Hyperscalers Could Face Backlash from Relying on Natural Gas if Future Projections Hold True

After investing heavily in renewable energy sources, tech giants such as Amazon, Google, Meta, and Microsoft are now turning to natural gas to fulfill the energy needs of their data centers that support their ambitious AI projects. However, a recent study indicates that they may face regret over this reliance on fossil fuels.

According to Noreva, an energy research firm, natural gas prices could potentially triple in various regions of the U.S. in the next few years. This surge is expected as an increase in hyperscaler demand clashes with a decline in supply growth and an uptick in liquefied natural gas exports. The hyperscalers may not be ready for the potential financial shocks that await them.

“The energy market appears to have lulled itself into believing that gas prices won’t rise,” said Peter Gardett, CEO of Noreva, commenting on the situation. “Basic calculations reveal that the gas market is much tighter now than it was a few years ago.”

Aggressive Investments by Hyperscalers

The low cost of gas has led these companies to claim substantial portions of the market. For instance, Meta recently announced plans for a 7.5-gigawatt natural gas plant in Louisiana to support its Hyperion data center. Shortly thereafter, Microsoft and Google each declared intentions to construct their own gas plants, each with a capacity of over one gigawatt, also in Texas. Not to be excluded, Amazon has plans for a 7.6-gigawatt facility in the same state.

These companies, which have typically avoided significant capital investments, are now rapidly expanding their physical infrastructure while also diving into unfamiliar energy markets.

According to Gardett, one investor expressed surprise at the level of natural gas price risk that these hyperscalers are willing to assume. “They’re taking on risks that are not typical for a buyer in this space,” he mentioned.

Noreva projects that natural gas prices could rise to over $10 per million BTUs at specific delivery points, in contrast to current prices that hover between $2 and $4.50 per million BTUs; for example, the prominent Henry Hub in Louisiana is currently priced just under $3.

Given that fuel costs account for roughly half of a large power plant’s operational expenses, a significant increase in natural gas prices could drastically elevate the operational costs of AI data centers. This could lead to higher token prices or prompt hyperscalers to rely more on the electricity grid, in turn increasing costs for consumers.

While natural gas prices seem stable for now with futures contracts showing a lack of significant anticipated changes, Gardett is skeptical about whether this will hold true in the long run.

Rising Demand Dynamics

Natural gas prices have remained consistent due to a combination of steady demand and the gradual addition of new supplies, which has balanced out declining production from older wells. Gardett notes that while energy companies will continue to introduce new supplies, the pace is not what it used to be, and drilling new wells is becoming increasingly costly.

“This alone wouldn’t shift the current economic balance. What’s causing changes is the connection of the domestic gas market to the global one,” he explained. “The demand generated by AI applications is also a significant factor.”

Hyperscalers have been drawn to regions like Texas and Louisiana because of their low natural gas prices. However, in West Texas, the focus has primarily been on oil production, where the natural gas produced has often been considered a secondary product. Until recently, there were limited pipelines to transport this gas, resulting in discounts for those who sought to use it. This situation is beginning to change.

“New pipelines have been constructed, enabling this gas to reach export markets,” Gardett noted.

With improved connectivity between West Texas and national as well as international natural gas markets, local demand will begin to impact prices elsewhere. Even minor fluctuations in price near the hyperscalers’ data centers could have widespread repercussions.

“There will be instances where one area has an abundance of gas while another does not, leading to significant price disparities,” Gardett pointed out. Those disparities may drive prices in certain regions above $10 per million BTUs for extended durations.

In such scenarios, even if hyperscalers can tolerate increased costs, their natural gas usage may trigger a backlash against data centers. Currently, 80% of consumers are concerned about the effects of data center operations on their utility bills, primarily related to electricity, and this concern could extend to natural gas as well.

As hyperscalers rush to power their AI data centers, they are becoming more entrenched in the fossil fuel sector, an area where they lack significant experience, but one that could have far-reaching impacts on their operations.

“In future earnings calls, you might hear companies like Alphabet discuss the link between natural gas prices and their financial results, which is quite an unusual scenario,” Gardett concluded.