Utah FORGE: The Geothermal Lab That Makes No Electricity

Utah FORGE is a geothermal project that does not make electricity, and that is the entire point. About 10 miles northeast of Milford, on the western flank of the Mineral Mountains in Beaver County, Utah, the U.S. Department of Energy funds a pair of deep wells drilled into hot granite with no commercially useful natural flow through it. The rock is not bone dry — the pilot well logged thousands of natural fractures — but its permeability is far too low to run a power plant on. The job is to find out whether engineers can manufacture a working reservoir in that kind of rock, measure it exhaustively, and publish the data so private companies do not each have to learn the same lessons at their own expense.

The operator's own words are the plainest summary: "We are not a geothermal plant." "We will not be producing electricity to send to the grid." There is not even a laboratory building. The laboratory is the wells, the surface equipment, the sensors, the experiments, and the volume of rock itself.

A new extended circulation test started on August 12, 2026 and was planned to run for roughly 90 days, possibly 120. As of September 12, 2026, the project had not announced results or confirmed whether the test had run without interruption.

What Utah FORGE Is

FORGE is the Department of Energy's dedicated field site for developing, testing, and accelerating enhanced geothermal system technologies. An enhanced geothermal system, or EGS, takes hot rock that lacks the water and permeability a conventional geothermal plant depends on and engineers both in: drill into it, crack it open under pressure, pump water through the cracks to pick up heat. Our explainer on enhanced geothermal systems covers the technology in full. FORGE is where the tools for doing it get tested at full scale, on the government's dime, with the results made public.

The operator, the University of Utah's Energy & Geoscience Institute, describes it as "the world's only full-scale field laboratory for de-risking and testing" EGS tools and technologies. "De-risking" is doing the work in that sentence. A geothermal power plant is a commercial bet on a resource. FORGE exists to make that bet less of a gamble for whoever comes next. It sits beside operating wind, solar, and conventional geothermal facilities, all of them selling electricity, and is itself an underground experiment that sells none.

Even the word "reservoir" means something different here. It does not mean a pooled body of water underground; the operator defines it as a system of small enhanced fractures through which injected water moves and is heated.

Why Beaver County

FORGE ran as a competition. Phase 1, in 2015 and 2016, had five candidate teams (Fallon, Newberry, Snake River Plain, Milford, and the west flank of Coso) sharing $2 million. Phase 2, from 2016 to 2018, narrowed it to Utah and Fallon, which instrumented, characterized, and permitted their sites. On June 14, 2018, after three years of planning, site characterization, and competition, DOE selected the University of Utah's Milford site. Phase 3 is the one that built what is there now: full-size drilling, reservoir stimulation and testing, site monitoring, and competitively selected research. DOE's phases page labels Phase 3 "2018–present", while the FORGE roadmap describes the June 2018 selection as beginning Phase 2C with Phase 3 following around mid-2019.

The site is on state land leased from the Utah Trust Lands Administration, reachable year-round by nearby paved and graded roads, in eastern Beaver County next to a wind farm and a solar photovoltaic power station.

How big is it? The published figures do not agree. A 2016 paper hosted by the operator says about 25 km². DOE's phases page says approximately 15 square miles. Another DOE page says less than five square miles. The figures may refer to different boundaries — the project area, the lease, the monitoring footprint or the operating zone — and none of the sources says which.

What is underneath is less ambiguous: crystalline granitic basement, or granitoid, beneath a layer of sedimentary basin fill. The 2016 paper describes the site as underlain by granite and gneiss at 175 to 225°C at depths of 2 to 4 km.

The Dry Hole: Acord-1

The case for this patch of ground rests on a failed well from 1979.

The Acord-1 exploration well was drilled nearby that year to 3.8 km. It hit granite and gneiss at 3.1 km and reached 230°C at total depth. And it produced no fluid at all. Its temperature profile was conductive, meaning heat was moving through the rock by conduction rather than being carried by circulating water, which told the geologists the deep crystalline rock was impermeable. For a conventional developer, that was the end of the story. No water, no plant.

For FORGE, it was the qualification. DOE's framing when it launched the program in 2015 was that the absence of commercially useful natural permeability is intentional: FORGE tests whether engineers can create and control fluid pathways in hot rock that otherwise cannot sustain conventional geothermal production. The experiment needs rock that does not already work. Acord-1 had proven this rock does not.

