Hyperscalers spent years buying into wind and solar projects. Now, as AI data centers demand huge amounts of power, some of the biggest technology companies are moving deeper into natural gas.
That shift could prove costly. Noreva, an energy research firm, says natural gas prices could triple in some parts of the U.S. in the coming years as data center demand meets slower supply growth and rising exports of liquefied natural gas.
Big AI plans are pulling tech deeper into gas
Amazon, Google, Meta, and Microsoft are all pursuing more power for data centers that support their AI ambitions. The new emphasis on natural gas marks a change for companies that have often tried to frame their energy strategies around renewable power.
Several large projects show how quickly the strategy is changing. In March, Meta said it would build a massive 7.5-gigawatt natural gas power plant in Louisiana to power its Hyperion data center. A few days later, Microsoft and Google each said they would build gigawatt-scale gas power plants in Texas. Amazon plans to build a 7.6-gigawatt gas power plant in Texas.
These moves are not just about buying electricity. They place technology companies into the physical infrastructure business at a much larger scale, and into energy markets where prices can move in ways that are difficult to predict.
Peter Gardett, CEO of Noreva, told TechCrunch that at least one investor he spoke with was “surprised” by how much natural gas price risk hyperscalers are accepting. He added: “They’re doing things that are not normal for an off-taker to do.”
The price risk is easy to overlook
Natural gas has looked inexpensive and relatively stable for years. Today, prices range from about $2 to $4.50 per million BTUs, while the widely traded Henry Hub in Louisiana is priced at just under $3.
Noreva’s forecast points to a very different possibility. The firm expects natural gas prices to rise above $10 per million BTUs in certain hubs, the delivery points used for futures contracts.
That matters because fuel represents about half the cost of electricity from a large power plant. If natural gas prices double or triple, “bring your own power” AI data centers become more expensive to operate. Those higher costs could show up in token costs, or hyperscalers could decide to connect more demand to the grid, which could push electricity prices higher.
For now, futures contracts are not pricing in major changes. Gardett called the current strategy “not an unreasonable bet,” but he also said he is not convinced the market is right.
Demand, supply, and exports are changing the market
Natural gas prices have stayed stable because demand has been relatively flat and new supplies have continued to offset declining production at older wells. Gardett expects energy companies to add more supply, but not as quickly as before. New wells are also becoming more expensive.
Two forces are changing the picture. One is the increasing connection between the domestic gas market and the global gas market. The other is AI demand from large data centers.
Texas and Louisiana have attracted hyperscalers partly because natural gas has been cheap there. West Texas has been especially notable because many wells have focused on oil, with natural gas produced as a byproduct. With limited pipeline capacity to move that gas away, producers sold it at a discount to buyers that could use it nearby.
That local discount is under pressure. Gardett said pipelines have been built in the region, and much of that gas is moving toward export markets.
As West Texas becomes more connected to national and international gas markets, local prices may no longer behave as separately as they once did. Demand near major AI data centers could affect prices elsewhere, while broader market movements could also reach back into places that hyperscalers chose for cheap gas.
The backlash could move beyond electricity bills
The data center debate is already tied to household utility costs. According to the source article, 80% of consumers are worried about data centers’ impact on their utility bills, mostly because of electricity.
If natural gas prices rise in regions near major AI infrastructure, that concern could expand. Data centers would not just be part of the electricity conversation. They could also become part of the natural gas bill conversation.
That would add another layer of scrutiny for companies that are already facing questions about the physical footprint of AI. The more hyperscalers depend on natural gas, the more their businesses may be exposed to fuel markets they do not fully control.
Gardett described the possible shift in stark terms: “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.”
The core issue is not whether natural gas can power AI data centers. It can. The risk is whether the current assumption of cheap, available gas holds as demand grows, exports rise, and supply becomes harder to expand at the same pace.