Noreva Flags Natural Gas Price Risk for AI Data Centers
Energy research firm Noreva has warned that natural gas prices could rise above $10 per million BTUs at some US trading hubs as hyperscalers expand gas-fired generation for AI data centers. Current prices range from roughly $2 to $4.50 per million BTUs, while the widely traded Henry Hub in Louisiana is priced at just under $3.
Amazon, Google, Meta and Microsoft have historically procured wind and solar projects, but are now making major commitments to gas generation. Meta said in March that it would build a 7.5-gigawatt natural gas plant in Louisiana for its Hyperion data center. Microsoft and Google each announced gigawatt-scale gas plants in Texas, and Amazon plans a 7.6-gigawatt gas plant in the state.
AI power demand meets a changing gas market
Noreva CEO Peter Gardett argues that the assumption of permanently cheap gas may no longer hold. He expects producers to add supply, but not at the pace seen in the past, while new wells are becoming more expensive. His assessment also centres on growing links between the domestic market and international LNG demand, alongside the additional pull from AI infrastructure.
West Texas has attracted data-center power projects partly because associated gas from oil wells was sold at a discount when pipeline capacity was limited. Gardett said new pipelines are changing that equation by moving more gas towards export markets. As the region becomes more connected to national and global markets, local supply conditions and demand elsewhere can affect prices more directly.
Exposure reaches beyond power procurement
Noreva expects regional price differences to become important, with some locations potentially seeing prices above $10 per million BTUs for extended periods. Futures markets do not currently anticipate large changes, and Gardett described the hyperscalers’ view as not unreasonable, but he questioned whether it would prove correct.
Fuel accounts for about half of the electricity cost from a large power plant. A doubling or tripling of gas prices could therefore materially raise the cost of operating data centers built around dedicated generation. Companies could pass some costs through to AI token pricing or seek more grid connections, which could increase pressure on electricity prices.
The energy commitments also deepen the commercial ties between cloud and AI providers. Amazon’s AWS relationship with Anthropic, reflected in Anthropic’s $5 billion Amazon investment and AWS spending, illustrates how large-scale AI expansion increasingly depends on physical infrastructure decisions as well as model development and cloud capacity.
Business implication
Businesses buying AI capacity or planning dedicated data-center power should assess gas-price exposure alongside compute availability, grid access and operating-cost assumptions, because fuel-market volatility could materially change the economics of gas-backed AI infrastructure.