The race to deploy artificial intelligence is complicating tech companies’ commitments to reduce greenhouse gas emissions, most of which come from the burning of gas, oil and coal and drive climate change. They say they must be flexible as they rush to build sprawling data centers that can consume more power than entire cities.

“Even if they haven’t officially revised their goals, they are starting to acknowledge that, ‘Yeah, we’re maybe not on track,’” said Patrick Huang, a senior analyst at Wood Mackenzie.

Now, Huang said, the companies must use whatever kinds of power they can to stay competitive — and increasingly that is natural gas, which is mostly methane, a planet-warming greenhouse gas.

Tech companies bought record amounts of clean energy in 2024 and 2025, according to the Clean Energy Buyers Association.

But total emissions have gone up over roughly the first five years of their climate commitments, according to companies’ sustainability reports. Google’s emissions jumped nearly 50%. Amazon’s rose by 33%, Microsoft’s more than 23% and Meta’s more than 60%.

Data centers used about 4.6% of total U.S. electricity in 2024, a share that could nearly triple by 2028, according to government estimates. Some analysts predict nationwide electricity use to rise as much as 20% in the next decade, with data centers a big reason.

Meanwhile, a backlog of proposed projects awaiting permission to connect to power grids and efforts by the Trump administration to sideline renewable energy may affect tech companies’ climate goals — and prolong reliance on fossil fuels, experts said.

“Each of these alone could be real challenges,” said Julie McNamara, associate policy director at Union of Concerned Scientists’ Climate & Energy program. “Together, it’s just creating a real near-term crunch on the system.”

Natural gas use spikes as AI soars

Tech companies say they’ve made significant progress on emissions through energy-efficiency measures, buying renewable energy credits and power from sources that don’t emit greenhouse gases and requiring suppliers to reduce their own emissions.

Yet natural gas in 2024 accounted for more than 40% of electricity powering U.S. data centers, while coal supplied 30% globally, the International Energy Agency said. And the trend doesn’t appear to be slowing. Utilities are planning natural gas plants around the country to help supply data centers, while some tech companies plan on-site gas plants built only to feed a data center.

“Companies are scrambling to try to get as much power as they can as quickly as possible,” said Lori Bird, director of the U.S. Energy Program at the World Resources Institute. “It’s a mad rush and a lot of competition for resources.”

Microsoft President Brad Smith told The Associated Press that he is “confident in our ability” to meet the company’s 2030 goal to remove more carbon dioxide from the atmosphere than it emits by investing in new sources of carbon-free energy, including nuclear, solar and hydropower.

In Wisconsin, for example, two new natural gas plants to help power a Microsoft data center will be offset by investment in solar elsewhere in the state. Similarly, three natural gas plants will provide electricity to a massive Meta data center in rural Louisiana, while the company invests in solar elsewhere.

Google says it’s investing in wind, hydropower, battery storage and advanced nuclear, though it also relies on natural gas. The company plans to buy electricity from a natural gas plant to be built at the Archer Daniels Midland corn processing plant in Decatur, Illinois, where carbon dioxide emissions would be captured and stored underground.

To help meet clean energy goals, tech companies count on such power purchase agreements and buying renewable energy certificates, a tradeable commodity that supports new and existing sources. But that could get more difficult under proposed changes to how greenhouse gases are reported, which would require that sources are in the same region as a company’s data center and match hours of operation — for example, solar credits could only be applied to daytime operating hours.

Although some new gas plants will replace dirtier coal plants, it takes about 30 years to recover the investment. That means delaying the overall transition to clean and renewable energy at a time when the United Nations Environment Programme warns that high-emitting countries are unlikely to meet their own targets for reducing greenhouse gas emissions. AI is blamed in part for a 2.4% uptick in U.S. fossil fuel emissions last year, according to a study by the Rhodium Group, an independent research firm.

And though other sectors of the economy also are electrifying, “it is only because of these data centers that these gas plants are being built,” McNamara said. “There are no two ways about it.”

Trump war on renewables complicates tech goals

Getting enough electricity was challenging even before President Donald Trump took office last year and took aim at renewable energy.

He’s canceled grants and permits for solar and wind projects and tax breaks for renewable energy, which advocates say can be built less expensively and more quickly than natural gas or nuclear plants, while ordering that several coal-fired power plants slated for retirement keep running.

Many companies set goals expecting federal tax credits would support wind and solar deployment, said Rich Powell, chief executive officer of the Clean Energy Buyers Association.

But those credits will end in July, after being eliminated by the Republican-controlled Congress and Trump.

Trump, who has called climate change a “hoax,” has argued that green energy is unreliable and expensive and could harm national energy independence.

Powell said his association has “been very, very clear with this Congress and this administration that all technology should be on a level playing field and that we’re putting both energy affordability and energy reliability at risk if we don’t do that.”

Josh Parker, sustainability chief for chipmaker Nvidia, said AI eventually will reduce electricity use because it’s more efficient than traditional computing. He said curtailing energy development could cause the U.S. to fall behind on AI.

“Our perspective is that we need an all-of-the-above approach to energy,” he said.

Tech companies would have been hard-pressed in 2020, when many set goals, to project current energy needs because much of the technology and equipment used to train machine-learning models — which use most data-center electricity — were just being introduced, said Jay Dietrich, who researches AI sustainability for the Uptime Institute and formerly led emissions goal-setting at IBM.

By 2023, he said, tech companies “had a pretty good idea things were going to get a lot more exciting … and that the numbers were going to grow quickly.”

He expects many will extend the timeline for emissions goals, based on a 2025 Uptime Institute survey that saw a 12% drop in the number of operators saying they’d meet a market-based 2030 carbon-neutral goal. However, even with increasing emissions, the largest companies should be able to afford enough renewable energy and offsets to meet carbon-neutral goals.

McNamara said the surge in electricity demand from data centers turned a challenge into “an outright crisis.”

“Tech companies are allowing implicitly or explicitly an enormous increase in fossil fuel dependence under their watch and because of their actions,” she said.

___

Associated Press writer Matt O’Brien contributed to this report.

___

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.