On September 27, 2026, a kilometer and a half beneath the surface of the Pacific Ocean, on the slope of an active undersea volcano three hundred miles off the coast of Oregon, a turbine the size of a dinner plate spun up to nearly twenty-six thousand revolutions per minute. For twenty-nine seconds it generated electricity, and for eight point seven seconds the plant produced more energy than it consumed itself. Sixty watt-hours in twenty-nine seconds. Median net power — around two kilowatts, peak — sixteen and a half. One electric kettle. Less than nine seconds of kettle time.
Andrew Redd, founder and CEO of the Seattle startup Endurance Energy, put it this way: “Goddard’s first liquid rocket flew for two and a half seconds. Our first run with a positive energy balance lasted about nine. Our approach is: first make it exist, then make it good.”
Both numbers in that sentence are accurate. Robert Goddard’s rocket rose twelve and a half meters above a farm in Massachusetts in March 1926 and flew for two and a half seconds. Ninety-two years after that flight, hundreds of satellite constellations were operating in orbit, and cargo missions were heading to Mars. The history of modern spaceflight begins with two and a half seconds of iron tearing skyward. The history of seafloor geothermal power may have begun on September 27, 2026 — eight seconds later, but in the same genre.
I got hooked on this story in a morning tech digest: a short line saying that the startup Endurance Energy had “generated electricity from the heat of an undersea volcano for the first time in the world.” I read it and caught myself not understanding where the line ran between “first in the world” and “in reality, nothing works yet.” Eight seconds of positive energy balance — is that a premiere, or a press release? And why build a power plant on an undersea volcano when, for now, there is nowhere for its electricity to go?
The longer I read, the more interesting it got. There is engineering exotica here: an organic Rankine cycle on the ocean floor, titanium heat exchangers, electronics in oil-filled chambers that withstand a hundred and fifty atmospheres of pressure. There is the seventy-year history of Pacific oceanographic programs — because the startup came to work alongside cables scientists had laid there long ago, and operated its plant from Seattle through a scientific fiber-optic network. There is venture logic in its pure form: $63 million, including a Series A led by Founders Fund — and the first commercial customer is not in California or Japan, but in the Kingdom of Tonga, which runs on imported diesel. And finally, there is the volcano itself: the only volcano on the planet whose eruption scientists publicly promised to predict in advance in 2024, missed by a year, and are now even more confident about.
Let’s peel it back layer by layer: what exactly those eight seconds proved, why a company full of ex-SpaceX people went down to the seabed, why an undersea power plant needs someone else’s scientific infrastructure — and why the greatest threat to this project may come not from the ocean, but from a budget request on Capitol Hill.
The place is Axial Seamount, a shield volcano on the Juan de Fuca Ridge in the northeastern Pacific, roughly 480 kilometers off the Oregon coast. It is the most active volcano in the Pacific Northwest: its eruptions in 1998, 2011, and 2015 were fully documented instrumentally, with data from before, during, and after. That is why it was chosen as the site for the first experiment — but not only for that reason.
Endurance Energy went out to Axial on a three-week expedition in September 2026 — aboard the vessel Ocean Guardian. It was already the company’s fifth deep-sea expedition in just one year of field operations. The timeline is fully documented in Endurance’s press release and retellings by Ars Technica, GeekWire, and Canary Media, and it reads like the shift log of a duty engineer.
It started with drilling. A subsea drilling lander bored two holes into the caldera floor a couple of meters apart, and both produced hydrothermal fluid. The second was fitted with casing, filled with cement, and sealed with a valve — creating the world’s first subsea geothermal well: 16 meters deep, 5 centimeters in diameter. For comparison: the initial casing on oil and land-based geothermal wells is 50–90 centimeters across, and their depths begin at hundreds of meters. After cementing, the well delivered fluid at 260 degrees Celsius. But there was not enough flow — the plant simply choked on it.
I’m stressing this point because it is the key to the whole story. The “world’s first well” turned out tiny not because the company could not make it any other way, but because that was the plan: the goal of the first trip was to test the chain itself — “drill, case, cement, open.” The task of “make the plant work” had not yet been set. First the chain, then the power.