The Wells

How many wells FORGE has depends on what you count. The operator's dashboard lists seven project wells: 58-32, 16A(78)-32, 16B(78)-32, 68-32, 78-32, 78B-32, and 56-32. Its 2023 year-in-review counts eight, because it includes the water well, 58B-32, which the dashboard shows separately at 1,200 ft. Three groups: the pilot well that proved the rock, the monitoring wells that listen, and the two operating wells that do the work.

The pilot well: 58-32

Well 58-32, originally MU-ESW1, is a vertical pilot and test well completed in 2017 to 7,536 feet. It penetrated more than 4,500 feet of granite and recorded a bottom-hole temperature that the project's own records give as either 386°F (197°C) or 390°F (199°C), depending on the entry. It was also the first well stimulated: testing in 2017 indicated hydraulic fracturing and shearing, followed by a three-stage program between April 19 and May 3, 2019.

The listeners: 68-32, 78-32, 56-32, 78B-32

Because the whole experiment is about cracking rock under pressure, FORGE needed to hear the rock crack. Four wells exist mostly for that. Two shallow ones went in during March 2019: 68-32, with a permanently installed geophone and accelerometer near 921 to 925 ft, and 78-32, which reaches granitic basement at about 2,600 ft and carries a distributed acoustic sensing cable cemented outside its casing to 3,268 ft. Distributed acoustic sensing, or DAS, uses a fiber-optic cable as a continuous sensor along its whole length; at FORGE the fiber reads acoustic activity, temperature and strain, with pressure measured by a separate downhole gauge. Drilling 78-32 also found 46°C, non-potable water at about 700 to 900 ft flowing at 200 gallons per minute. Two deeper ones came in 2021: 56-32, drilled vertically to 9,145 feet in February with methods that improved rate of penetration by more than 60% versus 16A-32 and 180% versus 58-32, and 78B-32, near-vertical, drilled June 27 through July 31 to approximately 9,500 ft per the dashboard, with DAS as its monitoring technology. The network was designed for a magnitude -2 detection limit.

The doublet: 16A(78)-32 and 16B(78)-32

The two operating wells form a doublet: one injects, one produces. 16A(78)-32 is the injection well and 16B(78)-32 is the production well.

16A spudded on October 30, 2020 (spudding is the moment the bit first breaks ground). The operator dates completion to January 2021 on one page and to December 31, 2020 in its January 2021 newsletter. It went down vertically through approximately 4,700 ft of sediments, then, after reaching 6,000 ft, turned to 65° from vertical. The tilt is deliberate: highly deviated trajectories are intended to intersect numerous subvertical fractures and maximize potential heat-exchange contact. A slanted well crosses more of the up-and-down cracks in the granite than a vertical one would. The operator's dashboard puts 16A at 10,987 ft measured depth (the length of the hole) and 8,559 ft true vertical depth (how far straight down it goes). Its own May 2024 stimulation release says 10,897 ft instead. The dashboard figure matches the drilling-completion release. Preliminary measurements placed the temperature at the toe above 442°F (228°C). Drilling finished 60 days ahead of schedule and recovered approximately 74 ft of core.

16B, the production well, spudded April 26, 2023 and reached total depth by June 20, 2023, at 10,947 ft measured depth and a vertical depth given as 8,357 ft in the drilling record and 8,391 ft in the project's 2024 annual report. It runs approximately parallel to 16A and 300 ft above it; both lateral sections are inclined 65° from vertical. Roughly 240 ft of 4-inch core was the objective between 4,855 ft and 10,493 ft; the detailed coring report puts the amount actually recovered at about 210 ft. The well also hosted a supplier trial: Eavor Technologies supplied 350 joints of insulated drill pipe for two consecutive bottom-hole assemblies, judged successful because the pipe reduced heat transfer into the drilling string and kept the bottom-hole assembly in cooler fluid. For how holes like these get made, see our guide to geothermal drilling and wells.

Making a Reservoir

Two wells a few hundred feet apart in hot granite get you nothing until water can move from one to the other. Stimulation is pumping water into the well at pressures high enough to open fractures in the rock. FORGE has done it in steps, each bigger than the last.

2019, pilot well. The three stages in 58-32 produced 424 locatable microearthquakes with magnitudes from -2.0 to -0.52.

2022, first trial in 16A. From April 14 through April 24, 2022, 16A got a 10-day, three-stage hydraulic-stimulation trial. All three intervals were deeper than 10,000 ft: the open-hole toe (the uncased end of the well) and two shallower, cased 20-ft intervals. The three stages took 4,261, 2,777 and 3,016 barrels of water under pressure. The trial showed that bridge plugs (tools that seal off one section of a well while another is treated) could be set and retrieved at geothermal pressures and temperatures, and that mapped fracture networks could be created near the toe of 16A. Maximum induced magnitude: M0.5.

