Company · Climate · AI
Zanskar
The geothermal company teaching machines to read the heat the surface hides.
The geothermal company teaching machines to read the heat the surface hides.
Somewhere in western Nevada there is a patch of ground that looks like every other patch of ground. No steam. No hot springs. No bubbling mud. Nothing that would tell a passing geologist there is a 250-degree reservoir sitting 2,700 feet below their boots. In the summer of 2025, Zanskar drilled there anyway - because a model told them to. The model was right.
That site, which Zanskar calls "Big Blind," is the first blind geothermal system the U.S. industry has confirmed as commercial in more than three decades. It is also the clearest evidence yet for the company's whole argument: that the cleanest, most reliable power on the planet has been hiding in plain sight, and the only thing missing was a better way to look.
Four numbers that would have sounded delusional in a 2021 pitch deck. They are now in press releases.
Geothermal energy has a flattering reputation and an unflattering track record. It is carbon-free, runs 24 hours a day, and does not care whether the wind blows or the sun shines. On paper, it is the baseload power source climate planners dream about. In practice, it has spent decades as the renewable that everyone admires and nobody scales.
The reason is unglamorous: finding it is a gamble. Traditional exploration leans on surface clues - hot springs, fumaroles, altered rock - and then bets millions on a drill bit hoping the heat and the permeability both show up at the same depth. Drill a dry hole and you have spent a fortune on an expensive vertical disappointment. Those "soft costs" and that dry-hole risk, not the engineering, are what kept the industry small.
Zanskar's founders looked at the federal resource estimates everyone quoted and noticed something awkward: they were built on studies more than 20 years old. If the maps were that stale, the conclusion drawn from them - that hydrothermal energy was a niche - might be stale too. Maybe geothermal wasn't scarce. Maybe it was just badly searched for.
Carl Hoiland and Joel Edwards are not energy executives who discovered data. They are geoscientists - Hoiland did a PhD in geological sciences at Stanford, was an NSF graduate fellow, then an Activate fellow at Lawrence Berkeley National Lab - who decided the smartest place to apply machine learning was the rock beneath everyone's feet. In 2021 they founded Zanskar in Salt Lake City, naming it after a Himalayan range where expedition and geology collide. The bet was simple to state and hard to prove: that the Department of Energy had undercounted hydrothermal potential by an order of magnitude, and that the resources nobody could see were the ones worth chasing. If you could mine decades of geological data with AI instead of squinting at the surface, you could find the systems that leave no calling card - the blind ones - and you could de-risk them before the expensive drilling began. That is the entire premise. Everything Zanskar has built since is an attempt to turn that guess into a balance sheet.
Carl Hoiland and Joel Edwards launch in Salt Lake City to find geothermal with AI instead of surface guesswork.
The company triples output at an existing New Mexico plant - without drilling a single new well.
Two intermediate-depth wells hit a ~250°F permeable reservoir at roughly 2,700 feet.
First blind geothermal system confirmed commercial by U.S. industry in 30+ years. 100+ MW potential.
Led by Spring Lane Capital. Total equity funding reaches $180M. Six plants enter the pipeline.
A development capital facility (scalable to $100M) co-led by Just Climate funds the first greenfield plants.
What Zanskar sells, at its core, is better aim. Its prospecting toolkit ingests enormous volumes of geological data and flags anomalies that signal exceptionally high heat flow. Then its AI-prediction engine sites the actual well targets - where to drill to test temperature and permeability - so that geologists are pointing the bit at the highest-probability spot rather than the most obvious one. That is how a site with zero surface signs becomes a confirmed reservoir.
Finds high-heat-flow anomalies in raw geological data, including blind systems with no surface clues, and sites viable well targets.
Stochastic resource modeling cuts dry-hole risk on greenfield and known sites - which lowers the cost of capital.
Six ~20 MW binary-cycle plants in the pipeline, sized to win faster interconnection in Western markets.
Squeezes more from existing assets - the Lightning Dock plant's output tripled without new wells.
Note the deliberately boring 20-megawatt unit size. It is a permitting hack dressed up as an engineering choice, and it is one of the cleverer things in the company.
Plenty of climate startups have a thesis. Fewer have a hole in the ground that confirms it and a syndicate of investors willing to underwrite the next one. Zanskar's $115M Series C - the largest venture round to date into AI-enabled geothermal discovery - drew Spring Lane Capital, Obvious Ventures, Union Square Ventures, Lowercarbon Capital, Munich Re Ventures and more than a dozen others. The round was oversubscribed and capped.
Bars scaled for readability, not to a single axis - dollars and megawatts refuse to share a ruler. Figures from company announcements, Jan-Apr 2026.
The validation is not only financial. Tripling Lightning Dock proved the software earns its keep on assets that already exist. Big Blind proved it can conjure new ones. And a $40M development facility co-led by Just Climate signals the part that matters most for a power company: capital that funds steel in the ground, not just slides in a deck.
Zanskar's stated goal is almost aggressively unromantic: to make geothermal "the most affordable form of widely-deployable, 24/7, carbon-free generation on the planet." Not the most exciting. The most affordable. The company is betting that the climate transition will be won by whatever clean power source becomes cheapest and most reliable, and that an always-on resource with no fuel cost has an unfair advantage - if only someone can stop wasting money finding it.
The competition is real and growing. Next-generation players like Fervo Energy, Sage Geosystems and Eavor are attacking the same opportunity from the engineering side, drilling enhanced systems where the heat is known. Zanskar's wager is different: that the bigger prize is in discovery - in the terawatt of conventional geothermal it argues the industry simply overlooked.
Data centers are hungry. Grids are straining. The demand for power that is both clean and constant has rarely been louder, and most of the obvious answers come with an asterisk about weather or storage. Geothermal does not. If Zanskar is even partly right that the resource is ten times larger than the textbooks say, the implication is not a better startup - it is a different energy map.
Which brings us back to that unremarkable patch of Nevada. A year ago it was empty desert with a secret. Soon it is meant to host the nation's first new conventional geothermal plant on a previously undeveloped site in nearly a decade - 100-plus megawatts of carbon-free power pulled from rock that gave away nothing at the surface. The ground did not change. The way we looked at it did.