On a remote stretch of tundra near the Norwegian border, in the Pechengsky District of Russia’s Kola Peninsula, there is a rusted steel manhole cover bolted into a slab of concrete. The cover is roughly the size of a small dinner table. The dozen large bolts holding it down have not been removed in more than three decades. Underneath is a pipe nine inches across that descends 12,262 meters — 40,230 feet, 7.6 miles — into the continental crust of the Earth. That is deeper than the Mariana Trench. It is the height of Mount Everest plus the height of Mount Fuji, stacked. It is the deepest artificial point ever made on the surface of the planet, and it has held that record continuously since June of 1990, which means that on the day Tim Berners-Lee proposed the World Wide Web, no one in the human species had ever drilled deeper than this hole, and on the day OpenAI released GPT-4, the same thing was still true.

The hole is called the Kola Superdeep Borehole, and it is the kind of infrastructure project that only happens when a country has a lot of money, no shareholders, an active geopolitical rivalry, and a willingness to spend twenty years drilling toward an answer it never actually got. The Soviets started it in 1970, drilled until 1992 when the rock got too hot and the country that was paying for it stopped existing, mothballed it through the rest of the decade, formally closed it in 2005, and abandoned the surface compound by 2008. The wooden derrick that once stood over the wellhead was dismantled. The buildings around it collapsed into the permafrost. In 2026, the site is a ruin in the Arctic, accessed via deteriorating roads in a closed military district about 150 miles from Murmansk. The hole itself is almost certainly deformed and partially collapsed in its deepest sections, which is what happens to a 23-centimeter-wide pipe when active circulation stops and the rock around it keeps cooking at 180 degrees Celsius. Nobody has been down there to check. Nobody has been to the surface compound in any organized way since the Russian invasion of Ukraine made foreign scientific exchange with Murmansk a non-starter.

And nobody — anywhere in the world, in 36 years of trying — has drilled a deeper vertical hole.

What it was actually for

The Kola project was the Soviet entry in a Cold War scientific contest that almost nobody remembers, because the other side lost interest and called it off. In 1957, the United States announced Project Mohole, an attempt to drill through the oceanic crust to reach the Mohorovičić discontinuity — the “Moho,” the boundary between the Earth’s crust and the underlying mantle. The Moho was the prize. Drilling to it would have produced direct samples of mantle rock for the first time in human history, settled decades of arguments about Earth’s deep structure, and demonstrated that humanity could reach the other side of the same kind of hard-rock boundary that defeats every other industrial process. Project Mohole drilled into the seafloor off Mexico to a grand total of 183 meters in 1961 and was canceled in 1966 by Congress for cost overruns. The Soviets, watching this, decided to do it from land — where the crust is thicker but the engineering is cheaper — and announced the Kola Superdeep Borehole as a national prestige project on May 24, 1970. The target was 15,000 meters. The rationale was scientific. The motivation was that the Americans had quit.

The project ran for 22 years and never reached the mantle. The Kola crust at that location is around 35 kilometers thick — roughly the thickness of the entire Baltic Shield — and at the bottom of the drilled hole the borehole had penetrated about a third of the way down. The original 1970 target depth of 15,000 meters was, in hindsight, geophysically arbitrary. The Soviets picked it because it sounded ambitious. The drilling team, led by geologist David Guberman and the team at the Kola Scientific Center, hit 11,662 meters in October 1982 — already a world record — drilled a side branch off the main hole, hit 12,262 meters in 1990, broke equipment, started a fifth hole from 8,278 meters, drilled a few hundred more meters, and stopped in 1994 because the country was out of money. The official cause of project failure depends on who you ask. The temperatures at the bottom were 180°C instead of the predicted 100°C, which meant the drilling fluid kept flashing into vapor and the steel kept softening. The rock at depth had started behaving plastically, oozing back into the borehole faster than the drill could clear it. The Soviet Union had stopped paying salaries. All three things were true at once.

What they actually found

The Kola Borehole was a scientific disappointment in exactly the sense that the first fusion reactor experiments were a scientific disappointment — it did not deliver the headline goal, and what it did deliver was so unexpected that almost everyone forgot how disappointed they were. The pre-drilling consensus was that beneath about seven kilometers of granite, the team would find a layer of basalt — the Conrad discontinuity — which had been inferred from 1923 seismic data and treated as textbook geology for half a century. They never found it. The granite kept going. What had looked like a basalt boundary in seismic data turned out to be a metamorphic transition inside the granite itself — the same rock, denser and more crystalline below a certain depth, and just dense enough to bounce seismic waves the way basalt would. Fifty years of geophysical models had to be quietly revised.