2023, first connection. Almost as soon as 16B was drilled, the team ran a brief circulation experiment between the two wells on July 19 and 20, 2023. It revealed some interwell connectivity. The maximum induced magnitude was M0.45.

2024, commercial scale. Beginning April 3, 2024 and running approximately two weeks, FORGE re-fractured previously treated sections of 16A, perforated and fractured additional intervals, and injected four stages in 16B, using 118,000 barrels (4,956,000 gallons) of non-potable water from a shallow well on the site. Fiber-optic signals in 16B identified likely fracture intersections created from 16A, which guided where to perforate and stimulate 16B. The program also evaluated proppant (material pumped into fractures to hold them open), custom frac plugs, multiple perforation clusters per stage, and different treating fluids.

The reported stage count is inconsistent. The operator's May 23, 2024 release says 16A was fractured in "eight different stages," then describes three old intervals re-fractured and seven new intervals perforated and fractured. The release does not resolve the discrepancy.

The 2024 Tests

Stimulation makes the cracks. Circulation tests whether they work as a heat exchanger. Two tests in 2024 answered part of that question.

The nine-hour test (April 2024)

A nine-hour circulation test injected water into 16A at rates up to 15 barrels per minute (630 gallons per minute). Production from 16B reached up to 8 barrels per minute, which the operator gives as 344 gallons per minute — though at the standard 42-gallon petroleum barrel, 8 bpm is 336 gpm. The operator characterized recovery as "around 70%" and reported produced-water temperature rising to approximately 139°C (282°F). The release does not reconcile those flow figures with each other. Microseismic observations, fiber-optic data, and spinner logging (a downhole tool that measures where along the well fluid is flowing) indicated multiple independent flow paths between the wells. The maximum induced seismic event during the April 2024 stimulation and short circulation was magnitude 1.9.

The extended test (August to September 2024)

The raw-data archive defines the 2024 extended circulation test as running from August 8 through September 5, 2024. During that nearly month-long run, water was injected at a consistent 420 gallons per minute. Tracers (chemicals added to the injected water so their arrival at the other well can be timed) tracked fluid movement, and the test monitored flow, temperature, microseismicity, fiber-optic strain and pressure, and well condition through downhole surveys.

The headline results, per the operator's September 9, 2024 report: "more than 90%" of the fluid came back, and the temperature remained approximately 370°F. That recovery figure is ambiguous: the operator writes it as a share of produced fluid without stating the denominator. Its August 12, 2026 retrospective describes the same result as "nearly 90%."

What they proved, and what they did not

Proved, in the operator's framing: engineered connectivity between the wells, circulation through hot dry granite, recovery of heated water, and short-term heat transfer.

Not proved: long-term reservoir sustainability. The operator said so directly, noting that longer tests were required to evaluate performance and sustainability. (If you are tracking how far commercial EGS has come alongside FORGE, see our news coverage of the enhanced geothermal commercial milestone.) A month tells you the reservoir works. It does not tell you whether it keeps working, whether it cools off, or whether the water keeps coming back.

What Is Happening Right Now

On August 12, 2026, Utah FORGE began a new extended circulation test expected to last "upwards of four months." The plan ramps injection in steps, 5 barrels per minute, then 7.5, then 10, over approximately 48 hours, and then holds. The closed-loop test is designed to circulate cooler, non-potable water from 16A to 16B at 10 bpm for roughly 90 days, with a possible extension to 120 days.

The test measures pressure and production stability, water recovery, heat extraction and thermal decline, per DOE's August 17, 2026 update, along with the scaling and corrosion that drive water treatment, named in the operator's own August 12 release. Principal Investigator Kristie McLin said in August 2026 that thermal breakthrough (the point at which cooler injected water reaches the production well and pulls the temperature down) and water loss over time remained unknown. DOE says extended operating data are needed to determine project economics and to design future commercial plants with greater confidence.

As of September 12, 2026, both Utah FORGE and DOE described the test as underway, in updates dated August 12 and August 17 respectively. No live operational log exists. Ninety days from August 12 runs into November, so the published plan has the test still circulating through the autumn. No results have been published, including recovery, temperature and seismicity figures. The next milestone is completion and analysis of roughly 90 days of continuous circulation, possibly extended to 120.

