In the summer of 1904, in a valley in southern Tuscany where the ground had hissed and steamed for as long as anyone could remember, an Italian prince named Piero Ginori Conti hooked a small generator to a pipe of natural steam and lit a handful of light bulbs. Within a decade the valley was feeding electricity to the towns around it, and apart from a pause during the Second World War, it has kept doing so ever since. Yet for most of the century that followed, geothermal power remained what it had been at Larderello, a gift of rare geography. It worked only where heat, water and broken rock happened to meet close to the surface, and even after California opened the first American plant at The Geysers in 1960, the handful that followed never supplied more than a sliver of the nation’s power.
It was never a shortage of heat. Deep inside the planet, uranium, thorium and potassium have been slowly decaying for billions of years, and about half of the energy flowing out of the Earth comes from them, a nuclear reactor of the planet’s own that runs in every season and every weather. Drill deep enough almost anywhere and the rock grows hot. The trouble was that hot rock alone was useless; without water moving through it, and paths for that water to follow, the heat simply stayed where it was.
In the early 1970s, scientists from Los Alamos set out to change that at a place called Fenton Hill, in the Jemez Mountains of northern New Mexico. They drilled into hot, dry granite, cracked it open with water, and circulated water down one well and up another, carrying heat out of rock that had never held a drop. The physics held. The economics did not. The wells were slow and costly, the money grew harder to find once oil turned cheap again, and in 1995, after two decades of work, the project was shut down, filed away as an idea that had arrived before its tools.
The tools were being built a few hundred miles to the east, by people with no interest in geothermal at all. For close to twenty years a Texas oilman named George Mitchell kept spending money on the Barnett Shale outside Fort Worth, rock most of the industry considered worthless, until his engineers found a cheaper way to crack it open and others learned to pair that method with wells that turn and run sideways for thousands of feet. American oil output went from about 5 million barrels a day in 2008 to a record 12.9 million in 2023, and a generation of drillers learned to do in weeks what had once taken months.
Those were exactly the tools Fenton Hill had been missing. In 2023, in the high desert of northern Nevada, a pilot built on horizontal wells and fractures opened in stages began making electricity from hot rock, and it held steady temperatures for well over a year. Drilling times fell sharply, some of the largest technology companies in the country signed on to buy the power, and the Department of Energy began to describe a future in which next-generation geothermal supplies 90 gigawatts or more by 2050, more than twenty times what geothermal supplies today.
It arrived just as the country ran short of power. In 2025 the United States generated more electricity than in any year in its history, and still the typical new power project took just over five years to connect to the grid. Much of the new demand comes from artificial intelligence, which is, beneath its software, an enormous physical industry taking two shapes. Training campuses, where hundreds of thousands of chips on a single site teach a model over months, can now draw a gigawatt or more each, and like the aluminum smelters that once followed cheap hydropower up the Columbia River, they have begun to move to the power rather than wait for it. Inference sites, where those models go to work, are smaller and far more numerous, drawing tens of megawatts each near the cities they serve. Both need electricity that does not pause for night or weather, and a geothermal plant can serve either one from the land right beside it.
Where that happens matters as much as whether it happens. Built in the wrong place, a power plant or a data center can take the water a valley depends on or crowd a town that never asked for it. Built in the right place, the same projects bring what a rural county rarely sees all at once: hundreds of millions of dollars of investment, decades of skilled jobs, a tax base for schools and roads, and electricity for whatever the town wants to build next. A geothermal plant burns nothing and asks for remarkably little land, and we treat that land as though we will be here in a hundred years, because we will.
Terrak was founded to build in those right places. We have built our own AI to read the ground, the rock, the rights and everything in between, so that we can find the best land to build on, where the heat runs close, the power has a path to the people who need it, and the town nearby will be glad to have us. There we drill with the tools the shale fields perfected, and then we build the plants and run them for decades, because owning the whole of it is the surest way we know to move quickly and keep our word. Our goal is power that never switches off and, in time, costs less than any other source.
The work takes a particular sort of person: people who have logged wells or mapped rock, who know how land and mineral rights truly change hands, or who have drilled horizontal wells in shale and wondered where those skills might go next. It is early and it is hard, and it will be the best work of our lives.
Our mission is to turn America’s hot rock into terawatts of always-on power, and then to do the same for the world.
If this future excites you, reach out to us at hello@terrakenergy.com