This week in Houston, natural gas became the main topic at CERAWeek—one of the biggest energy events in the world. Around 10,000 experts and industry leaders gathered to talk about one big question: how to power the fast growth of artificial intelligence.
A big part of the discussion focused on energy. AI systems, especially data centers, need huge amounts of electricity to run. And right now, many experts believe natural gas is one of the fastest ways to meet that demand.
Laurent Ruseckas from S&P Global said that gas-powered electricity is “critical” for AI development. Many sessions at the conference looked at how the gas industry can support AI’s growing needs, and also how AI can help improve gas production.
The problem is simple: data centers use a lot of power, and much of that power still comes from sources that produce pollution.
According to the International Energy Agency, natural gas is already the third-largest energy source for data centers, covering about 26% of their needs. Coal is still number one, followed by renewable energy.
In the United States, gas production has grown quickly over the years. Data from the U.S. Energy Information Administration shows that liquefied natural gas (LNG) exports have increased massively since 2016. With Donald Trump supporting fossil fuels, experts expect this trend to continue.
Some analysts say the rise in gas is directly linked to the growth of AI. Simply put, as AI expands, it needs more power—and gas is currently one of the easiest ways to supply it.
Industry groups like the Center for Liquefied Natural Gas also support this view. They argue that gas is reliable and can deliver power whenever it’s needed, which is important for data centers that must run 24/7.
In fact, more than one-third of U.S. gas capacity is already being used to power data centers, according to recent research.
But not everyone agrees that this is a good long-term solution.
Some experts warn that relying too much on gas could create problems. Mark Brownstein from the Environmental Defense Fund said that while gas will play a role, it may not be as big as many expect. He believes this approach is “not sustainable” because it is expensive and causes significant pollution.
Natural gas may burn cleaner than coal, but it still produces carbon dioxide. Even worse, methane leaks from gas systems can have a strong impact on global warming.
These concerns are already causing pushback. For example, in West Virginia, a project to build a gas-powered plant for a data center is facing opposition from local residents worried about health and environmental risks.
At the same time, big tech companies are under pressure. Many had promised to reach carbon neutrality by 2030 or 2040, but the rapid growth of AI is making that harder. Some experts now say those goals are being pushed aside because companies need quick and reliable energy.
Another option being discussed is nuclear power. It already provides about 15% of the electricity used by data centers worldwide. Officials from the U.S. Nuclear Regulatory Commission say nuclear is becoming an important part of the solution.
The challenge, however, is time. Nuclear plants take years to build, much longer than gas facilities.
Still, companies are moving in that direction. Meta recently signed deals with nuclear energy providers in the U.S., aiming to secure long-term power. But those projects are not expected to be fully ready until around 2035.
In the end, the message from the conference was clear: AI is growing fast, and the world needs to find enough energy to support it. Right now, natural gas is leading the way—but the debate over its long-term impact is far from over.
