What is geothermal energy?.
- The article does not describe geothermal, as a category, as near-zero emissions: lifecycle medians sit far below coal and gas, but a 2002 worldwide survey put flash plants' average operational venting at 122 gCO₂ per kWh — the near-zero claim belongs to binary and closed-loop configurations, and is scoped that way.
- On induced seismicity it names the two canonical failures, Basel and Pohang, alongside the managed record and its control protocols: the honest formulation is that the risk is manageable through protocols — not that it is predictable.
- Geothermal's structural argument is that it runs: on the EIA's final 2024 figures, US geothermal plants ran at a 64.6% capacity factor, against 34.3% for wind and 23.2% for utility-scale solar photovoltaics.
- Catoxy Energy is being built to develop, operate and optimise integrated geothermal projects on exactly this basis — and applies the same discipline to describing the technology as it would to developing it.
Geothermal energy is heat drawn from the rock beneath us. It is one of the few renewable resources that does not depend on the weather, the season or the time of day: the temperature two kilometres under a site is the same at midnight in February as at noon in July. Nearly everything worth knowing about the technology — its economics, its geography, its honest limitations — follows from that one fact and from the physics of where the heat comes from.
This is the primer we would want a sceptical reader to have. It covers the physics, the three system families, what the temperature bands actually mean, and straight answers on the three questions any serious assessment asks: is it clean, is it safe, and does it run out.
Where the heat comes from
Earth sheds 46 ± 3 terawatts of heat continuously. That is the field's current consensus estimate, carried in the most recent analyses of Earth's heat budget, and the heat has two sources: primordial heat left over from the planet's formation, and radiogenic heat from the ongoing decay of potassium, thorium and uranium in the mantle and crust. Geoneutrino measurements — detectors in Japan and Italy counting the particles that radioactive decay emits — put the radiogenic share at roughly half, with recent central estimates around 40–50%; the potassium contribution within that is still inferred from models rather than measured directly. The planet is, in a precise sense, a slowly discharging battery with a radioactive trickle-charger.
Two details matter for anyone assessing the resource. First, the three decaying elements supply almost all radiogenic heat in a roughly 2:2:1 ratio, and a reference-model estimate puts total radiogenic power at about 20.1 TW, of which the continental crust alone contributes some 6.8 TW — a third of the radiogenic output sits concentrated in the thin layer directly beneath our feet. Second, the heat is not evenly distributed. Where it concentrates — and how easily it can be reached — is a question of geology, which is why the industry describes resources in terms of geological play types: volcanic and plate-margin settings where hot fluid convects naturally, and the conduction-dominated sedimentary basins and rift systems that cover much of continental Europe.
What a geothermal system needs
A conventional geothermal system needs three things at once: heat, water and permeable rock. Where all three occur together naturally — mostly along tectonic plate boundaries — the result is a hydrothermal resource: drill, produce hot fluid, use it, reinject. This is the mature form of the technology, and it is geographically restricted. Most of the world's landmass has the heat but not the natural plumbing.
Engineering supplies two answers to that. An enhanced geothermal system (EGS) injects fluid deep underground under carefully controlled conditions to create new fractures and re-open existing ones, building a human-made reservoir in rock that was too tight to flow, and then circulates fluid between injection and production wells. The US Department of Energy estimates EGS-type resources exist in most, if not all, US states — but the technology's maturity deserves an honest label. The National Renewable Energy Laboratory's 2024 Annual Technology Baseline classifies EGS as nascent, with no commercial EGS plant operating or under construction as of its 2022 base year — and that snapshot is now dated in ways worth stating precisely. Fervo Energy's Project Red in Nevada — in the company's own words a limited-scope, proof-of-concept commercial pilot — has delivered power and revenue since 2023, and its first commercial-scale development, Cape Station, has 500 MW under construction with the first ~100 MW expected operational in early 2027, backed by 658 MW of binding offtake (all company-reported and, for Cape, forward-looking). The long-running research site at Soultz-sous-Forêts in the Upper Rhine Graben proved the concept decades earlier; a commercial industry is now being assembled, young enough that every number in it deserves its label.