At 6.7 kilometers down, in rocks dated to roughly 2 billion years old, the team found microscopic fossils of single-celled marine organisms — 24 species of preserved plankton, sealed in carbon and nitrogen compounds inside the metamorphosed rock, still recognizable. The Archean ocean had left fossils a third of the way through the continental crust, and they were still there. At nearly the same depth, the borehole encountered free water — liquid water inside fractures in crystalline rock — at depths where existing theory said no water could possibly exist. The drilling mud at depth bubbled with hydrogen, helium, nitrogen, and carbon dioxide. Soviet scientists described it as “boiling.” The hydrogen was probably the product of serpentinization — water reacting with deep iron-rich minerals to produce hydrogen gas — and it changed the field’s understanding of where hydrogen and abiotic methane come from in the deep crust. Three findings that should each have generated entire research programs were, instead, footnotes in textbooks because the country that ran the experiment fell apart in 1991 and no Western institution was set up to inherit the results.

Then there was the heat. The Soviet team had budgeted for about 100°C at 12 kilometers. They got 180°C. That extra heat was the immediate engineering constraint that stopped the drill — but it was also, in retrospect, the most commercially valuable thing the Kola Borehole ever discovered. The Earth was hotter at depth than anyone had modeled. The implication was that if you could reach those depths in commercial quantities, you would have access to a thermal reservoir vastly larger than any conventional geothermal field. The Soviets noted this and moved on, because their economy collapsed before they could capitalize on it. Thirty-six years later, three American startups are trying to build the entire next generation of carbon-free baseload power on the implication.

Nobody has beaten the record, but somebody is trying

The vertical record set at Kola in 1990 has stood for 36 years. Two projects have come close. In May 2008, the BD-04A well at Qatar’s Al Shaheen Oil Field reached 12,289 meters of total drilled length — 27 meters longer than Kola — but the BD-04A is an extended-reach lateral well, mostly horizontal, with a 10,902-meter horizontal section. It is the longest measured well, not the deepest vertical one. In February 2025, China National Petroleum Corporation completed Shenditake 1 in the Taklimakan Desert of the Tarim Basin in Xinjiang, drilling vertically to 10,910 meters — the deepest onshore well in Asia, the world’s second-deepest vertical well — and ending drilling 90 meters short of the planned 11,100-meter target when active oil and gas indications gave them an excuse to stop. The Shenditake 1 took 580 days to drill, 300 of those days for the last 910 meters. At 10,000 meters down, the temperature in the borehole exceeded 210°C — hot enough to vaporize cooking oil — and the pressure exceeded 130 megapascals, higher than the crushing force at the deepest point of the Mariana Trench. The CNPC engineer who led the project said drilling was “as difficult as the lunar exploration programs,” which is the kind of comparison a country makes when it is competing simultaneously for moonshot technology bragging rights and strategic energy reserves and doesn’t see a meaningful distinction between the two.

Shenditake 1 is the first serious vertical challenge to Kola since 1990, and it still came up 1,352 meters short — roughly the height of the Burj Khalifa, plus the height of the Empire State Building. The structural reason no one has matched Kola is the same reason Kola itself stopped: the deeper you go, the hotter the rock gets, the more the drill string deforms under its own weight, the more the borehole walls try to close in on the equipment, and the less any of the tools of conventional rotary drilling — the same drill bits and mud-pumping rigs that the oil and gas industry has refined over a century — actually work. Drill bits made of tungsten carbide and synthetic diamond can chew through granite, but they wear out, they need to be replaced, they require pulling thousands of meters of pipe out of the ground, swapping the bit, and lowering everything back down — a process that takes days each time and that gets worse the deeper the bit has gone. At 7.5 miles down, every meter of additional drilling consumes more equipment, more time, and more money than the meter above it. The marginal cost is going up at the same time that the engineering envelope is collapsing.

This is the engineering problem that Quaise Energy, a Houston-based startup spun out of MIT’s Plasma Science and Fusion Center in 2018, is attempting to render obsolete. Quaise’s drilling system uses a gyrotron — a high-power millimeter-wave generator originally developed for plasma heating in fusion reactors — to ablate rock instead of grinding it. The gyrotron beams a focused electromagnetic wave down a waveguide into the rock face, vaporizing the rock at the bottom of the hole; the vapor is then carried up the hole by a purge gas. There is no drill bit. There is nothing to wear out. The technology is, in principle, indifferent to depth, indifferent to rock temperature, indifferent to the hardness of the granite that defeated the Soviets at Kola. In July 2025, Quaise drilled 100 meters of Texas granite in a field test — a record for millimeter-wave drilling, and the first time the technology has produced a hole more than a few centimeters deep outside the MIT laboratory. The company has announced plans for a 10x more powerful gyrotron and a pilot superhot geothermal plant in the western United States by 2028, targeting commercial drilling to depths of 20 kilometers and rock temperatures of 400°C.

