The Third Way to Power AI

The real constraint on AI is no longer talent or chips. It's power. The IEA expects electricity demand from data centers to roughly double by 2030 (opens in new tab), with AI as the single biggest driver. Goldman Sachs puts the increase in data-center power demand at as much as 165% (opens in new tab) over the same stretch. The best models in the world are worth nothing if you can't plug them in.

And the obvious places to find more power are running out of room. On land, new power and data-center projects run into slow permitting and rising concern from local communities, and even approved projects are getting pulled back or scaled down by state and local governments.

So, some of the best-capitalized people in the industry are looking up to space, where orbital data centers have become a serious proposal. The trouble there is the economics: putting data centers on satellites that ride rockets into orbit is enormously expensive and risky.

There is a third way, and it runs along the bottom of the ocean.

Ocean Data Centers

Endurance Energy is building power on the seafloor. Half a kilometer below the ocean floor, near the ridges where the Earth's tectonic plates pull apart, sits geothermal fluid at around 400°C. The water on top of it is 2°C. That is the best thermal gradient anywhere on the planet, and a thermal gradient is the one thing you need to make electricity. The Earth does the heating. The ocean does the cooling. Endurance builds the machine that harnesses the energy.

This is the next move in a very old story. Every era reaches for a resource harder to get to and more abundant than the last. Whale oil gave way to petroleum. Petroleum pushed from onshore gushers into deep water, and then into shale. The seafloor is the next frontier, and it is the largest one yet.

The mid-ocean ridge system runs 65,000 kilometers through every ocean basin (opens in new tab), and it's where roughly three-quarters of the Earth's internal heat escapes. Endurance is starting in the Pacific's Ring of Fire, where the heat sits shallowest and island grids pay the most punishing prices for imported diesel. But the ridge circles the entire globe, most of it in international waters, free for anyone to tap.

An Engineering Problem, Not a Science Problem

Our conviction grew stronger once we saw that Endurance is solving engineering problems, not science problems.

We already know how to drill into the seabed. Oil companies go five kilometers down looking for oil they aren't even sure is there. Andrew Redd (opens in new tab), Endurance's founder and CEO, needs just half a kilometer to reach heat we know is there. The components to do this already exist in adjacent industries: the drilling, the turbines, the subsea cabling. The novelty is in the integration — putting it all together into a single unit you can lower onto the seabed and swap out when something breaks.

Built to Be the Lowest-Cost Producer

Right now, every energy project is getting funded thanks to the power demands of AI. The energy shortage is acute enough that capital will back almost anything that produces an electron. That won't last. When the rush settles, the projects still standing will be the ones with the lowest cost of production, just as the oil wells that survive a price crash are those that are cheapest to run.

Geothermal is already proving this on land. Fervo Energy (opens in new tab) signed the largest geothermal power purchase agreement ever and counts Google among its backers and customers (opens in new tab). It showed two things at once: first, that a SpaceX-style engineering culture can crack a resource the energy industry had written off; and second, that hyperscalers will sign up for firm, carbon-free power the moment a team can deliver it.

On a levelized-cost basis (opens in new tab), geothermal already ranks among the cheaper sources of around-the-clock power. Endurance runs on the same economics with a better starting point: the Earth's heat is free and the ocean's cooling is too, which pushes the marginal cost of the electron below almost anything else on the market. Endurance doesn't only win when demand is high. It survives when demand falls.

The Team and the Talent Window

The other reason we're backing Endurance is the team, and a hiring window that just opened.

For the last twenty years, if you were a brilliant aerospace or mechanical engineer, the place to do the hardest work of your career was one of Elon's companies. SpaceX trained a generation of people to solve brutal physical-engineering problems. As SpaceX moves toward the public markets, those people are, for the first time, available for problems that have nothing to do with rockets. Andrew is one of them, and he has attracted a remarkable number of others to his mission. When I met Andrew, what stood out was the clarity. He took every question I had down to the physics and back up to the spreadsheet without losing the thread.

Endurance extends our Global Resilience & Energy (opens in new tab) thesis: that the AI era is, underneath everything, an energy story. The companies that solve power generation will be among the most valuable of the coming decade. We're glad to be investors in Endurance, and we're early on purpose. The hard parts are still ahead, and we'll support Andrew and his team at every turn.

Energy has always belonged to whoever reaches the next frontier first. This one is half a kilometer under the seafloor, and Endurance is going down to get it.