Artificial intelligence may feel weightless. A prompt goes in, an answer comes out, and somewhere far away a server thinks really hard. But behind that digital magic is something very physical: buildings packed with chips, pulling huge amounts of electricity and producing massive amounts of heat.
Data centers used about 4.4% of all U.S. electricity in 2023, up from 1.9% in 2018, and that share could rise to 6.7–12% by 2028. Globally, data centers used about 415 terawatt-hours in 2024, and the International Energy Agency projects that could more than double to around 945 terawatt-hours by 2030.
That raises a big question: How do we power AI without the grid filing for emotional support? One promising answer is geothermal energy.
Why Geothermal Makes Sense for AI
Solar and wind are great, but they are variable. Solar fades at night. Wind comes and goes. AI data centers, however, need power constantly. Servers do not care if it is cloudy, calm, or 2:17 a.m.
Geothermal energy is different. It uses heat from inside the Earth to generate electricity, and because Earth’s internal heat is always there, geothermal plants can provide steady, around-the-clock power. The U.S. Department of Energy says geothermal plants often operate with a capacity factor around 90%, making them useful for 24/7 loads like data centers.
That is why geothermal is suddenly getting attention from Big Tech. AI does not just need clean power. It needs clean power that does not blink.
How a Geothermal-Powered AI Center Could Work
A geothermal-AI facility could work in a few different ways.
The simplest version is grid-connected. A geothermal plant produces electricity and sends it into the same regional grid that serves the data center. The data center still gets electricity through the grid, but the geothermal project adds clean, steady power to match its demand.
A more ambitious version is “behind-the-meter,” where the geothermal plant is built close to the data center and directly supplies much of its power. That could reduce grid congestion and shorten the wait for new transmission lines.
The most interesting version combines power generation and cooling:
Deep geothermal wells
↓
Geothermal plant generates electricity
↓
Electricity powers AI servers
↓
Liquid cooling captures server heat
↓
Heat is reused, stored underground, or partly converted back to electricity
↓
Cooled fluid returns to the system
This is where it gets fun. Data centers do not just consume power; they also produce a lot of heat. Nearly all electricity used by servers eventually becomes heat. Liquid cooling can capture that heat more efficiently than air cooling.
Could that heat be turned back into electricity? Technically yes, using systems like an Organic Rankine Cycle. But there is a catch: server waste heat is usually low-temperature heat, often better for heating buildings, greenhouses, or industrial processes than for making large amounts of electricity. So the best use is probably not “power the whole data center with its own waste heat.” It is more like recovering some value instead of dumping all the heat into the atmosphere.
Who Has Bought Into This?
Google has been one of the clearest early movers. In 2023, Google announced that its partnership with Fervo Energy had produced operational enhanced geothermal power in Nevada, adding carbon-free electricity to the grid serving Google’s data centers in the state.
Meta has also jumped in. In 2024, Meta partnered with Sage Geosystems on a project meant to provide geothermal power for Meta data centers. Meta described it as a first-of-its-kind project using Sage’s geopressured geothermal system.
Then in 2025, Meta signed another deal with XGS Energy to develop 150 MW of next-generation geothermal power in New Mexico. XGS says its system is designed for zero operating water use, which matters a lot in dry western states. That project is expected to come online in phases by 2030.
Microsoft and G42 also announced plans for a geothermal-powered data center in Kenya as part of a $1 billion digital investment, though reports in 2026 said Kenya had suspended that project.
How Many AI Data Centers Already Use Geothermal?
This is where the answer gets less shiny.
As of now, geothermal-powered AI/data-center projects are still early-stage. Google’s Nevada project is the clearest operational example tied to data-center electricity supply. Meta’s Sage and XGS projects are announced or under development, but not fully online yet.
There is no reliable public count of “AI data centers powered by geothermal,” partly because most companies buy electricity through grids and power purchase agreements rather than directly plugging a data center into one plant.
