The expansion of artificial intelligence is no longer just a race to develop more powerful models; it’s also a race to secure the massive amounts of electricity needed to run data centers. As some power grids struggle to keep up with this accelerating demand, natural gas is emerging as a key solution for providing the energy these facilities need.

With the rapid expansion of AI data centers, natural gas is emerging as a solution to provide the electricity needed to run them, at a time when some power grids are struggling to keep up with the accelerating pace of building these facilities.

According to a report published by The Washington Post, natural gas has shifted from a temporary solution to overcome delays in connecting data centers to a choice that could last for decades, with long-term fuel contracts and the creation of dedicated power plants and pipelines being signed.

Why does AI consume so much electricity?

AI systems require massive data centers housing thousands of specialized processors, particularly graphics processing units (GPUs) used to train and run models. But the problem doesn’t stop at powering the chips; these devices produce a lot of heat, making cooling systems an essential part of energy consumption within data centers.

As the number of users increases and AI applications expand, companies need to build new data centers quickly, and each new center means an additional demand for electricity.

The International Energy Agency estimates that global data center consumption was around 415 terawatt hours in 2024, accounting for about 1.5% of global electricity consumption, but is projected to more than double by 2030 to reach approximately 945 terawatt hours, with artificial intelligence being the main driver of this growth.

The power grid does not keep pace with the speed of data center construction

Technology companies can build a new data center in a much shorter time than it takes for some electrical infrastructure projects to add transmission lines, substations, transformers, and new generation capacity.

The Washington Post notes that obtaining connection to the power grid can take 4 to 7 years in some cases, which is a long time for companies competing to build artificial intelligence infrastructure quickly.

Therefore, some companies and developers have started thinking differently: instead of waiting for the grid, why not produce electricity near the data center itself? And here natural gas comes into play.

Why gas specifically?

The main advantage of gas is that it can be used to generate electricity continuously and its production can be controlled according to demand, unlike sources such as sun and wind which depend on weather conditions.

Gas plants can also be developed and utilized in a relatively shorter period compared to some alternatives that require many years before becoming widely available.

The newspaper notes that technology companies are looking at nuclear and geothermal energy as long-term options, but some of these projects may need 5 to 10 years before they become widely available. This explains why gas has become an attractive option in the current stage.

Gas boom is not small

The U.S. Energy Information Administration expects dry natural gas production in the United States to reach 111.7 billion cubic feet per day in 2026, then rise to 115.9 billion cubic feet per day in 2027, at new record levels.

The agency also expects U.S. gas consumption to reach record highs, driven partly by increased use for electricity generation.

Meanwhile, the Energy Information Administration predicts that U.S. electricity consumption will reach a record high in 2026 and 2027, with AI-intensive data centers being one of the main drivers of this growth.

What about renewable energy?

This does not mean that artificial intelligence is entirely heading towards fossil fuels. The International Energy Agency expects renewable energy sources to meet nearly half of the increase in electricity demand associated with data centers by 2030.

But gas will remain a key player in the coming years; the agency expects natural gas to provide more than 130 terawatt hours of additional electricity annually for U.S. data centers by 2030 compared to 2024 levels.

Here is the paradox: tech companies are investing in solar, wind, nuclear and storage, but at the same time they need a source that can provide power quickly and reliably as AI infrastructure expands.

A bet that could last for decades

The problem is that what started as a temporary solution could become a permanent infrastructure. Building a large gas plant or a dedicated pipeline for a data center does not necessarily mean a short-term investment.

The Washington Post notes that fuel contracts extending 10 to 20 years, along with significant investments in generation plants and infrastructure, could make it more difficult to phase out gas later. This opens up a new battle behind the AI race: will the AI boom cause natural gas to live longer, even as tech companies pledge to shift to low-carbon energy sources?