to the elevators -- so they use less power. It is already remaking transportation, planning routes and timetables to cut down on miles traveled and fuel guzzled, based on real-time data.

Here is a closer look at some of the ways AI may make a big dent in energy use and greenhouse-gas emissions.

Transportation gets smarter

AI has already had striking effects on the transportation business. AI-driven route planning has helped some of the major U.S. freight companies cut fuel use in ground vehicles -- in some cases by as much as 5% to 10% -- by simply lowering the miles they must travel. Recent studies suggest the whole ground-freight industry could cut its emissions by 10% to 15% by using AI-led dynamic route optimization in all vehicles.

AI can analyze traffic in real time, and is starting to get better at guiding vehicles away from busy areas, according to major delivery companies, reducing the fuel wasted by stop-and-go driving. And sitting in traffic adds up to a lot of pointless emissions: Americans wasted 3.3 billion gallons of gasoline and diesel fuel in 2022 -- over 215,000 barrels a day of petroleum -- sitting in traffic, according to Texas A&M.

Some aspects of route planning are much less visible but make a big difference in carbon emissions. We all know that e-tailers cluster deliveries together to save miles traveled. But even before the trucks are loaded with packages, a crucial form of routing goes on behind the scenes. AI-enabled logistics predicts what goods people will be ordering, and where and when. That way, e-tailers can stock their distribution centers according to probable local demand, which means fewer miles spent on deliveries and less emissions from overproduction of goods that aren't able to be sold.

The potential savings don't stop on the ground. Airlines are also using AI to help with dynamic route planning. For instance, they are using AI to analyze wind conditions in real time to find ways to use the least amount of fuel -- planning the best routes to minimize wind resistance, as well as finding the best airspeed and altitudes to travel and even the best type of plane for the weather conditions. Likewise, AI can pick routes and altitudes that make the most of tailwinds and reduce turbulence. These practices and others have reduced emissions by between 3% to 10% a year, airlines say.

Marine freight is getting the same kind of logistical help. AI intelligence can calculate the best times for ships to "slow steam" -- lower their speed -- which can greatly boost efficiency: A 10% drop in speed cuts fuel use by 20%. Improving traffic at ports can also cut down on wasted fuel. Ships burn as much as seven to 10 tons a day of fuel while anchored near ports, waiting for congestion to clear. AI-assisted programs help shippers lower the waiting period by timing their arrivals at port efficiently, and the ports themselves use AI to better schedule vessels, among other things, to clear up congestion.

Taken together, those strategies -- among many others -- could make a substantial difference in canceling out the emissions from AI data centers. In its "Widespread AI Adoption" scenario, the International Energy Agency suggests that the spread of AI in the transportation sector alone could slash 900 million metric tons of carbon emissions by 2035. Other studies suggest the savings could be higher. In comparison, the agency expects emissions from data-center electricity use to rise to 300 to 500 million metric tons by 2035, up from 180 million metric tons today.

Running buildings by algorithm

Big buildings churn out a lot of carbon. Looking at the whole lifetime of buildings -- from construction through daily operation, a concept called "embodied carbon" -- the sector contributes almost 40% of all global greenhouse-gas emissions. Here, too, AI holds a lot of potential to help. A study suggests that using AI to run operational systems in buildings could cut the sector's carbon emissions by 8% to 19% by 2050, depending on the pace of adoption.

Already, commercial and large residential buildings are upgrading energy-management systems, using AI to reduce energy use and lower operational costs. For instance, sensors can track occupancy in real time and shut down some elevator banks and turn off lights that aren't needed when there aren't many people around, saving energy and slashing emissions. AI-powered HVAC systems can precool buildings ahead of forecast heat waves to lower energy use. Smart window-shading systems can respond to sun angles to avoid glare and reduce heating effects in summer months.

Another solution is already in play in buildings -- large and small -- that generate their own power on-site. AI can charge a building's batteries and use the regular electric grid when rates are low. Then the AI can tap in to the batteries in the late afternoon and early evening, when solar energy is fading and demand on the wider grid is peaking. This not only saves building owners money but reduces the amount of nonrenewable energy needed from the wider grid.

(Utilities are also using this kind of AI scheduling to find the best times to send more renewable resources to the grid. AI can even predict solar radiation and wind speeds, to figure out when renewables will be generating the most and least power.)

Raw materials go green

The raw materials we need in manufacturing are another big source of carbon -- starting with emissions cranked out during extraction and continuing as materials are transformed into usable products like cement, steel and plastic. And the list goes on.

AI can make that long process much cleaner -- first, by accelerating material-science innovation. Much the way AI is speeding pharmaceutical research, engineers are now able to use AI to calculate the properties of new materials based on their chemical composition and structure. That means shortened research-and-development timelines and faster discovery of green materials that can replace conventional ones in areas like construction, as well as manufacturing and petrochemicals.

AI is also reducing carbon emissions by coming up with recycling solutions that replace the high-emissions manufacturing of new products. AI, for instance, can find ways to use metal scraps to reduce waste, and identify biodegradable alternatives to substitute for materials like plastics that now overflow landfills.

Making better use of nature

One of the best weapons against greenhouse-gas emissions is nature itself, in the form of carbon sinks -- areas like forests that absorb more carbon than they churn out. AI can help strengthen that natural defense by finding the best ways to preserve the carbon sinks we have, and restore the ones that we've lost. To give an idea of the scale involved, according to one database of carbon credits issued, "U.S. forest carbon-offset projects have issued nearly 204 million credits [for adding forested land] throughout their history" -- the equivalent of nearly 204 million metric tons of carbon dioxide.

And AI offers a lot of room for improvement. AI systems let us analyze satellite, drone and ground-based imagery to monitor land conditions so we can focus reforestation and sequestration efforts on the most promising places. AI can also study soil quality to remediate soil erosion and land degradation to advance soil's natural carbon-collection properties. Remote-sensing systems can detect disease and pests early, to minimize crop and forestland losses.

Of course, the low-carbon AI future is far from guaranteed. Among many other potential issues, there is one that we're wrestling with right now: privacy. Do most people really want tech companies to know when we are at home or not and what temperature we prefer?

Those questions -- along with other concerns -- need to be addressed. But AI also holds tremendous promise to solve the very problems it creates. We should take advantage of that potential.

Amy Myers Jaffe is director of the Energy, Climate Justice, and Sustainability Lab and a research professor at New York University's School of Professional Studies, and author of "Energy's Digital Future." She can be reached at reports@wsj.com.

(END) Dow Jones Newswires

September 16, 2025 12:00 ET (16:00 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.