Energy · Geothermal

Why Geothermal Is Finally Scaling — and What It Means for the Grid

Q: Why are data centres the first big customers?

Because they need firm, around-the-clock clean power and are willing to pay for it. Solar and wind are cheap but intermittent, and batteries are expensive beyond a few hours, so buyers with hourly clean-energy targets have limited options. On 1 September 2026 Fervo signed a 396 megawatt power purchase agreement with Google, described by the company as the largest enhanced geothermal deal to date, with an option for about 600 megawatts more by June 2030. Fervo also announced a 500 megawatt agreement with Shell Energy in 2025. Geothermal competes with gas peakers and long-duration storage rather than with solar on price.

📅 Sep 16, 2026 🏷️ Explainer · Evergreen 🕐 8 min read

By Luminesca · Updated Sep 16, 2026
Analysis compiled from public reporting with AI-assisted drafting. See our editorial policy.

Geothermal has spent decades as the energy transition's most reliable also-ran: clean, constant, and almost impossible to build at scale outside a handful of volcanic hotspots. That is changing — not because the heat underground became easier to reach, but because the drilling did. Developers are now borrowing oil-field rigs, horizontal wells and hydraulic fracturing to manufacture heat reservoirs in ordinary hot rock, turning a geography problem into an engineering one. The results have attracted bank debt, a public listing and power contracts from some of the world's largest electricity buyers. The question is no longer whether it works, but whether it can be repeated cheaply and safely enough to matter.

What next-generation geothermal actually means

Conventional geothermal needs hot rock, water and permeable rock to line up naturally. That combination is rare: the United States has roughly four gigawatts of geothermal capacity across 93 plants, mostly in seven western states. Against a grid the Department of Energy says needs 700 to 900 gigawatts of clean firm power, that is a rounding error.

Next-generation geothermal stops waiting for the geology to cooperate. In an enhanced geothermal system, or EGS, developers drill kilometres into hot, low-permeability rock, fracture it in stages, then pump water down one well and up another. A second family, closed-loop systems, circulates fluid through sealed pipes instead. The Department of Energy's Office of Geothermal puts conventional resources at about 40 gigawatts and next-generation potential at up to 5,500 gigawatts. Unlike solar and wind, the output is firm, dispatchable and fuel-free on a small land footprint, with lifecycle emissions estimated by the National Renewable Energy Laboratory at a median 32 grams of CO2 per kilowatt-hour, against 490 for gas and 820 for coal.

The drilling cost curve is the whole story

Every credible argument for geothermal now rests on one metric: how fast and how cheaply a well can be drilled. Fervo Energy offers the clearest evidence. Its first commercial well at Project Red in Nevada reached 11,220 feet in 70 days in 2022 and has supplied up to three megawatts on average since late 2023. Its Cape Station project in Utah has compressed that timeline sharply: in July 2026 the company announced that its Sawtooth 7 well reached 19,448 feet, with a 7,500-foot horizontal section in 460-degree-Fahrenheit rock, in 21 days — 143% faster than its first Cape Station well, on a harder design.

Costs followed. Fervo reports that drilling across its first four horizontal Cape wells fell from 9.4 million dollars to 4.8 million dollars per well. Cape Phase One ran near 7,000 dollars per kilowatt of capacity; Phase Two is now on track for 5,500 dollars, against a long-term target of 3,000. That is why banks arrived, followed by roughly two billion dollars in a May listing. The caveats deserve equal weight: reporting on the drilling record notes that Phase One capacity has yet to reach the grid, and independent comparisons still place next-generation projects well above the 4,800 to 6,300 dollars per kilowatt typical of conventional plants.

What the Utah FORGE tests proved — and what they did not

Underneath the commercial story sits a publicly funded experiment that changed how the industry understands reservoir creation. Utah FORGE, a 218-million-dollar Department of Energy project run by the University of Utah near Milford, fractured rock in 12 stages across a pair of deep wells. The circulation test that followed recovered about 70% of the injected water at 282 degrees Fahrenheit, and a later 28-day test — summarised in a technical update for the Electric Power Research Institute — sustained commercial flow rates with roughly 90% of the water returned at 190 degrees Celsius. The most transferable finding was mundane: diamond drill bits designed for hard, hot crystalline rock raised the rate of penetration by about 50% and cut drilling costs by roughly 30%.

What the tests did not prove is longevity. A circulation test shows that fluid can travel between two wells; it does not show how a reservoir behaves over the decades a power plant needs to repay its capital. The failure mode engineers fear is thermal breakthrough: if fractures link the wells too directly, injected water returns before it has absorbed enough heat.

The risk that ended earlier projects

Enhanced geothermal has a graveyard, and its occupants died of the same cause. High-pressure injection at Basel in Switzerland triggered seismicity up to magnitude 3.4 in 2006 and the project was abandoned. The magnitude 5.5 Pohang earthquake in South Korea in 2017 was linked to EGS stimulation and ended that project too. The industry now operates under traffic light protocols: continuous seismic monitoring, a green band below roughly magnitude 1, injection paused or reduced in an amber band around magnitude 2, and a full stop above it. Utah FORGE recorded nothing larger than magnitude 1.9. Because the volume, rate and placement of injected fluid relative to critically stressed faults drives the seismic response, site selection carries more weight than any protocol applied afterwards.

