Geothermal isn’t tapped out it has been hard to see. On a scrubby stretch of western Nevada desert, there are no steaming pools, no geysers, and no tourist-board hints that anything unusual is happening below. Yet a few thousand feet down sits a hot, permeable reservoir that a geothermal startup says it identified using artificial intelligence-an approach aimed squarely at the industry’s long-standing problem: promising resources that leave almost no surface clues.
Zanskar Geothermal & Minerals calls its Nevada target “Big Blind,” a name borrowed from the geoscience shorthand for “blind” hydrothermal systems-reservoirs with little to no surface expression and, often, no local history of geothermal prospecting. The company says it drilled to roughly 2,700 feet and encountered porous rock at 250°F, meeting basic thresholds for a utility-scale prospect. Zanskar’s CEO and co-founder, Carl Hoiland, described the search challenge as one that has shaped the sector’s story for decades: “The idea that geothermal is tapped out has been the narrative for decades.”
Conventional geothermal using naturally occurring hot water and permeability has always had a paradoxical value proposition. On the one hand, it can offer predictable, baseload power with zero direct emissions but only at locations where three stars happen to align: heat, fluid, and rocks that allow that fluid to move. Miss any one of those and developers often find themselves spending millions to learn, too late, that the subsurface isn’t cooperative. As Zanskar’s co-founder and CTO Joel Edwards put it, “It’s a ‘classic needle in the haystack problem.’”
The modern twist is not that the Earth changed, but that pattern-finding did. Zanskar’s models are trained on locations where blind systems are already known many discovered accidentally during decades of oil, gas, and minerals drilling. The company then stacks and cross-checks disparate signals-geology, geophysics, and remote sensing-looking for combinations that correlate with hidden reservoirs. As Hoiland puts it, AI has gotten really good over the last 10 years at being able to pull out those types of signals out of noise.
One reason this matters is scale: Nevada and its neighbors sit within the Great Basin, a vast region where heat flow is often favorable but surface indicators are inconsistent. In that context, academic estimates frequently highlight how much of the resource is effectively invisible: Hoiland has pointed to a landscape dotted with opportunities, and Nevada geoscientist James Faulds told CNN that more than three-quarters of US geothermal resources are blind. If that proportion is even directionally right, then the “known geothermal map” is less a full inventory than a collection of places that were easy to spot.
But by far the most interesting detail in Zanskar’s account, from an engineering perspective, has little to do with the architecture of the model. A regional scan identifies candidate zones; field teams gather higher-resolution data; only then does the company commit to deep confirmation drilling. That staged approach echoes how other high-cost subsurface industries reduce risk-thin out the search space early, spend money late. In a separate description of the process, Zanskar’s team highlighted that it used models trained on known hot spots and simulations, then narrowed prospects with targeted field campaigns before the deeper wells that confirmed Big Blind’s temperature and permeability.
The next bottlenecks are less glamorous. Turning Big Blind into a grid-connected project still requires permitting, interconnection, and financing, says Zanskar, with a first-power estimate of three to five years. That timeline is not uncommon for new generation, but geothermal often can face an added mismatch: exploration risk is front-loaded, while the grid often demands certainty. As one recent permitting overview put it, geothermal development often requires “numerous permits, authorizations, and other regulatory requirements” that affect “project timelines, costs, and risks,” especially in areas where federal lands and leasing occur via the Bureau of Land Management’s program.
Meanwhile, the geothermal conversation has had two parallel tracks. The buzz that is loudest is over next-generation methods chiefly EGS, which create or expand permeability in otherwise tight hot rock. The U.S. Department of Energy sums up EGS as a means to “human-made reservoir” by injecting fluid to open fractures and enable circulation where natural permeability is insufficient. Zanskar’s pitch runs opposite: ample conventional resources remain, many of them “blind,” and the industry’s poor historic hit rate discouraged systematic exploration.
The stakes transcend one Nevada lease. Electricity demand is growing, and data centers are among the most discussed new loads, partly because they demand reliable, around-the-clock supply. Baseload resources those that can run regardless of weather are relatively scarce in the clean-energy toolkit. That is why geothermal’s small present-day footprint in the United States is so striking: even though the United States leads the world in installed geothermal capacity, geothermal is still about 0.4% of the U.S. electricity mix, according to figures cited in the main reporting. The constraint has not been a lack of heat; it has been the difficulty and cost of repeatedly turning subsurface uncertainty into bankable projects.
Zanskar’s approach puts a new emphasis on discovery as an engineering discipline: how quickly a developer can move from broad possibility to drillable confidence. If the company can repeat its result across numerous sites Hoiland has said the team has identified “hotspots” with Big Blind-like signatures the biggest shift may be psychological as much as technical. Investors and utilities do not need geothermal to be limitless; they need it to be repeatable.
For now, Big Blind reads as a proof of method: a reservoir with 250°F rock at 2,700 feet, found without surface signposts, validated by drilling, and positioned as a conventional geothermal project, rather than some next-gen experiment. The broader implication is that the map of geothermal potential in the American West may be less complete than it looks and that seeing what is already there could be as consequential as inventing entirely new ways to make heat usable.