A closed-loop system takes the opposite bet. Instead of engineering a reservoir, it keeps the working fluid — water or supercritical CO₂ — inside a sealed wellbore by design, in a U-shaped loop or a coaxial pipe-in-pipe arrangement, with conduction-dominated heat exchange through the surrounding rock. The Department of Energy's analogy is a radiator. The design removes the dependence on formation fluid and natural permeability, and pays for it in physics: with no fracture network to sweep heat from, output is limited by the small contact area of the pipe itself. NREL's modelling is blunt — a closed-loop system with only a few kilometres of borehole is limited to electricity output in the tens to hundreds of kilowatts even in 500 °C rock, and multi-megawatt electric output requires tens of kilometres of downhole completion. A modelled coaxial retrofit of a real 4.2 km well at Mol in Belgium sustains roughly 0.5 MW of thermal output over twenty years — a simulation of a real well, not measured twenty-year operation, and a number that does not generalise to multilateral designs. In every one of NREL's forty modelled closed-loop cases, using the heat directly beats converting it to electricity. The field moved to partial commercial operation for the first time in December 2025, when Eavor's project at Geretsried in Germany began delivering power to the German grid. Claims that closed-loop makes geothermal viable anywhere are, in NREL's careful phrasing, what proponents highlight — not an established capability. Where natural permeability is good, open systems far out-produce closed ones; the two converge only in tight rock. Closed-loop is the geology-insensitive option, not the higher-output one.
Heat, water and permeable rock occur together; one well produces, one reinjects into the same formation.
Fluid injected under controlled conditions re-opens and creates fractures, making a reservoir where nature left the rock tight.
Working fluid stays in the pipe by design — U-loop or coaxial — trading reservoir risk for a hard limit on heat transfer area.
Three schematic panels. First, a hydrothermal doublet: production and injection wells into a permeable aquifer. Second, an enhanced geothermal system: two wells connected through engineered fractures in hot, tight rock. Third, a closed loop: a single sealed U-shaped wellbore with no fluid exchange with the rock, shown alongside a coaxial pipe-in-pipe variant.
What the temperature bands mean
There is no internationally standardised temperature classification for geothermal resources. Reviews of the field are explicit that the published schemes draw their category boundaries in different places, and one widely used teaching classification — from the Geo-Heat Center at the Oregon Institute of Technology — simply calls resources low below 90 °C, moderate from 90–150 °C and high above 150 °C. What exists instead of a standard is a set of working bands used by the US agencies, and they are worth knowing precisely.
Direct use — heating things rather than generating electricity — spans roughly 20–150 °C. For power, the 2024 Annual Technology Baseline draws the line at 200 °C: binary-cycle plants, which pass the geofluid through a heat exchanger to drive a secondary working fluid, generally serve resources below 200 °C, while flash plants, which boil the geofluid itself into steam, serve resources at or above it. That 200 °C line is a plant-selection convention, not a minimum temperature for power — binary cycles generate from roughly 100–110 °C. Depth follows temperature and geology rather than a fixed rule: the ATB's representative hydrothermal cases sit at 175 °C and 1.5 km for binary and 225 °C and 2.5 km for flash, with its deep-EGS cases at 3–3.5 km, and the frontier beyond — so-called superhot systems above 375 °C — deeper still.
Chart of depth versus temperature. Direct-use applications span 20 to 150 degrees Celsius; binary power plants serve resources below 200 degrees; flash plants serve 200 degrees and above. Representative cases: a binary plant at 175 degrees and 1.5 kilometres depth, a flash plant at 225 degrees and 2.5 kilometres, a deep EGS binary case at 175 degrees and 3 kilometres, and a deep EGS flash case at 250 degrees and 3.5 kilometres.
Europe's geography spans the whole range. The continent's conduction-dominated sedimentary basins — the Paris Basin, the North German Basin, the Molasse Basin south of Munich — host aquifer doublets producing hot water for district heating, while rift settings like the Upper Rhine Graben host the deeper, hotter projects, including Europe's pioneering EGS work.
What geothermal is not: heat pumps
The term "geothermal" gets applied to a technology that works on entirely different physics, and the conflation inflates statistics, so it is worth being precise. A ground-source heat pump exploits the fact that the upper few metres of the ground hold a nearly constant 10–15 °C year-round — stored solar warmth, not deep Earth heat — and uses electricity to move heat between that stable reservoir and a building, cutting heating and cooling electricity consumption by 25–50%. The Congressional Research Service classifies ground-source heat pumps as "an energy efficiency technology", not energy production, and the distinction matters when reading industry statistics: of the 70,329 MWt of worldwide geothermal direct-use capacity recorded in the 2015 world review, ground-source heat pumps were 70.9% of installed capacity and about 55% of the energy delivered. A global "geothermal direct use" total is, by capacity, roughly seven-tenths heat pumps — a fact any honest use of those totals should carry.