So Adélie — a three-ton plant the size of a small van, named after the Adélie penguin — was moved to a nearby natural hydrothermal vent blasting out of the seabed at 281 degrees Celsius. The hydrothermal fluid passed through a titanium heat exchanger, boiled a closed loop of organic working fluid — a classic organic Rankine cycle, the same one that drives hundreds of land-based geothermal plants — and the turbine produced electricity. The turbine reached nearly 26,000 revolutions per minute.
The first run lasted 29 seconds: the battery pack caught a current spike during spin-up and stopped responding to commands, and the system went into safe shutdown. Ten minutes later, the team tried to restart the plant — and immediately found that the power cable feeding the working-loop pump was dead. The cause was a defect in the outer jacket: seawater had seeped inside and shorted two conductors. That was the end of the energy part of the expedition: three weeks of work, two wells, one plant, one blindingly brief run on a natural vent, and a collection of engineering lessons that were the whole point of the exercise.
Endurance Energy was founded in January 2025. Its creator, Andrew Redd, had previously worked as an engineer at SpaceX — and left to build his own business at precisely the moment geothermal went from an academic subject to a venture one. In a year and a half, the company raised $63 million; $54 million of that was a Series A led by Founders Fund, Peter Thiel’s fund, known for betting on “hard engineering” — from defense drones to helium mining on the Moon. The team of roughly three and a half dozen people was assembled according to the same principle as the product itself: twelve SpaceX alumni, plus veterans of fusion company Helion and Blue Origin. The investor list also explains the company’s signature ambition: do things that look on paper like Ph.D. material, but in practice ought to get a taste of revenue within five years.
The name was not chosen by accident. The company’s subsea plants are named after penguins — as capacity grows, the chain of names moves up the ocean food chain: from the Adélie penguin, which lends its name to the hundred-kilowatt prototype, through gentoo — the next, upgraded version due to go to the seafloor this winter — to the two-megawatt “Leopard” now in development: Leopard Seal, the penguins’ main predator. Each successive device eats whatever came before it. It is the most honest nomenclature in energy I have ever seen.
Inside the product, things are more prosaic. The organic Rankine cycle is the same machinery used in land-based geothermal stations since Italy in 1904, only folded into a three-ton box lowered to the seabed and placed over a well. Titanium heat exchangers — because hydrothermal fluid is acidic brine full of dissolved metals that eats ordinary stainless steel in weeks. Electronics in pressure-balanced oil-filled chambers: the oil inside the housing compresses by exactly as much as the water outside, and the housing remains intact a kilometer and a half down. No magic — just the exponentially rising cost of the relationship between depth and sealing.
The well is drilled with filtered seawater, without drilling mud. That both minimizes chemical contamination of the surrounding ecosystem — tubeworms, crabs, mollusks, bacterial mats live around the vents — and simplifies the legal side: the project brought along a voluntary environmental monitoring program, from seabed imaging to water samples before and after every operation, and cross-checked its data against eleven independent scientific instruments on the same volcano. The loudest sound recorded during the work was the hum of the vessel’s main engine — within the normal range for research ships that have worked in these waters for decades.
During the test, the plant was not feeding energy into an onshore grid — because there simply is no onshore grid it could connect to. It was operated through the Regional Cabled Array: a network of 900 kilometers of fiber-optic and power cables launched by the U.S. National Science Foundation in 2014. It runs across the entire Juan de Fuca Ridge so that one hundred and fifty scientific instruments — thermometers, seismometers, cameras, chemical probes — can send their data ashore in real time. Adélie drew power from that same network and reported through it: operators in Seattle could see temperatures, flow rates, turbine speed, and current in real time, as if the plant were standing in their server room.
This is one of the most elegant and deceptive details in the entire story. Elegant — because a newcomer startup was able to reach the ocean floor 480 kilometers offshore without laying a meter of its own infrastructure: instead of spending hundreds of millions on its own cable, Endurance leased someone else’s, built with American taxpayer money for fundamental science. Deceptive — because the very infrastructure that made the experiment cheap exists in only one place on the planet and is funded through a single line in the federal budget, over which a separate war is now being fought in Washington. More on that below.