If Quaise works at commercial scale — and that “if” is doing a lot of structural load — the Kola record will be obsolete by the end of the decade, the entire deep-geothermal industry will become a direct competitor to the alternative carbon-free baseload technologies that hyperscalers are currently signing power purchase agreements with, and the Earth’s interior heat will become accessible at depths and temperatures the Soviets reached once, painfully, by accident, in 1990. The proof of concept for the commercial opportunity is buried in a Russian field report from 1985 that almost nobody has read since the Cold War ended.

The site today

You cannot, in 2026, simply drive to the Kola Superdeep Borehole. The site is inside the closed Pechengsky District near the Russian-Norwegian border, in a region that has become significantly less accessible since 2022. The international scientific cooperation that brought German, Finnish, and American geophysicists to the wellhead in the 1980s and 1990s has been on indefinite hiatus since the invasion of Ukraine, and the larger Cold War-era research apparatus that produced the project no longer has a Russian institutional successor that anyone in the West is talking to. The handful of journalists and explorers who have visited in the last several years describe a derelict compound: collapsing pre-fabricated barracks, rusted-out machinery half-buried in tundra, the wooden tower long gone, and at the center of it all the sealed steel cap and its dozen bolts. The Pechengsky District is in permafrost. The structures sink, the buildings warp, the windows go missing, and the surrounding scrap metal slowly disappears as locals haul off anything that can be sold. The hole itself is still there. It is just no longer obviously a hole — it is a manhole cover in a clearing, surrounded by the bones of the infrastructure that once supported it.

This is the rare Pipe Dreams subject that did not survive its own success. The Manhattan steam grid is 144 years old and still heating skyscrapers because the buildings above it were designed around it. The Iranian qanats are 2,500 years old and still flowing because the engineering is too simple to break. The Wuppertal Schwebebahn has been carrying commuters since 1901 because the valley below it has no other transit option. The Berlin Rohrpost outlasted five regimes because the pipes were already in the ground. The Mumbai dabbawalas have been running their lunchbox delivery network for more than a century because the system requires no infrastructure beyond human labor and the Mumbai rail timetable. The Kola Superdeep Borehole survived its drilling phase, set the deepest-vertical-hole record, made fundamental scientific discoveries, and was then abandoned because the discoveries themselves did not generate a commercial follow-up. Nobody was buying mantle samples. Nobody was selling 180°C steam. The infrastructure existed to do something — drill to the Moho — that did not happen, and once the drilling stopped, there was no second use for the site, the way Manhattan’s steam grid found a second use heating hospitals or the Falkirk Wheel found a second use connecting two canals that had been derelict for decades. Kola was a single-purpose machine. Its purpose ended. The machine stopped.

What 2026 actually looks like at Kola, and underneath it

The deepest hole on Earth, in 2026, is a record holder by default — held by a sealed Soviet artifact in an Arctic ruin that no Western scientist has visited in years, surrounded by the rusting metal that was once the support infrastructure of one of the most ambitious scientific drilling projects ever attempted, in a closed Russian military district that is, geopolitically, more isolated than it has been at any point since the project began. The record itself has stood for 36 years not because the engineering is impossible — China demonstrated in February 2025 that 10,910 meters of vertical drilling can be done with modern equipment in 580 days — but because the economic and scientific motivation that drove the original project has not reassembled. The Cold War rivalry that funded Kola is gone. The Moho is still out of reach. The mantle samples are still hypothetical. The geothermal opportunity that the heat readings hinted at in 1985 is being pursued not by re-entering Kola but by drilling new holes elsewhere using fundamentally different technology, with drill bits replaced by gyrotrons and conventional drilling replaced by ablation.

