A geothermal power plant makes electricity by pulling heat out of the ground and using it to spin a turbine, the same basic principle behind a coal or gas plant, minus the fuel and the smokestack. The heat comes from hot water or steam trapped in underground rock, and how a plant handles that fluid once it reaches the surface depends entirely on how hot it is.
Three ways to turn underground heat into power
There are three commercial designs, and resource temperature decides which one a site uses.
Dry steam plants are the oldest kind, first built in Larderello, Italy, in 1904, and they work when the reservoir produces fluid that’s already mostly steam. That steam runs straight to the turbine, no processing needed. The Geysers in Northern California is the largest single geothermal source in the world and still uses this design.
Flash steam plants are the most common type running today. They tap fluid hotter than 182°C, or 360°F, and as that pressurized water rises toward the surface and hits lower pressure, part of it flashes into steam almost instantly. That steam spins the turbine, and the leftover liquid can be flashed a second time to pull out more energy before it goes back underground.
Binary plants are what let developers build where the resource isn’t scalding hot. The geothermal fluid, usually below 182°C, never touches the turbine. It passes through a heat exchanger and heats a second fluid with a much lower boiling point, often a hydrocarbon like isobutane, which vaporizes from that modest heat and drives the turbine on its own closed loop.
| Plant type | Resource temperature | How it drives the turbine |
|---|---|---|
| Dry steam | Steam already dominant underground | Steam piped directly to the turbine |
| Flash steam | Above 182°C (360°F) | Pressurized fluid flashes to steam at the surface |
| Binary cycle | Below 182°C (360°F) | Heat exchanger boils a separate low-boiling-point fluid |
Where the world’s geothermal capacity actually sits
Geothermal meets less than 1% of world energy demand, per the International Energy Agency’s 2024 assessment, but it runs hard where it exists. Plants worldwide averaged over 75% capacity utilization in 2023, against under 30% for wind and under 15% for solar.
The United States leads in raw capacity at 3,953 MW as of year-end 2025. Indonesia is second at 2,742 MW and the Philippines third at 2,034 MW. Türkiye is fourth at 1,797 MW and is now Europe’s largest geothermal producer. New Zealand sits at 1,259 MW and Kenya at 980 MW, with Iceland rounding out the group at 808 MW. Those ten countries hold more than 93% of the roughly 17,173 MW installed worldwide.
Capacity numbers only tell part of the story. Kenya generates a higher share of its electricity from geothermal than any other country, about 43% as of 2022 data from the U.S. Energy Information Administration, because the Rift Valley gives it an unusually productive resource relative to its grid size. Indonesia, despite running the world’s second-largest fleet, pulls only about 5% of its own electricity from geothermal, since its overall power demand dwarfs what those plants produce. In the U.S., California accounts for 68.6% of national geothermal generation and Nevada another 24.7%, though geothermal is still just 5.2% of California’s mix.
What makes enhanced geothermal different
Every plant above depends on a natural hydrothermal reservoir, water-bearing rock with enough permeability for that water to circulate on its own. Most of the planet’s heat isn’t packaged that way. Enhanced geothermal systems, usually shortened to EGS, target hot rock that has the temperature but not the plumbing. Engineers drill into it, pump water in under pressure to open or widen existing fractures, then circulate fluid between an injection well and a production well, pulling heated water back up to run a plant, typically a binary one.
If EGS works at scale, geothermal stops being limited to volcanic regions and becomes available almost anywhere a company can drill deep enough. The IEA’s 2024 report put the ceiling high: up to 800 GW of geothermal capacity worldwide, producing close to 6,000 terawatt-hours a year, roughly matching current U.S. and Indian electricity demand combined. The same report projects EGS costs could fall 80% by 2035, to around $50 per megawatt-hour. Fewer than 30 countries currently have a geothermal-specific policy on the books, against more than 100 for wind and solar, which the IEA flags as the real bottleneck.
The Utah projects testing whether EGS scales
The clearest evidence that EGS has moved past the lab sits in Beaver County, Utah.
