A $115 million raise for a geothermal startup doesn’t sound like the kind of thing that should matter to the AI boom. It’s not a flashy model launch or a new chip announcement. But Zanskar’s latest funding round is a reminder of something the tech world is starting to bump into more often: AI’s future isn’t only limited by algorithms — it’s limited by electricity.
And not just any electricity. Training and running large models takes a huge amount of power, and data centers don’t just need it in bursts. They need steady, around-the-clock energy that doesn’t disappear when the wind dies down or the sun sets. That’s where geothermal starts to look a lot more relevant — it’s one of the few clean sources that can deliver reliable “always-on” power.
Zanskar is betting that the geothermal industry has been stuck in the past, not because the resources aren’t there, but because we’ve been looking for them the wrong way. The company’s core argument is bold: conventional geothermal — the kind that taps naturally occurring hot rock and fractures underground — could be far bigger than most experts assume. Zanskar CEO Carl Holland says we may be underestimating the opportunity by an order of magnitude or more, and that with better tools, it could go from “tens of gigawatts” to something approaching a terawatt-scale category.
That’s a huge claim, but the reason investors are taking it seriously is that Zanskar has already shown early proof. In its last round, the company explored three different geothermal locations and hit on all three. One project helped bring a struggling power plant back to life in New Mexico, and the other two were entirely new discoveries — together totaling more than 100 megawatts of potential generation. As CTO Joel Edwards put it: “Three of three.” The obvious question now is what happens when they run the same playbook ten times.
The “playbook” here is really two layers of AI. First, Zanskar uses machine learning trained on historical geothermal finds — including the messy reality that a lot of geothermal systems were discovered accidentally. In fact, the company says roughly 95% of geothermal systems don’t have obvious surface clues like hot springs or volcanoes, which helps explain why conventional geothermal has stayed stuck at around 4 gigawatts in the U.S. for years. Their approach is essentially: stop waiting to stumble into geothermal and start systematically hunting it.
Once Zanskar identifies a promising site, they validate it on the ground and move into development using more advanced modeling tools, including a technique called Bayesian evidential learning. The goal is to reduce uncertainty, pressure-test the team’s assumptions, and narrow in on the most likely outcomes before committing serious capital to drilling and buildout.
The $115 million Series C — led by Spring Lane Capital, with backers ranging from Lowercarbon and Union Square Ventures to Munich Re Ventures — gives Zanskar the runway to scale that approach. The company’s next milestone isn’t just finding more heat underground. It’s building enough confirmed capacity to unlock project finance — the kind of long-term infrastructure capital that can take geothermal from “startup story” to “real grid asset.” That’s also how climate companies escape the valley where a lot of pilots die: they prove the model works repeatedly, then bring in bigger money to build at scale.
If Zanskar is right, this could be a turning point for geothermal. For years, conventional geothermal has been treated like a mature, slow-growth corner of the energy market. Zanskar is arguing it’s actually an underexplored opportunity hiding in plain sight — especially across the U.S. West — and that AI is the tool that finally makes it searchable, repeatable, and investable.
And if they pull it off, the impact goes beyond geothermal. It’s a blueprint for how AI doesn’t just consume power — it can help unlock more of it.
This analysis is based on reporting from techbuzz.
Image courtesy of Unsplash.
This article was generated with AI assistance and reviewed for accuracy and quality.