Water is the second constraint. An EGS loop does not recover every litre it injects, and in the arid western United States that matters. Developers have responded with non-potable water for stimulation, air-cooled plants, and steel casing and cement to isolate deep wells from freshwater aquifers. Reviews also flag that hot circulation through granitic rock can mobilise trace elements such as arsenic and boron.

Why data centres are the first real customers

The most revealing signal is commercial rather than technical: who is willing to sign. On 1 September 2026 Fervo announced a 396-megawatt power purchase agreement with Google, described by the company as the largest enhanced geothermal deal to date, with an option for roughly 600 megawatts more by June 2030.

The pattern is not accidental. Hyperscale buyers with hourly clean-energy targets need power available every hour, and their alternatives are gas plants, nuclear, or batteries that become prohibitively expensive beyond a few hours. Geothermal's selling point is therefore not cheapness but firmness: it substitutes for the gas peaker and the long-duration battery, not the solar farm. Cape Station reflects that logic — 500 megawatts under construction, first power targeted for late 2026, roughly 100 megawatts operating by early 2027, and an assessed site potential of 4.3 gigawatts.

What it means

Set against the scale of the problem, the sector is still small, and it has been shrinking: between 2016 and 2021, seven new geothermal plants totalling 186 megawatts came online in the United States while 11 older plants retired, taking 103 megawatts off the grid. The policy machinery has shifted up a gear — 171.5 million dollars opened by the Department of Energy's Office of Geothermal in February 2026 for next-generation field tests, and a 14-million-dollar project to convert an existing Pennsylvania shale gas well into an eastern test site.

The honest conclusion is that geothermal's bottleneck has moved. For fifty years the binding constraint was geology: you could only build where nature had assembled heat, water and permeability. What limits the industry now is execution at repetition, an operating record that does not yet exist, and capital that must be spent before the reservoir is proven. Government projections put next-generation geothermal between 90 and 300 gigawatts by 2050. If the drilling curve keeps bending, 90 gigawatts displacing gas would avoid roughly 325 million tonnes of CO2. If it flattens, geothermal remains an excellent resource in a small number of very fortunate places.

Frequently asked questions

What is the difference between conventional and enhanced geothermal?

Conventional geothermal needs three things to line up naturally: hot rock, water, and rock permeable enough for that water to circulate. That combination is rare, which is why the United States has only about four gigawatts of geothermal capacity and 93 plants concentrated in seven western states. Enhanced geothermal systems create the missing conditions instead. Developers drill into hot, low-permeability rock, fracture it hydraulically, inject water down one well and bring heated fluid up another. The Department of Energy estimates conventional resources at roughly 40 gigawatts and next-generation potential at up to 5,500 gigawatts.

Does enhanced geothermal cause earthquakes?

It can, and two projects show the cost of getting it wrong. High-pressure injection at Basel in Switzerland triggered seismicity up to magnitude 3.4 in 2006 and the project was abandoned; the 2017 magnitude 5.5 Pohang earthquake in South Korea was linked to EGS stimulation and ended that project too. The industry response is the traffic light protocol: continuous seismic monitoring with a green band below roughly magnitude 1, injection paused or reduced in an amber band around magnitude 2, and a full stop above it. At the Utah FORGE test site the largest event recorded during stimulation was magnitude 1.9, with no felt seismicity. Site selection, especially distance from critically stressed faults, matters more than any operational safeguard.

How much does enhanced geothermal cost today?

It is still more expensive than conventional geothermal and far more expensive than solar or wind. Conventional plants cost roughly 4,800 to 6,300 dollars per kilowatt of capacity, while next-generation projects have been reported at around 15,000 dollars per kilowatt. Fervo Energy says Cape Station was near 7,000 dollars per kilowatt in Phase One and is on track for 5,500 in Phase Two, with a long-term target of 3,000. The Department of Energy's goal is about 45 dollars per megawatt-hour by 2035; current geothermal contracts sign in the 70 to 100 dollar per megawatt-hour range, a premium buyers pay for around-the-clock delivery.

Why are data centres the first big customers?

Because they need firm, around-the-clock clean power and will pay for it. Solar and wind are cheap but intermittent, and batteries are expensive beyond a few hours, so buyers with hourly clean-energy targets have limited options. On 1 September 2026 Fervo signed a 396-megawatt power purchase agreement with Google, described by the company as the largest enhanced geothermal deal to date, with an option for about 600 megawatts more by June 2030. Fervo also announced a 500-megawatt agreement with Shell Energy in 2025. Geothermal competes with gas peakers and long-duration storage rather than with solar on price.

What still has to be proven?

Two things: repeatability and reservoir life. Fervo's first commercial-scale plant is not yet generating power, with first delivery targeted for late 2026, and no enhanced geothermal reservoir has operated for the two to three decades that would show how heat output holds up. Thermal breakthrough, where fractures connect the injection and production wells so directly that cool water returns before it absorbs enough heat, is a known design failure mode. The other open question is cost discipline at scale: a fast well proves a rig can perform, not that an entire fleet of reservoirs can be built on budget.