Big Tech's gas bet could backfire badly, energy researchers warn
Amazon, Google, Meta and Microsoft are building massive natural gas power plants for their AI data centres. A new energy research report says prices could triple, and the companies may not be ready for the shock.
Key points
- Energy research firm Noreva forecasts natural gas prices could exceed $10 per million BTUs in some US regions, up from roughly $2 to $4.50 today.
- Meta announced a 7.5-gigawatt gas plant in Louisiana in March; Amazon plans a 7.6-gigawatt plant in Texas.
- Fuel accounts for about half the cost of electricity at a large power plant, so a price tripling would sharply raise the cost of running AI data centres.
- New export pipelines are connecting cheap West Texas gas to global markets, removing the local price discount that attracted hyperscalers in the first place.
- A survey cited in the research found 80% of consumers already worry about data centres pushing up their utility bills.
Amazon, Google, Meta and Microsoft, the giant cloud companies often called hyperscalers because of the enormous scale at which they operate, spent years buying wind and solar power. Now they are reversing course. All four have announced plans to build large natural gas power plants to keep their AI data centres running.
The pitch makes sense on paper. Gas is cheap right now. But a report from Noreva, an energy research firm, suggests the companies may be locking themselves into a fuel that is about to get much more expensive.
Why could prices spike so sharply?
Noreva expects prices to more than double, and in some areas triple, within a few years. Three forces are colliding at once: AI data centres are pulling in huge amounts of new demand, the growth in fresh gas supplies is slowing down, and new export pipelines are shipping more American gas overseas as liquefied natural gas.
West Texas is the clearest example. Most drilling there targets oil, and natural gas comes out as a cheap byproduct. There were not enough pipelines to move it elsewhere, so local producers sold it at a discount. New pipelines have changed that. "They've finally built some pipelines out there, and a lot of that is headed towards export markets," Peter Gardett, CEO of Noreva, told TechCrunch. Once West Texas gas connects to global prices, that regional bargain disappears.
Today, the widely traded Henry Hub benchmark in Louisiana sits just under $3 per million BTUs. Noreva sees some regional delivery points climbing above $10.
What does this mean for your bills and your AI tools?
Higher gas prices hit data centre operators through their electricity costs. Fuel covers roughly half the cost of generating electricity at a large plant, so a tripling of gas prices could make running AI services significantly more expensive.
That cost has to go somewhere. It could push up the price of AI tools that charge per query (each query is called a "token", the small unit of text an AI model processes). It could also push data centres back onto the shared electricity grid, raising bills for everyone nearby. Noreva notes that 80% of consumers already worry about data centres inflating their electricity costs, and natural gas bills could soon join that list.
For context, here is how the four main gas plant announcements stack up:
| Company | Plant size | Location |
|---|---|---|
| Meta | 7.5 GW | Louisiana |
| Amazon | 7.6 GW | Texas |
| Microsoft | Gigawatt-scale | Texas |
| Gigawatt-scale | Texas |
Gardett is blunt about the risk. "On future Alphabet earnings 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."
Futures markets, which are traders' best guesses about where prices are heading, are not pricing in a big jump yet. So the hyperscalers are not necessarily wrong today. They are just, according to Noreva, not prepared for what happens if Gardett's arithmetic turns out to be right.
Common questions
Will this make ChatGPT or Google AI more expensive to use?
Possibly, over time. If gas prices rise sharply and data centre costs climb with them, AI companies may raise per-use prices or reduce free tiers. Nothing is changing immediately, but it is worth watching.
Do these gas plants affect people who do not use AI?
Yes, potentially. Large gas consumers near major pipelines can influence regional prices for everyone, including households that heat with gas or whose electricity comes from gas-fired generators.
Is there a cheaper alternative the companies could switch to?
Renewable energy such as wind and solar is cheaper per unit in many places, but it is not always available when demand spikes. That reliability gap is the main reason hyperscalers say they need gas, at least for now.
