Major cloud computing companies betting on natural gas to power AI data centers could face a sharp increase in energy costs in the coming years. According to a new report from energy research firm Noreva, natural gas prices could exceed $10 per million British thermal units at some delivery hubs in the United States, compared with a current range of approximately $2 to $4.50, while the Henry Hub price in Louisiana is just under $3.
The warning is based on the convergence of three factors: growing gas demand from data centers, slowing growth in domestic supply, and rising liquefied natural gas exports. Peter Gardett, Noreva’s chief executive, believes energy markets have become accustomed to assuming that gas prices cannot rise very far, but simple calculations could point to a much tighter market than existed a few years ago.
Huge bets on dedicated power plants
After years of purchasing wind and solar power projects, companies such as Amazon, Google, Meta, and Microsoft have begun securing large quantities of natural gas to meet the needs of data centers linked to their AI ambitions.
In March, Meta said it would build a massive 7.5-gigawatt natural gas power plant in Louisiana to supply power to the Hyperion data center. Days later, Microsoft and Google announced plans to build gas plants with capacity of at least one gigawatt each in Texas. Amazon also plans to build a 7.6-gigawatt gas plant in the same state.
These plans mean that companies that have historically avoided major capital expenditures are investing directly in physical assets and energy markets, areas in which they do not have the same expertise they possess in software and cloud services. Gardett said one investor he spoke with expressed surprise at the scale of the gas-price risks that data center companies are willing to accept, noting that they are taking steps unusual for a large energy consumer.
How could gas prices be passed on to the cost of AI?
Fuel accounts for approximately half the cost of electricity produced by a large power plant. Therefore, a doubling or tripling of gas prices could make it far more expensive to operate AI data centers that generate their own power. The impact could appear in the form of higher costs for the tokens used in AI services, or it could prompt companies to reconnect to the electric grid, potentially increasing pressure on electricity prices.
Futures contracts currently do not anticipate major price changes in the foreseeable future, which makes betting on gas economically plausible, according to Gardett. However, he questions whether these expectations can reflect all the changes underway in the energy market.
Domestic demand connects to the global market
Gas prices have remained stable in recent years thanks to relatively steady demand and the addition of new supplies that offset declining production from older wells. Gardett expects energy companies to be able to add new volumes, but not at the previous pace, while drilling new wells has also become more expensive.
The most important change, according to Noreva’s analysis, is the connection between the domestic gas market and the global market, coinciding with the wave of demand driven by AI. Low gas prices have attracted data center companies to Texas and Louisiana, particularly West Texas, where most wells have focused on oil and the resulting natural gas was a byproduct that lacked a broad market.
Limited pipeline capacity in the region also forced producers to sell gas at a discount to anyone able to use it. But the construction of new pipelines changed this situation, with a large share of the gas now heading toward export markets. As West Texas becomes connected to national and international gas networks, demand in the region can affect prices elsewhere, and vice versa.
Regional risks and growing social pressure
Gardett believes price differences between regions could become a key factor in driving prices above $10 per million British thermal units for extended periods. Areas with abundant gas could sit alongside areas facing shortages, resulting in large price disparities even if the market average remains more stable.
Even if major cloud computing companies are able to absorb these increases, their intensive gas consumption could add a new dimension to objections against data center expansion. Already, 80% of consumers are concerned about the impact of data centers on utility bills, with most of this concern focused on electricity. The same dissatisfaction could extend to gas bills if the new demand affects prices.
The report concludes that companies such as Amazon, Google, Meta, and Microsoft have, in their rapid effort to secure power for AI data centers, become more closely tied to the fossil-fuel market. If Noreva’s forecasts materialize, gas prices could shift from an operational detail to a factor influencing the costs of AI services and the business performance of major technology companies.