If Quaise works, or Fervo’s enhanced geothermal systems work, or Sage Geosystems’ pressure-geothermal pilots work, then sometime before 2030 some American startup will drill a hole somewhere in Utah or Texas or Nevada that quietly surpasses Kola’s depth — not as a national prestige project, not as a Cold War statement, not even as a scientific endeavor, but as a power-generation project meant to sell round-the-clock carbon-free electricity to a hyperscale data center running a language model trained on every text humans have ever written. The record will fall. It will fall in service of a use case nobody at the Kola Scientific Center could have anticipated in 1970, on the back of microwave technology that did not exist when the Soviets were drilling, to power computational systems that did not exist when the Soviet Union itself existed. And when it does fall, the Kola Superdeep Borehole will become exactly what it has been trying not to become for 36 years: the second-deepest hole on Earth, sealed under a rusted manhole cover in a closed Russian district, at the bottom of a tundra clearing that nobody visits, in a country that is no longer on speaking terms with the people building the holes that will eventually go deeper. Twelve thousand two hundred and sixty-two meters down, in the dark and the 180-degree heat, the granite is still there. The microfossils are still there. The water is still there. The hydrogen is still bubbling out of the rock the way it has been for 2 billion years. The hole at the surface is nine inches wide, the cover is bolted down, the bolts are rusting, the permafrost is creeping, the wood is gone, the country is unrecognizable, the science is settled, the record is held, the drill is gone, and the only thing the deepest hole in the world is doing in 2026 is waiting to be made shallow by comparison.

The Narva power plants burn oil shale — a sedimentary rock that produces roughly one-third the energy of coal per tonne but generates more CO2, more particulate matter, and more ash per kilowatt-hour than any other fossil fuel in commercial use. Estonia, a country of 1.3 million people, has one of the highest per-capita carbon emission rates in Europe, and the Narva complex is the reason. In 2007, the plants generated 95% of Estonia’s electricity. They made Estonia one of the only countries in the EU whose power supply was essentially independent of Russian gas — which is why, when Russia invaded Ukraine in 2022 and the rest of Europe scrambled to find alternatives, Estonia’s energy minister could shrug. The dirtiest power plants in the EU are also the most geopolitically independent power plants in the EU. And they sit five kilometers from the Russian border, in a city that is 97% Russian-speaking, on the most exposed section of NATO’s eastern flank, in a region where the primary employer is the state energy company that operates the plants the EU wants shut down. The Narva oil-shale complex is where three irreconcilable imperatives — energy security, climate targets, and the social stability of an ethnic minority on a frontline border — collide inside a single facility. Nobody has figured out how to satisfy all three.

The fuel nobody else uses

Oil shale is not shale oil. It is a fine-grained sedimentary rock containing kerogen — an organic compound that, when heated, decomposes into synthetic petroleum products and combustible gas. Estonia sits on one of the world’s largest oil shale deposits — the Baltic Oil Shale Basin — and has been mining and burning it since 1924. The Balti Power Plant was built between 1959 and 1965. The Eesti Power Plant was built between 1963 and 1973. The newer Auvere Power Plant, completed in 2015 at a cost of €540 million, uses circulating fluidized bed combustion to extract more energy per tonne and reduce — though not eliminate — the emission penalties. Together, the Narva complex has a nameplate capacity of approximately 1,615 megawatts.

The rare earth deposits and critical minerals that sustain the global technology supply chain are defined by concentration — the resource exists in commercially viable quantities in very few places. Estonian oil shale is the same: it is commercially viable almost nowhere else, because the energy return on investment is so poor that most countries with oil shale deposits (the U.S., Brazil, Jordan, Morocco) have never found it economical to exploit at scale. Estonia burns it because Estonia has it, because the Soviet Union built the plants to burn it, and because the alternative — importing Russian gas through pipelines controlled by Gazprom — was a dependency that three decades of post-Soviet independence have been dedicated to eliminating. The fuel is terrible. The alternative was worse.

The grid disconnection

On February 8, 2025, Estonia, Latvia, and Lithuania disconnected from the BRELL grid — the Soviet-era electrical network linking Belarus, Russia, and the three Baltic states. On February 9, the three countries synchronized with the Continental European Network. “Today, history is made,” European Commission President Ursula von der Leyen said at a ceremony in Vilnius. “This is freedom, freedom from threats, freedom from blackmail.” The disconnection had been planned since 2009. Russia’s invasion of Ukraine accelerated the timeline by nearly a year. The cost of the necessary infrastructure — submarine cables, grid reinforcements, the LitPol Link connection to Poland — totaled approximately €1.6 billion, largely funded by the EU.