Why This Could Become a Big Deal
A Rhodium Group analysis found that behind-the-meter geothermal could potentially meet 55–64% of projected hyperscale data-center growth in a clustered siting scenario, and if companies intentionally build data centers near the best geothermal resources, geothermal could meet all projected growth in the early 2030s in their modeled scenarios.
That does not mean it will happen automatically. It means the resource potential is real. For AI companies chasing enormous power loads, geothermal is attractive because it acts more like a clean power plant than an intermittent renewable source.
The Water Question
Data centers can also use massive amounts of water for cooling. Some large data centers may consume up to 5 million gallons per day, depending on size, climate, and cooling design.
Traditional cooling towers lose water through evaporation. That water is not destroyed, but it leaves the local watershed as vapor and may fall as rain somewhere else later. This is why geothermal projects that reduce water use are especially interesting. XGS, for example, says its New Mexico project uses technology designed for zero operating water use, decoupling geothermal production from natural groundwater resources.
That matters because the AI boom is not just an energy issue. It is also a water issue.
The Disadvantages
Before we crown geothermal as AI's energy savior, it's worth remembering that hot rocks aren't a silver bullet. Geothermal projects are expensive to drill, slow to develop, and heavily dependent on local geology. Not every region has the right temperatures or rock conditions, and some enhanced geothermal systems can trigger small earthquakes if not carefully monitored. While data center waste heat can be reused, it generally isn't hot enough to generate much electricity on its own. Perhaps the biggest challenge is scale: modern AI campuses can require as much power as a small city. Geothermal can help meet that demand, but for now it's more likely to be a key player on the team rather than the entire roster.
So Is Geothermal AI’s Sustainable Power Solution?
Whether geothermal becomes AI's long-term energy solution remains uncertain. The technology offers reliable, around-the-clock power and aligns well with the enormous energy demands of data centers, but challenges related to cost, geology, and scale still remain. What is clear is that geothermal is gaining serious attention from some of the world's largest technology companies. As AI continues to grow, the future may involve a mix of energy sources rather than a single answer, but geothermal could prove to be one of the most important pieces of the puzzle.
References
Energy Efficiency and Sustainable Energy (EESI). (2024). Data centers and water consumption. Retrieved from https://www.eesi.org/articles/view/data-centers-and-water-consumption
Fervo Energy. (2025). The enhanced geothermal data center corridor. Retrieved from https://fervoenergy.com/fervo-uipa-the-enhanced-geothermal-data-center-corridor-july-2025/
Google. (2023). Google and Fervo Energy bring next-generation geothermal power online. Retrieved from https://blog.google/company-news/outreach-and-initiatives/sustainability/google-fervo-geothermal-energy-partnership/
International Energy Agency (IEA). (2025). Energy and AI. Paris, France: International Energy Agency.
Lawrence Berkeley National Laboratory. (2024). United States data center energy usage report. Berkeley, CA: Lawrence Berkeley National Laboratory. Retrieved from https://eta-publications.lbl.gov/sites/default/files/2024-12/lbnl-2024-united-states-data-center-energy-usage-report_1.pdf
Meta Platforms, Inc. (2024). New geothermal energy project to support our data centers. Retrieved from https://about.fb.com/news/2024/08/new-geothermal-energy-project-to-support-our-data-centers/
Rhodium Group. (2025). Can geothermal power meet growing data center electricity demand? Retrieved from https://rhg.com/research/geothermal-data-center-electricity-demand/
Reuters. (2024). Microsoft and G42 invest $1 billion in Kenya to build data center. Retrieved from https://www.reuters.com/technology/microsoft-g42-invest-1-billion-kenya-build-data-center-2024-05-22/
U.S. Department of Energy. (2025). Geothermal energy and data centers. Washington, D.C.: U.S. Department of Energy. Retrieved from https://www.energy.gov/hgeo/geothermal/geothermal-and-data-centers
XGS Energy. (2025). XGS Energy and Meta partner on 150 MW advanced geothermal project. Retrieved from https://www.xgsenergy.com/xgs-energy-and-meta-to-partner-on-150-mw-advanced-geothermal-project/
Add comment
Comments