The startup could afford that luxury precisely because the experiment was not “the first plant,” but “the first proof-of-concept test.” There is another path. Others are already on it: in 2016, a Chinese team drew several watts of power from a small hydrothermal vent in the southwestern Indian Ocean — a closed publication in an applied-energy journal that did not become the beginning of an industry. Iceland’s IDDP program, which drilled the world’s first well into supercritical fluid at the Reykjanes volcanic field in 2016–2017, found that supercritical water at 400+ degrees eats through everything but expensive alloys within months — the program has existed for a decade and a half and remains a research effort. The lessons are the same: at the boundary of “volcano plus engineering,” the main enemy is not temperature itself, but its aggressive chemistry, mineral deposits, and corrosion, which consume equipment faster than the generated electricity can pay for it.
But back to the question you may already have asked: why put a power plant on the ocean floor when solar is cheaper on land, wind is cheaper, and land-based geothermal is already a mature industry?
Andrew Redd answers this very carefully: preserve the Earth’s surface. “We don’t have that much space on the surface of the Earth for energy infrastructure. Anything we can shift into the ocean — a place where nobody lives — is a win for everyone.” There is a great deal to argue with in that claim, and we will get to it, but there are honest engineering arguments behind it too. The ocean at a depth of a kilometer and a half is a free, infinite refrigerator, making the thermodynamic cycle efficient without cooling towers or fans. Hydrostatic pressure at that same depth keeps the system naturally pressurized: the pipes do not collapse because pressure outside and inside is balanced. The resource is right there: hydrothermal systems along mid-ocean ridges continuously vent the Earth’s interior heat into the ocean, operating around the clock in any weather, with a capacity factor neither solar nor wind can see even in their dreams.
The Kingdom of Tonga has been chosen as the first commercial customer. This is not a whim, but a calculation aimed at the weakest link in the global energy balance. Tonga is an archipelago in the South Pacific with 100,000 residents, whose electrification relies almost entirely on imported diesel: in 2024, fossil fuels supplied nearly 88 percent of the islands’ generation. Diesel is shipped to the islands by sea — its price hits the national budget, and after the war in the Middle East pushed oil prices up worldwide in 2026, the Tongan government publicly asked for assistance as an “island state dependent on fuel imports.” The entire country’s installed capacity is around 40 megawatts, with annual consumption of roughly 70 gigawatt-hours. Endurance plans to bring six two-megawatt plants to Tonga’s shores: 12 megawatts of continuous generation — around a third of the country’s total electrification, running in a mode that depends neither on cloud cover nor on calm winds.
Put another way: humanity’s first commercial contract for subsea energy will probably be signed not where the most money is, but where diesel costs the most. Tonga is to subsea geothermal what island grids are to solar-plus-storage: a market where the alternative is so expensive and so fragile that even an experimental technology looks like a sensible bet.
Now for the second layer of this story — the reason Axial is worth every word. Axial Seamount is not just a convenient test site; it is the best-studied undersea volcano on Earth. It erupts abnormally often: 1998, 2011, 2015 — at intervals of four to thirteen years. Each eruption follows the same mechanics: magma accumulates beneath the caldera, the seafloor swells — over the years, uplift reaches meters — then comes the release: the seafloor drops by two and a half meters, and the volcano discharges in lava flows across the seabed. The danger to people is zero: beneath a kilometer and a half of water and a hundred and fifty atmospheres of pressure, lava cools into pillows of basalt within hours.
Since 1997, the volcano has been draped in seismometers and pressure sensors, and since 2014 it has been connected to that same regional cable network through which Endurance later operated its plant. One hundred and fifty instruments, real-time data, round-the-clock monitoring of the tilt and elevation of the caldera floor. To paraphrase specialists at Oregon State University, it is the only volcano on the planet whose condition is updated every fifteen minutes.
That is why, here for the first time in the history of volcanology, they dared do what is usually considered professional suicide: make a public eruption forecast accurate to within a year. In December 2024, volcanologist William Chadwick announced that the caldera floor had swollen to the level preceding the two previous eruptions, and that an eruption should be expected before the end of 2025. The press immediately translated this into universal human language: “a volcano is preparing to explode beneath Oregon.” Chadwick had to separately explain to frightened coastal residents that the volcano was three hundred miles away horizontally and a kilometer and a half down vertically, and no wave would reach the shore.