Fervo Energy proved the concept first in Nevada. Its Project Red pilot, completed in July 2023, used horizontal wells drilled to about 8,000 feet true vertical depth with roughly 3,250-foot horizontal laterals, the first time anyone had applied horizontal drilling to an EGS project. It produced 3.5 MW of continuous power from flow rates up to 63 liters per second.
Cape Station, Fervo’s follow-on project near Milford, is a different scale entirely. The company broke ground in September 2023 targeting up to 500 MW in its initial phase, with permits for as much as 2 GW. A 30-day test on one well hit a flow rate of 107 kilograms per second at temperatures above 428°F, enough to support more than 10 MW from that single well, roughly three times Project Red’s output and, Fervo says, beyond what the National Renewable Energy Laboratory had projected the industry would reach before 2035. Southern California Edison and the Clean Power Alliance have signed power purchase agreements with the project, along with Shell Energy North America for a smaller 31 MW slice. In September 2026, Google signed on for nearly 400 MW of the second phase to help power a planned Utah data center.
A few miles away sits Utah FORGE, the Department of Energy’s dedicated EGS test site, run by the University of Utah. It isn’t a commercial plant, just a place for researchers to try stimulation and monitoring techniques without revenue riding on the outcome. In September 2024, FORGE completed extended circulation tests showing solid connectivity between its wells and temperatures around 370°F, and the DOE renewed its funding through 2028, adding $80 million.
Drilling has gotten measurably cheaper along the way. A 2025 Stanford Geothermal Workshop study by researchers Akindipe and Witter found geothermal well costs down 12% to 24% for vertical wells and 18% to 26% for horizontal ones versus a 2017 baseline, mostly from borrowing oil-and-gas hardware like diamond drill bits and pad drilling that spreads fixed costs across several wells.
Why EGS still isn’t cheap or fast
None of that means EGS is about to get solar-cheap on solar’s timeline. A November 2025 analysis from CleanTechnica laid out the case that EGS can’t repeat the cost collapse solar and batteries went through. Directional drilling and hydraulic stimulation are both mature, decades-old technologies, so EGS mostly assembles existing tools rather than inventing new ones with room for a steep learning curve. Solar panels and batteries have roughly doubled their cumulative output every one or two years, while a single EGS project takes four and a half to seven years from first drilling to commercial power, leaving only a couple of true global doublings per decade. Oil and gas has already drilled millions of wells. EGS might drill a few thousand globally before it counts as a mature industry, too small a volume to bend its cost curve the way solar’s did. Today’s EGS projects run $10,000 to $15,000 per kilowatt of capacity, against $800 to $1,200 for utility solar, and even an aggressive 40% cut over the next couple of decades leaves it well above where solar sits now.
Induced seismicity is the other hard limit, and it has already killed a project outright. The Basel, Switzerland EGS project began hydraulic stimulation in December 2006 and triggered more than 13,500 recorded seismic events, several above magnitude 3, the strongest hitting 3.4. A magnitude 2.9 quake six days into stimulation tripped the project’s automatic shutoff, and insurers eventually processed roughly 2,700 damage claims worth 7 to 9 million Swiss francs. A three-year study concluded the site would likely keep producing small quakes through its planned 30-year life, and Basel’s operators canceled the project in December 2009. That’s why today’s EGS developers run real-time seismic monitoring with traffic-light protocols that force a slowdown if shaking crosses a set threshold, and why regulators weigh a site’s proximity to faults and population centers before permitting it.
Permitting on U.S. federal land has been the slower-moving obstacle, and it’s only recently started to loosen. Most of the country’s best geothermal ground sits on land managed by the Bureau of Land Management, which oversees roughly 245 million acres with geothermal leasing authority plus another 104 million acres under Forest Service jurisdiction. In April 2024, the BLM adopted categorical exclusions that let certain exploration work skip a full environmental assessment. That speeds up scouting and testing a site, but building the plant itself still triggers the standard review process.
None of this should surprise anyone who’s spent years around drilling rigs and turbines. EGS doesn’t need new hardware, just oil and gas drilling tricks pointed at a different resource. What’s new is the buyer: data center operators who need round-the-clock power badly enough to sign long contracts and help someone build it. Whether that demand carries EGS past its cost and seismicity hurdles at real scale is still open, and Cape Station’s output over the next few years will answer more of that than any model can.