The Narva plants’ role in the newly independent grid is paradoxical. The Delta Works protect a country that would vanish without them. The Narva plants provide the dispatchable — on-demand, adjustable — power that stabilizes a grid that has just severed its connection to the system that previously provided frequency regulation from Moscow. Renewables supply an increasing share of Estonia’s electricity (Eesti Energia’s Enefit Green subsidiary operates wind farms across the Baltics), but wind and solar are intermittent. When the wind stops and the sun sets, the grid needs a plant that can ramp up immediately. The Narva complex is that plant. The dirtiest power in the EU is also, in the immediate aftermath of the grid disconnection, the dispatchable backbone that keeps the lights on in three NATO countries.

The city the plants built

Narva — Estonia’s third-largest city, population approximately 55,000 — is 97% Russian-speaking. The city exists because of the oil shale industry. The plants, the mines, and the support infrastructure employ a significant portion of the workforce. The Fergana Valley enclaves exist because Soviet ideologues drew borders that didn’t match the populations. Narva’s Russian-speaking majority exists because Soviet industrial planners imported Russian workers to staff the oil shale complex they built in the 1960s. The infrastructure created the community. The community depends on the infrastructure. The EU wants the infrastructure shut down.

Ida-Virumaa — the county that contains Narva and the oil shale region — has the highest unemployment rate, lowest incomes, and weakest economic diversification in Estonia. The Schwebebahn is transit infrastructure that became a city’s identity. The Narva plants are energy infrastructure that became a community’s livelihood — and shutting them down without replacing the jobs risks destabilizing a Russian-speaking border city that the Kremlin’s propaganda apparatus already targets as a potential zone of influence. After Russia’s invasion of Ukraine, Estonia’s president, prime minister, and defense minister visited Narva in rapid succession — not to inspect the power plants but to demonstrate that the Estonian state had not abandoned its Russian-speaking citizens. The visits were a security measure disguised as an inspection tour.

The closure timeline

Eesti Energia has committed to ending oil shale combustion for electricity by 2030. The Estonian government has pledged to phase out oil shale entirely by 2040. The EU Industrial Emissions Directive requires the closure of the older generation units, which have already exceeded their 50-year operational lives. The semiconductor supply chains and gallium processing facilities that sustain the chip industry face similar timelines — infrastructure that must be replaced before it fails, on schedules set by regulation rather than engineering. The Narva plants will not wear out. They will be regulated out.

But in 2025, a new oil shale plant was granted an environmental permit valid until 2035 — a decision that climate analysts called contradictory and that signals the tension between commitment and reality. The Great Man-Made River depletes a non-renewable aquifer because Libya has no alternative water source. The Narva plants burn a dirty fuel because Estonia’s grid, in the transitional years between Russian disconnection and full renewable buildout, has no dispatchable alternative at the required scale. The fuel is bad. The grid needs it. The timeline says 2030. The permit says 2035. The contradiction is the policy.

The ash fields

The environmental legacy of a century of oil shale combustion is physically visible from space. Ash deposits from the Narva plants cover approximately 20 square kilometers of the surrounding landscape — alkaline waste piles that leach heavy metals and sulfates into groundwater and waterways. The Aral Sea was deleted by irrigation infrastructure whose environmental costs arrived decades after the economic benefits. The LA Aqueduct drained a lake and is spending $2.5 billion on dust remediation. The Narva ash fields are the Baltic version: a century of energy production that left a toxic deposit the region will be managing long after the last turbine stops spinning. The Chicago River Reversal’s ecological consequences are still accumulating 126 years later. The Narva ash fields will be leaching into Estonian groundwater for comparable timescales.

Estonia’s Just Transition Plan — funded partly by the EU’s Just Transition Fund — allocates resources for retraining workers, diversifying the Ida-Virumaa economy, and remediating the environmental damage. The plan is modest relative to the scale of the problem: a 97% Russian-speaking city on NATO’s border, an economy built on a fuel the EU has mandated out of existence, ash fields that will require decades of remediation, and a grid that still needs dispatchable power that the renewable portfolio cannot yet provide. The dabbawalas face a demographic transition that threatens their customer base. The qanats are being killed by the wells that replaced them. The Narva plants face a regulatory transition that will eliminate their function — and the city, the community, and the border security implications that depend on that function have no replacement plan that satisfies climate physics, grid engineering, and NATO’s eastern flank simultaneously.

The infrastructure that survives is the infrastructure that solves only one problem. The Narva plants solve three — energy independence, grid stability, and regional employment — and the EU’s climate mandate requires eliminating the mechanism that solves all three, on a timeline that the grid hasn’t matched with alternatives, in a city that has no other economy, on a border that NATO cannot afford to destabilize. The plants are scheduled to close. The contradictions are not.