How geothermal gets built

The first boom died of exploration risk. US geothermal power starts in the 1960s at the Geysers field in northern California, scales faster than any other renewable of its era, reaches gigawatts of capacity through the 1980s, then essentially stops for two decades. A modest revival in the late 2000s and early 2010s added hundreds of megawatts, much of it offsetting declines at the older steam fields rather than adding net capacity. Two things killed it. The early technology could only use very high temperature steam, around 200 Celsius and often hotter, and such places are rare; resources get more abundant the further down the temperature scale you go, which is where the technology could not follow. The bigger problem was that developers kept drilling into what looked like sure things and coming up empty.

Why they came up empty is the counterintuitive part. The planet gets hotter with depth everywhere, typically around 25 Celsius per kilometer, so an ordinary site needs four or five kilometers of drilling to reach steam temperatures. What makes a commercial resource is either volcanic heat brought close to the surface or, across much of the western United States, fractures and permeable zones that convect hot water upward from depth. Hot springs are one visible manifestation of that, but they are the outliers. Most convective systems are blind: no spring, no volcano, nothing at the surface, and historically they were found by drilling into them by accident. So the industry’s founding heuristic, drill where it steams, was selecting on a weak signal. The Geysers happened to be the case where the surface sign sat on a world-class resource, and Hoiland thinks it unlikely another gigawatt-scale conventional field of that type remains undiscovered in the US. Oil companies including Chevron, Unocal, Phillips and Hunt spent hundreds of millions in the late 1970s and early 1980s drilling test holes every few miles looking for another Geysers, never found one, and in the process stumbled onto smaller systems only now being developed. That failure split the industry’s response, much of it funded by the Department of Energy, into two paths: get better at finding the blind systems, or stop depending on finding them by engineering the missing ingredients in place.

That distinction organizes everything after it. A conventional, or hydrothermal, resource needs three things naturally present: enough heat to boil the working fluid, permeability so fluid can circulate through the rock, and water to do the circulating. Enhanced geothermal systems came out of the recognition that plenty of wells were hot but had no permeability or no water, so the rock can be stimulated to create flow paths and the water added. Hoiland’s analogy is conventional versus unconventional oil and gas: whether you can drill a well and have what you need, or must modify the subsurface. Either way the development sequence is the same ladder of narrowing uncertainty. Cheap shallow geophysics picks the targets. Temperature gradient holes, a few hundred to a thousand feet at tens to hundreds of thousands of dollars, confirm heat is present. A conceptual model of what drives the system, volcanic, sedimentary or fault-hosted, guides where to go deeper. Then slim wells at a few thousand feet and one to four million dollars each test whether the resource will actually flow, followed by a flow test long enough to show the flow holds up. Each step costs more and tightens the distribution of outcomes, and the classic historical failure was skipping the slim well entirely and putting a five, ten or fifteen million dollar production well into rock that turned out to be hot but dry.

Even a flowing well does not finish the job, because what makes a project financeable is the decline rate, and that cannot be measured until the injection well exists. Extraction cools the water, and the reinjected water has to pick the heat back up from the rock before it returns. Spacing the wells is the art: too far apart and the water may never come back while reservoir pressure drains away, too close and it returns before it has reheated. Only with both wells running can you build the reservoir model that says what a second or third well pair would do to the decline. That data-then-decide rhythm is one of two reasons projects historically took over five years and sometimes ten. The other was permitting: five separate National Environmental Policy Act reviews on federal land. Exploration activities have recently been granted a categorical exclusion, meaning that class of work no longer needs its own environmental review, with reform of the construction stage only potentially ahead. Stripped to the bare bones, Hoiland gets one to two years to explore and confirm plus one and a half to two to build and interconnect, so three to four years total, which he says is now realistic in certain states. What does not loosen is geography. Near-term technology, conventional and enhanced alike, still needs tectonically active or high heat flow ground, about a third of most continental land masses, because drilling is the primary cost driver and shallower heat is cheaper heat. Cheaper drilling, or strong enough demand for clean firm power, would widen that, on a timeline he puts on the order of decades.