이미지: TechCrunch AI
Summary
- Energy research firm Noreva forecasts that natural gas prices in some parts of the US could rise as much as threefold within the next few years
- Meta, Microsoft, Google, and Amazon have each recently announced plans for gigawatt-scale, self-built gas power plants in Texas and Louisiana
- Gas accounts for roughly half of electricity costs at large power plants, meaning a price spike could directly affect AI data center operating costs and per-token pricing
- 예측 기관
- 에너지 리서치사 노레바(Noreva)
- 예상 가격
- 일부 허브에서 MMBtu당 10달러 이상(현재 2~4.5달러, 헨리허브 약 3달러)
- 메타
- 루이지애나에 7.5GW 가스발전소 건설(3월 발표, Hyperion 데이터센터용)
- 아마존
- 텍사스에 7.6GW 가스발전소 건설 계획
- 구글·마이크로소프트
- 텍사스에 각각 기가와트급 가스발전소 건설 발표
A Tripling Forecast Puts Big Tech on Edge
Energy research firm Noreva has forecast that natural gas prices in some regions of the US could rise as much as threefold over the next few years. The firm argues that surging demand from Big Tech to power AI data centers, combined with slowing supply growth and rising liquefied natural gas (LNG) exports, will tighten the market far more than currently expected. Noreva projects that prices at certain trading hubs (delivery points where futures contracts are settled) could exceed $10 per MMBtu (one million British thermal units). US gas prices currently range from $2 to $4.50 depending on region, with the benchmark Louisiana Henry Hub trading below $3. "Everyone had convinced themselves gas prices couldn't go up," Noreva CEO Peter Gardett told TechCrunch.
Why Big Tech Bet on Gas
After years of pouring money into wind and solar development, Big Tech has recently changed course. In March, Meta announced it would build a 7.5-gigawatt (GW) natural gas power plant in Louisiana to supply its Hyperion data center. Days later, Microsoft and Google each unveiled plans for their own gigawatt-scale gas power plants in Texas, and Amazon followed with plans for a 7.6GW plant, also in Texas. The shift toward gas reflects the reality that solar and wind, being weather-dependent, struggle to reliably support AI servers that run around the clock. The problem is that fuel costs—namely gas purchases—account for roughly half of electricity expenses at large power plants. If gas prices double or triple, the operating costs of these self-generated data centers would rise proportionally. Big Tech companies, which until recently had been reluctant to make large capital investments, now find themselves not only building physical power generation infrastructure but also taking on price volatility risk in an unfamiliar energy market. One investor who spoke with Noreva said they were surprised by the scale of gas price risk Big Tech companies are willing to absorb. As seen in SpaceX's case—delaying the removal of unpermitted turbines at its xAI data center by another year—securing power has already become a core bottleneck in the AI infrastructure race, one that is now outpacing the permitting process itself.
So What Changes
The gas futures market currently shows no signs of anticipating major swings, so there's no basis yet to conclude that Big Tech's bets are misguided. But if gas prices do rise as forecast, the effects could spread through two channels: rising operating costs at self-powered data centers getting passed through to per-token pricing for AI services, and Big Tech shifting away from self-generation toward greater reliance on regional power grids, which could push up electricity bills for households and businesses alike. As competition over AI model performance moves beyond securing chips and into managing price risk in energy markets, this forecast points to a financial vulnerability lurking behind the wave of data center construction announcements.