2025 passed. The volcano swelled above its 2015 level, earthquakes continued, no lava emerged. The forecast failed — and the failure itself became the most useful scientific result in a decade: it turned out that the level of inflation is not a threshold but an indicator, and that the volcano can accumulate magma above its “critical point” without erupting. “In hindsight, we just got lucky in 2015,” Chadwick said at the American Geophysical Union conference in December 2025. His new model shifted from the height of inflation to the frequency of earthquakes: by his estimate, the eruption will happen in 2026, when volcanic tremors arrive in waves of five hundred a day. In other words, the forecast remains in force — it has simply moved by a year and changed its mechanism.
For Endurance, this is not academic curiosity. It will return to Axial this winter with the upgraded Gentoo plant, which is meant to remain on the seabed for months. If an eruption happens while the plant is sitting on a vent, Endurance will get unique data: what a power plant standing on a volcano does when the volcano wakes up. And volcanologists will get the same thing from the other side: their monitoring becomes infrastructure for someone else’s business.
Skepticism about the whole venture does not come from internet comments, but from people who count energy money professionally.
Roland Horne, a Stanford University professor and one of the most cited experts in geothermal energy, answered briefly: “Probably technically feasible. But surely more expensive than on land.” Land-based geothermal already balances on the edge of profitability, and working in the middle of the ocean adds costs at every step — the design stops adding up against the simplest equation.
Emily Pope, a senior fellow at the Center for Climate and Energy Solutions, was even harsher: “I find it hard to see how projects like this will make economic sense.” For islands, it is cheaper to install solar panels with batteries than to send drilling vessels to an undersea volcano. And the question is not just the price of a kilowatt-hour, but the fact that the main expense of a subsea project is not the plant but the cable to shore: a dedicated subsea transmission line costs hundreds of millions of dollars and requires years of permitting. The first commercial project in Tonga will have to prove otherwise.
Against these arguments, the project has its own structural answer, and it is not entirely obvious.
What matters is that the startup is not going after the U.S. market, where electricity is cheap and regulation dense, but after places where power is most expensive. Tonga imports 100 percent of the fuel for its generation and pays several times more per kilowatt-hour than an American consumer on the coast. Twelve megawatts of continuous clean generation near the islands covers roughly a third of the country’s needs — and it is sold not in competition with a gas turbine, but against diesel brought in by sea in barrels. This is beachhead logic as old as the world: first sell where the alternative is awful, and only then move into the mainstream.
The second answer is physical. Hydrothermal vents operate around the clock and do not depend on weather, while deep water is a free, infinite refrigerator that lets the thermodynamic cycle work efficiently without cooling towers. The weight of ocean water at a depth of a kilometer and a half keeps the system under pressure by itself — no need for massive wellhead equipment and pumping stations, which make up a noticeable share of costs on land. A titanium heat exchanger instead of a concrete cooling tower: less concrete, less surface area, more salt.
And the third answer is about the Earth’s surface. Solar and wind farms take up land, and the amount of usable, unoccupied land is finite. An undersea plant with a footprint of a few square meters on the seabed and continuous generation offers a power density that only a nuclear plant can deliver on land. Redd puts it more ambitiously than anyone: “The Earth’s surface is too valuable to occupy with energy infrastructure. Anything that can be moved into the ocean should be moved into the ocean.” There is only one way to test whether that is true: build it and count. That is, in fact, what Endurance is doing.
Eight seconds of net-positive energy is proof of principle, not functioning power generation. The only real energy milestone is sustained continuous output, and that has not yet been achieved even once. The distance to a historic breakthrough here is the same as the distance between the Wright brothers’ first flight and a transatlantic crossing — and that is the best possible news, because in 1903 nobody understood how to begin mass production either.
The well played no part in the only energetically meaningful run: the plant was fed from a natural vent because the drilled well turned out to be too thin. In practice, the experiment proved two things at once — that we know how to drill on the seabed, and that for now we are taking energy directly from nature.
The pump’s power cable burned out ten minutes after the emergency shutdown, while the battery could not handle its startup current. The engineering problems are so ordinary that it is almost reassuring: not the physics of the undersea world breaking down, but ordinary electronics that can be fixed in any workshop.
