After years of buying renewable projects, major hyperscalers — Amazon, Google, Meta and Microsoft — are increasingly planning to run large AI data centers on natural‑gas power. A new report from energy research firm Noreva warns that this shift could leave them exposed to much higher regional gas prices.
What the report finds
Noreva says that in parts of the United States, natural gas prices could triple as hyperscaler demand collides with slower supply growth and rising liquefied natural gas (LNG) exports. Peter Gardett, chief executive of Noreva, told TechCrunch that many market participants have been ‘‘lulled into a sense that gas prices can’t go up.’’
The firm projects prices above $10 per million British thermal units (MMBtu) in some delivery hubs. Current wide ranges are about $2 to $4.50 per MMBtu, with the widely traded Henry Hub in Louisiana priced just under $3 per MMBtu.
Large bets by cloud providers
Cheap regional gas has encouraged hyperscalers to lock in on‑site fuel supplies. Recent announcements include:
- Meta’s plan to build a 7.5‑gigawatt natural‑gas power plant in Louisiana to feed its Hyperion data center.
- Microsoft and Google each announcing plans for gigawatt‑scale gas power plants in Texas.
- Amazon’s plan for a 7.6‑gigawatt gas power plant in Texas.
These investments mark a shift for companies that historically avoided large physical capital expenditures, pushing them deeper into energy markets where they have less operational experience.
Why higher gas prices matter
Fuel represents about half the cost of electricity from a large power plant, so a doubling or tripling of natural gas prices would materially raise operating costs for data centers that ‘‘bring your own power.’’ Higher fuel costs could increase prices for AI compute (for example, token costs), or force hyperscalers to rely more on the grid — which could in turn push up electricity prices for other customers.
Although current futures markets are not pricing in dramatic changes, Gardett cautions that futures aren’t always a reliable predictor and that linking the domestic gas market to global markets, plus the new demand from AI, change the fundamentals.
Supply, pipelines and exports
Stable gas prices until now were supported by relatively flat demand and steady additions to supply that offset declines at older wells. Gardett expects more supply additions, but not at previous rates, and notes new wells are becoming more expensive.
A key change is that previously isolated gas‑rich regions such as West Texas are becoming better connected by pipelines, directing more gas toward export markets. Historically, much of the gas there was a byproduct of oil wells and was sold at a discount because there was insufficient pipeline capacity. As pipelines have been built, that gas can reach broader domestic and international markets, tightening regional balances.
As regional markets interconnect, local demand near large data centers can have outsized effects on prices elsewhere. Gardett warns that local differentials will emerge: areas with abundant gas close to areas with little supply can create pronounced price gaps, and in some hubs those gaps could push prices above $10 per MMBtu for extended periods.
Corporate and consumer implications
Even if hyperscalers can absorb higher fuel costs, their increased gas consumption could exacerbate public concern about data centers’ impact on utility bills. Today, roughly 80% of consumers are worried about data centers’ effects on household utility costs, mainly electricity — and that anxiety could extend to natural gas bills if prices rise.
By moving quickly to secure on‑site gas power, hyperscalers are embedding themselves in the fossil fuel sector — a space where they have limited experience but where price moves could soon affect their business results and the cost of services.
Conclusion
Noreva’s analysis highlights a trade‑off: on‑site gas power provides control for massive AI compute, but it also introduces exposure to volatile regional gas markets increasingly linked to global LNG flows. If Noreva’s scenario materializes, hyperscalers could face far higher operating costs, with knock‑on effects for electricity markets, consumers and corporate earnings.
