The ground beneath our feet holds so much energy that experts at the Department of Energy think geothermal power could generate 60 gigawatts — or nearly 10% of U.S. electricity — by 2050.
Zanskar co-founder and CEO Carl Holland thinks that lofty number is too low, mostly because it’s discounting conventional geothermal’s potential.
The DOE’s figures assume advances in enhanced geothermal, which uses fracking techniques to access hot rock deep underground. Companies like Fervo and Sage Geosystems have been pursuing that angle, and experts agree it has tremendous potential. By contrast, conventional geothermal, which taps naturally fractured hotspots, has been stagnant, generating just 4 gigawatts in the United States, up only about a gigawatt in the last decade or so.
Conventional geothermal has been held back by outdated assumptions, Holland told News.
“They underestimated how many undiscovered systems there are, maybe by an order of magnitude or more,” Holland said. With modern drilling techniques, “you can get a lot more out of each of them, maybe even an order of magnitude or more from each of those. All of a sudden the number goes from tens of gigawatts to what could be a terawatt-scale opportunity.”
Zanskar is relying on AI to help break conventional geothermal out of its malaise. In the process, the startup has resuscitated a flagging power plant in New Mexico and discovered two new sites with over 100 megawatts of combined potential.
Those successes have helped Zanskar land a $115 million Series C led by Spring Lane Capital with participation from the All Aboard Fund, Carica Sustainable Investments, Clearvision Ventures, Cross Creek, GVP Climate, Imperative Ventures, Lowercarbon Capital, Munich Re Ventures, Obvious Ventures, Orion Industrial Ventures, Safar Partners, StepStone Group, Susquehanna Sustainable Investments, Tranquillion, Union Square Ventures, University Growth Fund, and UP.Partners.
Techcrunch event
San Francisco
|
October 13-15, 2026
Many potential geothermal sites have been overlooked because people have been looking for evidence on the surface like a hot spring or a volcano, Holland said. But about 95% of all geothermal systems lack such a tell, he added. “We just keep finding them by accident. That is a great application for AI.”
To find new geothermal resources, Zanskar first feeds supervised machine learning models a range of data, including past accidental discoveries. When they have identified a promising site, the company sends a team to the site validate the discovery.
Zanskar then has to come up with a plan to develop it. For that, the team uses another AI approach, one developed for this very purpose called Bayesian evidential learning (BEL). In BEL, existing data helps build a series of assumptions known as priors, and then models work to falsify those hypotheses, spitting out probabilities for each. Where there are gaps, the startup has built a geothermal simulator to fill in the blanks.
So far, Zanskar’s approach has been working. The previous funding round allowed the startup to explore three sites, each of which it considered a success. “Three of three,” said Zanskar CTO Joel Edwards. “What does it look like when you try 10?”
The company has enough sites in the pipeline to support at least a gigawatt of generating capacity, Holland said. For now, they’ve been focused on the U.S. West, which has the most potential. He would like to find at least 10 confirmed sites to help Zanskar court project finance investors, which provide access to lower cost capital than VCs. If the company can make that work, then it might get through the valley of death that has claimed many other climate tech startups.
Holland is not convinced that Zanskar has solved all the challenges associated with finding geothermal resources, but he’s optimistic the company is on the right path.
“We now know this is the future of exploration,” he said. “This is going to change geothermal in very short order.”