Thermal pollution, noise, mineral discharge, and risks to the unique ecosystems around hydrothermal vents — Endurance’s voluntary monitoring is good, but “voluntary” and “independent” are different words, and there is still nobody to regulate subsea energy: laws for it do not exist in any country in the world.
And most importantly — infrastructure dependency. Adélie was powered and operated through the National Science Foundation’s scientific cable network, and in 2025 the U.S. administration proposed cutting the budget for oceanographic observation programs by eighty percent. The program survived until the summer of 2026 on one-off measures — what comes next is unclear. If the network is cut back, everyone loses: volcanologists and their forecasts, and startups building their energy systems along someone else’s scientific cables.
The story of eight seconds is not a story about energy. It is the story of what the very beginning of a technology looks like — one that still has no name in textbooks, no regulator, no market, but has already required drilling a well into the ocean floor.
In 1926, a lone enthusiast on a farm in Massachusetts made iron fly for two and a half seconds. Perhaps the main takeaway from these eight seconds on the Pacific floor is this: humanity has, for the first time, managed to close an energy loop on the ocean floor — physically, technically, live for one hundred and fifty scientific cameras. What grows from that will be shown by the next eight minutes, eight hours, and eight years of plants operating on the seabed. But a world in which power plants grow out of volcanoes has stopped being science fiction and become an engineering agenda for decades ahead.
First. Press releases do not lie with numbers, but with framing. Every key number in this story is honest: 60 watt-hours, 8.7 seconds, 16.5 kilowatts at peak — all independently repeated in three publications and in company documents. Even the word “first” is honest: the first subsea geothermal well, the first end-to-end run of the entire chain on the seafloor. There is no need to lie: the frame “we built volcano energy” turns nine seconds of kettle time into a headline by itself. The answer to my original question — “premiere or press release?” — turned out to be boringly engineering-like: a premiere, but a premiere of system behavior rather than energy generation. Exactly the same difference as between “the rocket flew” and “the rocket delivered cargo.”
Second. The combination of “SpaceX people + Founders Fund + a scarce physical resource” is already an established launch pattern for hard-tech startups, and it deserves attention in its own right. In a year and a half, a team of twelve ex-SpaceX people went from a sheet of paper to the first undersea well in history and a plant on a volcano — on $63 million. For comparison: a single deep well in a land-based geothermal project can consume tens of millions. The speed is explained not by talent but by test discipline: each expedition tests exactly one hypothesis and documents every failure. “First make it exist, then make it good” is not a slogan, but the operating procedure of rocket programs, applied to seafloor energy for the first time.
Third. The most underrated element in the whole story is not the turbine or the well, but the cable. Scientific infrastructure paid for by taxpayers to study a volcano suddenly turned out to be ready-made logistics for private business — and conversely, private business effectively conducted a field test of the scientific network with its experiment. It is a rare kind of symbiosis. The budget threat to this network, discussed below, could break that symbiosis at its most interesting point — and that, I think, is the story’s main political lesson: science that people have stopped considering useful sometimes turns out to be the wire through which someone else’s energy flows.
Fourth. The engineering in this story teaches the same thing the entire history of energy has taught: the physical limit is not temperature or pressure, but corrosion. Instinctively, you expect something heroic to break on a volcano. Instead, the two most mundane components failed: the battery could not digest its own startup current, and water leaked into the cable jacket. This is exactly what kills land-based geothermal in Iceland and casing strings in oil fields: not thermodynamics, but chemistry. The cost of subsea energy will be determined not by the efficiency of the Rankine cycle, but by the price of how long a titanium pipe can survive on the ocean floor. The question “how long does a seal last?” matters more than the question “what is the power output?”
And fifth, the most human part. The plant has a penguin name, the team has a puppy named Maple in its corporate photos, and its leader compares his plant to Goddard’s first rocket — not because it is pretty, but because it is a precise way of explaining to investors and to themselves that eight seconds is normal. Technologies begin with funny little numbers, and greatness is defined not by the first digit but by the slope of the curve growing out of it. Goddard’s rocket flew for two and a half seconds — and that was enough. We will see whether eight is enough.