One resource, used in stages
We have written before about why one resource yielding several energy forms changes project economics; here it is enough to show the mechanism. Applications stack by the temperature they need — the classic Líndal presentation — and a single well stream can descend that ladder, doing electrical work first and heating work afterwards. The benchmark for what the lower rungs alone can carry is Iceland: about 90% of the energy used for domestic heating there comes from geothermal energy, and district heating utilities supply 95% of the population with hot water.
Vertical temperature scale from 30 to above 200 degrees Celsius. Power uses: flash steam at 200 degrees and above, binary cycles from roughly 100 to 200 degrees. Direct uses: industrial process heat from 100 to 150 degrees, district and space heating from 60 to 100 degrees, greenhouses and aquaculture from 30 to 60 degrees.
Is it clean? The lifecycle numbers
The defensible claim is specific, and it is worth making specifically.
On full lifecycle accounting — construction, operation, decommissioning — the major assessments agree. The IPCC's Fifth Assessment Report puts geothermal electricity at a median of 38 gCO₂eq/kWh (range 6.0–79) against 820 for coal and 490 for combined-cycle gas — roughly 95% below coal and 92% below gas, in the same band as utility-scale solar photovoltaics at 48, and above onshore wind at 11 and nuclear at 12. NREL's independent harmonisation, which screened roughly 3,000 lifecycle studies and accepted fewer than 15% of them, lands in the same place: geothermal at a median of 37, against coal at 1,001, gas at 486, solar PV at 43, and wind and nuclear at 13.
But the category median hides a technology split that honest copy has to surface. NREL's geothermal-specific review found hydrothermal binary plants at a median of 11.3 gCO₂eq/kWh, EGS binary at 32.0 — and hydrothermal flash at 47.0, with far more variability. The reason is structural: a flash plant is an open loop, and CO₂ that arrives dissolved in the geofluid vents to the atmosphere after the turbine. A 2002 worldwide survey of operating flash plants, reported in that review, put average operational CO₂ emissions at 122 g per kWh, with a range of 4–740 — the top of which overlaps gas-fired generation. The survey is dated and measures operational venting rather than full lifecycle emissions, but it remains the broadest census the field's own review literature works from. That is why we do not describe geothermal, as a category, as "near-zero emissions". The near-zero claim belongs to binary and closed-loop configurations, where the geofluid stays sealed and operational emissions are essentially nil. Technology choice, not branding, is what makes geothermal clean — which is precisely why it is an engineering decision to get right at design time.
Is it safe? Seismicity and water, honestly
Creating an EGS reservoir means raising fluid pressure in deep rock, and raising fluid pressure on a stressed fault can make it slip. This is not a hypothetical: the industry's two canonical failures are part of its public record. In Basel in 2006, stimulation of the crystalline basement beneath the city triggered an event of local magnitude (ML) 3.4; the project was halted immediately and later abandoned. In Pohang, South Korea, a moment magnitude (Mw) 5.5 earthquake struck in November 2017 with an epicentre about 510 metres from an EGS site, after five stimulation campaigns over the preceding two years — and about two months after the final one. The Pohang event exceeded the widely used bound relating injected volume to maximum magnitude, and its timing carries its own lesson: injection volume does not reliably cap event size near a critically stressed fault, and stopping injection does not instantly end the risk.
The same peer-reviewed record that names the failures also names the projects that managed the risk — Soultz-sous-Forêts, Helsinki, Blue Mountain in Nevada, Utah FORGE — operating under traffic-light protocols that tie injection to real-time seismic monitoring and stop work at pre-agreed thresholds, one layer of a discipline that begins with site screening and fault mapping and continues through post-injection monitoring. The honest formulation is that induced seismicity is manageable through protocols — not that it is predictable. Anyone who tells you the maximum magnitude of an induced event can be forecast with confidence is ahead of the science.
Closed-loop systems change this calculus by design: with no reservoir stimulation and no fluid injected into the formation, the mechanism behind Basel and Pohang is absent during operation. Scoped claims survive scrutiny, so two caveats belong here: a 2013 event of magnitude 3.5 at St. Gallen was triggered during well-control operations — the drilling phase carries its own residual risk regardless of system type — and 2025 modelling work argues that decades of cooling can, in principle, contract stiff rock enough to trigger microseismicity on critically stressed faults. Removed-by-design applies to the stimulation mechanism, not to every conceivable pathway.
