Those investments make future gas prices increasingly relevant to the cost of running AI workloads. Fuel accounts for about half of the electricity cost at a large power plant, according to the reporting. If prices double or triple, data centers supplying their own gas-fired power could face substantially higher operating expenses than originally expected.
Noreva CEO Peter Gardett told TechCrunch that hyperscalers are accepting an unusual degree of exposure to those fluctuations. “They're doing things that are not normal for an off-taker to do,” Gardett said.
Current markets do not indicate that such a price surge is imminent, and Gardett acknowledged that existing data can support the companies' decisions. Noreva's forecast instead focuses on changes that could tighten supply and produce sharp regional differences over time.
One factor is the increasing cost of bringing new gas wells online as production growth slows. Another is rising demand from liquefied natural gas exports. AI data centers would add another source of consumption as technology companies seek large amounts of dependable electricity for their computing infrastructure.
West Texas illustrates how those forces could interact. Natural gas produced alongside oil has historically been available locally at steep discounts because producers had limited ways to move it elsewhere. That cheap supply helped make the region attractive for power-intensive data centers.
New pipeline infrastructure is changing that equation by providing routes from West Texas toward export terminals. As previously constrained supplies gain access to other buyers, Noreva expects regional gas markets to become more exposed to competing demand.
The result does not necessarily mean gas prices would rise uniformly across the country. Noreva instead expects pronounced differences between individual markets, with some locations potentially experiencing sustained prices above $10 per million BTUs. “You will get places where you get a lot of gas next to someplace where there's none, and so you'll get those big differentials,” Gardett said.
For hyperscalers, that volatility could affect more than the cost of running individual power plants. Higher fuel expenses could increase the cost of supplying AI compute or encourage operators to draw more electricity from the grid instead of relying entirely on dedicated generation. Either outcome would complicate the economics behind infrastructure investments designed to operate for years.
The issue could also extend to consumers. According to figures cited by Noreva, about 80% of consumers are already concerned about the effect data centers could have on utility bills. Additional demand for natural gas could broaden those concerns beyond electricity prices if it begins affecting fuel used for residential heating.
The companies have ways to reduce some of their exposure, including long-term hedging, fuel storage and plants capable of using more than one fuel. Nuclear and geothermal generation could provide additional alternatives over longer periods. The immediate challenge is that large gas projects are moving ahead while their owners cannot fully control the future price of the fuel they require.
That creates a direct connection between commodity markets and AI infrastructure costs that has previously received less attention than spending on chips, servers and data centers. As hyperscalers expand their own power generation, movements in natural gas prices could increasingly affect the economics of the AI services running on top of that infrastructure.
Gardett expects that connection to become visible even in how technology companies discuss their financial performance. “On future Alphabet earning calls, you will hear them talk about the correlation between natural gas pricing and Google results, which is strange, but that's where we are.”
This analysis is based on reporting from the tech buzz.
Image courtesy of BOP Products.
This article was generated with AI assistance and reviewed for accuracy and quality.