The Earthquake Question

Any project that fractures deep rock under pressure gets asked about earthquakes. FORGE fractures rock deliberately. The small earthquakes are the expected consequence, and a published protocol sets out what happens as they get bigger. The record runs from the 424 microearthquakes of the 2019 pilot-well stimulation (magnitude -2.0 to -0.52), through M0.5 in 2022 and M0.45 in 2023, to a maximum of magnitude 1.9 during the April 2024 stimulation and short circulation. Negative magnitudes are events far too small to feel; the network detects down to magnitude -2 precisely so the cracking can be mapped.

The rulebook is a traffic-light system, Version 2 of which is published in the Geothermal Data Repository. It works in tiers:

• Adaptive trigger: 10 events of magnitude 1 or greater within 24 hours and within 3 km of the FORGE reservoir. Where those events fall inside the stimulated volume, the response is a 10% step-down in injection and a reassessment rather than an immediate stop; events outside the FORGE reservoir can allow operations to continue. A cluster propagating along an imaged fault is treated differently.
• Magnitude 2 or greater within 3 km, inside the FORGE reservoir: controlled shut-in and suspension until a continuation plan is approved. An event of the same size outside the reservoir generally does not stop work, unless a deep cluster suggests it is tracking an imaged fault plane.
• Magnitude 3 or greater within 5 km: controlled flowback, securing the well, cessation of operations, and approval from the Science and Technology Analysis Team and DOE before anything resumes.
• Magnitude 3 or greater at 5 to 15 km: a University of Utah Seismograph Stations press release, but no operating change.

The largest event on record at the site, the 1.9 in April 2024, sat below the M2 shut-in line. Four wells at Fervo Energy's neighbouring Cape Station project terminate at the Utah FORGE lease boundary, less than 2,000 ft from FORGE's own injection well 16A. There is currently no evidence of a hydraulic connection between the two reservoirs.

Who Pays and Who Runs It

FORGE is a DOE program managed by the University of Utah's Energy & Geoscience Institute. The money has come in tranches:

• April 27, 2015: five Phase 1 projects share $2 million; DOE plans up to $31 million for the first two phases, including up to $29 million for Phase 2.
• June 14, 2018: Utah wins the final competition; DOE announces up to $140 million in continued funding over five years.
• February 24, 2021: 17 Utah FORGE R&D projects eligible for up to $46 million; a second solicitation later offered up to $44 million for reservoir characterization, creation, and sustainability research.
• October 1, 2024: an agreement takes effect extending Utah FORGE through 2028 with an additional $80 million over four years.

By July 2026, DOE reported that Utah FORGE had supported $90 million in research and technology testing. The current agreement runs through 2028.

DOE's FY 2027 budget request defers funding for a further Utah FORGE extension until FY 2028, after the planned long-term circulation tests and their analysis have concluded. The government wants to see what the current test says before deciding whether to keep paying.

What FORGE Has Given the Industry

The product of a laboratory is data, and FORGE's is unusually public. As of May 2026, DOE reported more than 133 terabytes of FORGE technical data available through the Geothermal Data Repository: end-of-well reports, stimulation records, seismic catalogs, the traffic-light protocol and the raw circulation-test archives, all downloadable by anyone, including the companies FORGE is meant to help.

The clearest beneficiary is next door. Fervo Energy was selected for DOE-funded research at FORGE on multistage hydraulic stimulation, and DOE says Fervo reported benefiting from FORGE research and data when selecting its adjacent Cape Station development. Fervo CEO Tim Latimer put it this way at the Cape Station groundbreaking: "Thanks to cutting edge research and data collection from FORGE, Fervo can accelerate the production of the region's geothermal resources." Fervo Energy is applying the same techniques at its commercial Cape Station project.

Fervo is not the only company chasing hot dry rock. Quaise Energy and XGS Energy are pursuing their own approaches. Neither company has a documented connection to FORGE; they are parallel efforts rather than FORGE spinoffs.

What Is Still Unproven

FORGE has shown that you can drill two deviated wells into hot granite of the kind that gave Acord-1 nothing in 1979, fracture the rock between them, and run water through the gap for nearly a month — injecting at a steady 420 gallons a minute and getting most of it back at around 370°F.

What it has not shown, in DOE's own August 2026 framing: commercially durable heat extraction, acceptable long-term water loss, long-term thermal performance, or commercial economics. The principal investigator says thermal breakthrough and water loss over time remain unknown. Those are the questions the test running right now was designed to answer, and the next funding decision is parked until it does.

As of September 2026, the reservoir has circulated water for weeks. Whether it can do so for months is what the current test is scheduled to establish.