On water, the comparison runs in geothermal's favour: EGS has relatively low water demands compared with other thermal power technologies, and can use non-potable water for stimulation and operations. And because the working parts of the system are kilometres underground, what occupies the land surface is a plant, wellheads and pipework rather than a fuel cycle.
Does it run out?
At planetary scale the resource is not meaningfully depletable — the 46 TW flux flows whether or not anyone taps it. The real question is local, and the honest answer has two halves.
Local thermal drawdown is real. The longest-running demonstration is also the most instructive: geothermal energy has supplied district heating networks in the Paris Basin for more than 40 years, from doublets in the Dogger aquifer — and the operators' own science is candid that reinjecting cooled brine leads to progressive cooling of the resource at the local doublet scale. What four decades of operation have produced is not a depleted basin but a management discipline: doublet spacing, reinjection design and lifetime forecasting that model the "thermal breakthrough" — the eventual arrival of cooled water at the production well — decades in advance. Sustainability, in geothermal, is an engineering property of the scheme, not a free attribute of the resource.
What a managed system delivers in return is the statistic that defines the technology. On the EIA's final 2024 figures, US geothermal plants ran at a 64.6% capacity factor, against 34.3% for wind and 23.2% for utility-scale solar photovoltaics. The Department of Energy separately notes that plants are typically available to generate 95% or more of the time — availability and capacity factor are different measures, and both belong in the picture, correctly labelled.
Why we wrote this down
Every claim in this article has a source, and the limits of each claim are stated next to it — which technologies the emissions numbers belong to, which magnitude scale a quoted event uses, which data year a capacity factor comes from. That is not caution for its own sake. Firm, dispatchable, low-carbon heat and power from a resource under the customer's feet is a strong enough proposition without embellishment, and the reader we care about — the engineer, the official, the analyst doing their diligence — is precisely the reader who checks.
Catoxy Energy is being built to develop, operate and optimise integrated geothermal projects on exactly this basis. We apply the same discipline to describing the technology as we would to developing it.
Sources: Sammon & McDonough, "Quantifying Earth's radiogenic heat budget", Earth and Planetary Science Letters 593:117684 (2022), for the 46 ± 3 TW surface heat loss; KamLAND Collaboration, "Partial radiogenic heat model for Earth revealed by geoneutrino measurements", Nature Geoscience 4 (2011); Huang et al., "A reference Earth model for the heat-producing elements and associated geoneutrino flux", Geochemistry, Geophysics, Geosystems 14 (2013); US Department of Energy, Enhanced Geothermal Systems and Tribal Energy Guide: Geothermal; Congressional Research Service, R47256, "Geothermal Energy: Frequently Asked Questions" (2022); NREL, Annual Technology Baseline 2024, Geothermal; Beckers et al., "Tabulated Database of Closed-Loop Geothermal Systems Performance", NREL/CP-5700-84979, OSTI 1961369 (2023); Fervo Energy, Form S-1 registration statement (March 2026), for Project Red, Cape Station and offtake status (company-reported); Eavor, first electricity production at Geretsried (December 2025); Leontidis et al., "Techno-economic modelling of a coaxial closed-loop geothermal system at Mol", Geothermal Energy 13 (2025); Jalilinasrabady, "Geothermal Resources Classification — A Review", GRC Transactions 45 (2021); Lund & Boyd, "Direct Utilization of Geothermal Energy 2015 Worldwide Review", Proceedings World Geothermal Congress 2015; Orkustofnun, District Heating in Iceland (2020 figures); IPCC, AR5 WG3 Annex III, Table A.III.2 (2014); NREL, "Life Cycle Greenhouse Gas Emissions from Electricity Generation: Update", NREL/FS-6A50-80580 (2021); Eberle et al., "Systematic Review of Life Cycle Greenhouse Gas Emissions from Geothermal Electricity", NREL/TP-6A20-68474 (2017), which reports the Bertani & Thain (2002) worldwide flash-plant survey; Yeo, Brown, Ge & Lee, "Causal mechanism of injection-induced earthquakes through the Mw 5.5 Pohang earthquake case study", Nature Communications 11:2614 (2020); Zhou et al., "Managing Induced Seismicity Risks From Enhanced Geothermal Systems: A Good Practice Guideline", Reviews of Geophysics 62 (2024); a 2025 closed-loop modelling study on thermal-contraction seismicity and closed-loop economics; Lopez et al., "40 years of Dogger aquifer management in Ile-de-France, Paris Basin, France", Geothermics 39 (2010); US EIA, Electric Power Annual, Table 4.8.B (final 2024 data).