Artificial intelligence has caused concern for its tremendous consumption of water and power. But scientists are also experimenting with ways that AI can help people and businesses use energy more efficiently and pollute less.
Data centers needed to fuel AI accounted for about 1.5% of the world’s electricity consumption last year, and those facilities’ energy consumption is predicted to more than double by 2030, according to the International Energy Agency. That increase could lead to burning more fossil fuels such as coal and gas, which release greenhouse gases that contribute to warming temperatures, sea level rise and extreme weather.
But when AI’s computing power is used to analyze energy usage and pollution, it can also make buildings more efficient, charge devices at optimal times, make oil and gas production less polluting and schedule traffic lights to reduce vehicle emissions.
Experts say that if uses like these continue to grow, they could help offset the energy consumed by AI.
“I am pretty optimistic that while more and more AI use is going to continue to increase,” said Alexis Abramson, dean of the Columbia University Climate School, “we’re going to see our ability to process be much more efficient and as a result, the energy consumption won’t go up as much as some are predicting.”
Building efficiency: Maintenance, cooling
AI can be used to make buildings more energy-efficient by automatically adjusting lighting, ventilation, heating and cooling based on weather data, electricity usage and other factors, said Bob French, chief evangelist at the building automation company 75F. Around one-third of U.S. greenhouse gas pollution comes from homes and buildings.
Letting AI schedule air conditioning and heating around workers’ arrivals and departures can be more efficient than manually adjusting the thermostat. Otherwise, a worker’s instinct might be to blast the air to quickly adjust the temperature. Automated thermostats can be particularly useful for smaller buildings where it’s not cost-effective to overhaul the entire heating and cooling system.
For building ventilation, automation can balance the intake of outside air against how much heating or cooling is needed to maintain indoor temperatures.
AI can also monitor the maintenance needs of HVAC systems and other equipment to predict and detect failures before they lead to costlier repairs.
Combined, these automations can reduce a building’s energy consumption by between 10% and 30%, experts said.
“That’s literally a super low-hanging fruit,” said Zoltan Nagy, professor of building services at Eindhoven University of Technology.
Finding energy- and cost-efficient times for EV charging
AI can schedule the most efficient charging of electric vehicles and other devices such as smartphones.
This means setting a schedule for when it is best to draw power from the grid, such as overnight, when demand and rates are lower so it’s less likely to make the grid burn more fossil fuels.
“Let’s say it’s a peak period when everybody’s got their air conditioning on, and I walk in my house and I plug in my car and I have it set up such that my car doesn’t start charging right away because it’s peak period time,” Abramson said.
The charging status screen is shown on a Tesla car at a home in San Francisco, Aug. 25, 2022. (AP Photo/Jeff Chiu, File)
In California, a pilot program shifted charging to times where there was more renewable energy available, and saved customers money.
AI can also help optimize how homeowners with solar panels store excess energy in batteries.
Reducing methane flaring from oil and gas operations
Boston-based Geminus AI uses deep learning and advanced reasoning to help oil and gas companies reduce methane flaring and venting, and reduce the amount of energy they use in extracting and refining.
Reducing methane emissions is among the fastest pathways to avoid the worst impacts of climate change, according to the United Nations Environment Programme. Methane is a powerful greenhouse gas responsible for about 30% of today’s global warming.
A motorist drives past the CHS oil refinery Sept. 28, 2024, in McPherson, Kan. (AP Photo/Charlie Riedel, File)
When pressure in oil and gas pipes builds up, some of the gas is released and burned to relieve the pressure, harming the planet and wasting money.
Geminus CEO Greg Fallon said they can monitor the network of wells and pipes and use AI-driven simulations to suggest changes to compressor and pump settings that eliminate the need for venting and flaring. Geminus does this in seconds. Traditionally it takes engineers about 36 hours to run simulations that make similar recommendations, Fallon added.
“As we scale this across the industry, there’s a massive opportunity to reduce greenhouse gas emissions,” Fallon said.
Finding geothermal hot spots
Salt Lake City-based geothermal energy startup Zanskar has built AI models to understand the Earth’s subsurface. It’s using that modeling to find overlooked geothermal hot spots and target drilling.
Geothermal creates electricity cleanly by making steam from the Earth’s natural heat and using it to spin a turbine. It’s one renewable energy the Trump administration favors.
Zanskar co-founders Carl Hoiland and Joel Edwards say they simulate and assess a huge number of possible subsurface scenarios to estimate where there are pockets of very hot water. From this, they pick optimal locations and drilling directions.
“AI is becoming the solution to its own energy problem,” Hoiland, the CEO, said. “It’s showing us a way to unlock resources that weren’t possible without it.”
Last year, Zanskar purchased an underperforming geothermal power plant in New Mexico. Their AI modeling successfully indicated there was an untapped geothermal reservoir that could repower the facility.
Next, Hoiland and Edwards focused on another site in Nevada, despite industry experts telling them it was too cold to support a utility-scale power plant. They drilled and announced their second geothermal discovery in September at that site.
Reducing traffic emissions
Google is using artificial intelligence and Google Maps data to identify traffic light adjustments that can reduce stop-and-go traffic to lower pollution. Passenger cars and small trucks account for about 16% of U.S. greenhouse gas emissions, according to Environmental Protection Agency data.
Launched in 2023, Project Green Light is now in 20 cities on four continents. The most recent is Boston, which has notoriously bad traffic.
Vehicles drive along a highway July 30, 2025, in Cincinnati. (AP Photo/Joshua A. Bickel, File)
Each city gets AI-generated recommendations. City engineers determine which to implement. Google says Project Green Light can reduce stop-and-go traffic by up to 30%, which cuts emissions by 10% and improves air quality.
“We’re just scratching the surface of what AI can do,” said Juliet Rothenberg, Google’s product director of Earth and resilience AI.
___
Read more of AP’s climate coverage.
___
The Associated Press’ climate and environmental coverage receives financial support from multiple private foundations. AP is solely responsible for all content. Find AP’s standards for working with philanthropies, a list of supporters and funded coverage areas at AP.org.
