Hyperscalers May Reevaluate Natural Gas Usage in Light of Recent Forecasts
Following significant investments in renewable energy, leading tech giants like Amazon, Google, Meta, and Microsoft are increasingly relying on natural gas to fuel their data centers that underpin extensive AI projects. However, a recently published research report raises concerns over the potential detrimental effects of this growing dependency on fossil fuels for these companies.
Noreva, an energy research firm, reports that natural gas prices in various U.S. regions could potentially triple in the coming years. This surge is likely due to increased demand from hyperscalers, stagnant supply growth, and soaring liquefied natural gas exports, which could expose hyperscalers to sudden price fluctuations.
“I believe many within the energy sector have been misled regarding future gas prices,” stated Peter Gardett, CEO of Noreva, in an interview with TechCrunch. “A straightforward calculation indicates that the gas market is much tighter than it was a few years ago.”
Major Investments by Hyperscalers
The low cost of natural gas is encouraging hyperscalers to secure substantial market positions. In March, Meta announced plans to develop a significant 7.5-gigawatt natural gas facility in Louisiana to support its Hyperion data center. Shortly after, Microsoft and Google unveiled plans for gigawatt-scale gas plants in Texas. Additionally, Amazon revealed intentions for a 7.6-gigawatt gas plant in Texas.
These companies, usually prudent about substantial capital expenditures, find themselves compelled to invest heavily in physical infrastructure while navigating complex energy markets.
According to Gardett, at least one investor was taken aback by the level of risk regarding natural gas prices that hyperscalers are willing to tolerate. “They are behaving in ways that are atypical for an off-taker,” he noted.
Noreva predicts that natural gas prices could exceed $10 per million BTUs at specific delivery points, a stark contrast to current prices ranging from $2 to $4.50 per million BTUs, with the well-known Henry Hub in Louisiana priced just below $3.
As fuel costs account for nearly half of the electricity expenses for large power plants, a doubling or tripling of natural gas prices could significantly raise operational costs for AI data centers that “bring their own power.” This scenario could lead to higher token prices or compel hyperscalers to connect to the grid, further increasing their electricity costs.
In the near term, natural gas prices appear stable, as futures contracts do not indicate significant volatility. “It’s a reasonable bet,” remarked Gardett, though he remains skeptical about its accuracy.
Rising Demand
Natural gas prices have largely remained stable due to years of consistent demand and limited supply growth, which mitigates declining production from aging wells, according to Gardett’s analysis. He believes that energy companies will manage to boost supplies, albeit not to former levels, and that drilling new wells is becoming increasingly expensive.
“This alone wouldn’t drastically change the economics. The crucial factor is that we are finally connecting the domestic gas market with the global market,” he elaborated. “Moreover, there’s an uptick in demand driven by AI.”
Hyperscalers have been drawn to Texas and Louisiana due to lower natural gas prices. For instance, in West Texas, many wells focus on oil extraction, producing natural gas as a byproduct that has limited access to the market. Traditionally, inadequate pipeline infrastructure hindered the transportation of this gas, forcing producers to sell it at reduced prices. However, significant enhancements are in progress.
“Recently constructed pipelines are largely aimed at export markets,” clarified Gardett.
As West Texas integrates better with both national and international natural gas markets, local demand will start to influence prices elsewhere and vice versa. Even minor price shifts near hyperscaler data centers could have far-reaching implications in other regions.
“There will be areas with abundant gas and others where it is scarce, leading to significant price differences,” Gardett noted. Such discrepancies could push prices in certain locations above $10 per million BTUs for extended periods.
In this environment, even if hyperscalers can manage rising costs, their escalating consumption of natural gas could provoke increasing public opposition to data centers. Currently, 80% of consumers are concerned about the impact of data centers on their utility bills, particularly regarding electricity costs. This concern may extend to natural gas expenses as well.
By rapidly powering their AI data centers, hyperscalers are becoming more intertwined with the fossil fuel sector—a domain in which they have relatively limited expertise, yet one that could heavily impact their operations.
“In future earnings calls for Alphabet, you may hear discussions about the correlation between natural gas prices and Google’s performance, which is unusual but indicative of the reality we face today,” Gardett concluded.
When you click on links in our articles, we may earn a small commission. This does not affect our editorial independence.
