A line flickered through the September news feed, easy to scroll past: astrophysicists calculated that even under the most generous estimates of lunar water, a million-person city on the Moon would run dry in roughly a hundred years. A few weeks before these calculations, China sent its most complex lunar mission yet to the south polar region — hunting for precisely that water. The headline and the mission collided head-on: some say "not enough," others fly to check.
What caught my attention wasn't the number itself, but the shift in question. For a decade and a half the debate was "is there water on the Moon" — and that was settled. But the work by Martin Elvis and Jonathan McDowell, published September 14, 2026 in Frontiers in Space Technologies, asks the next, far more uncomfortable question: how long will it last. This is no longer the romance of discovery, but a water utility audit, and its result is unpleasantly chilling: a city will hit the water limit before any other problem.
The least obvious part of this story isn't "not much water" (that's familiar), but where exactly the fault line runs. The fuel business on lunar water passes the audit surprisingly easily: a billion tons is enough to refuel rockets for a thousand years. What doesn't fit the budget is people — with their showers, laundry, and salad for dinner. Let's break down how this arithmetic works out.
The Apollo missions brought back samples so dry that for half a century no one doubted: our neighbor is a desert drier than any earthly wasteland. The poles changed everything. The Moon's axis is tilted only 1.5° to the orbital plane (Earth's is 23.5°), so the Sun there perpetually skims the horizon, and the floors of polar craters haven't seen direct sunlight for about four billion years.
In these "permanently shadowed regions" the stable temperature doesn't rise above 110 kelvin — at such deep freeze, ice sublimates at no more than a millimeter per billion years. The polar craters are cryogenic vaults that have stood locked for four billion years since asteroids and comets delivered water to the Moon. Breaking open this piggy bank began in 2009–2010: the Mini-SAR radar on India's Chandrayaan-1 gave the first hundreds of millions of tons of "possible water," and NASA spectacularly fired a rocket upper stage into Cabeus crater — and spectrometers spotted ice in the ejecta cloud.
By 2013 a series of orbital missions had built confidence: in the polar cold traps there could be up to a billion tons of water. In 2020 the stratospheric SOFIA observatory found traces of molecular water even on the sunlit surface of Clavius crater — not only the permafrost turned out to be wet, but formally "dry" regolith too. Against this backdrop a new lunar dream was born: if there's water, there's drinking supply, oxygen, and rocket fuel — meaning you can have a base, then a settlement, then a city. The dream was publicly stoked by Jeff Bezos with plans to move heavy industry to the Moon and Elon Musk with "self-sustaining cities."
Elvis and McDowell approached such plans from the position of someone who was first to calculate asteroid mining economics: Elvis from the Harvard-Smithsonian Center for Astrophysics made his name in the 2010s with the estimate of "how many profitable asteroids even exist." Their paper isn't a forecast or manifesto, but an honest Fermi estimate: no exotic technologies, just income minus expenses. The first audit item, contrary to expectations, came out upbeat.
A little-known fact of lunar geography: the Moon's poles are the best place in the Solar System for solar panels after orbit. Because of those same 1.5° of tilt, the rims of polar craters ("peaks of eternal light," as they're somewhat inaccurately called) are illuminated almost continuously. And low gravity allows vertical towers with photovoltaics up to a kilometer high — tower-masts slowly rotating to follow the sun along the horizon.
The authors' calculation: a million-person city needs 800 MW just for living (American norm 0.25–0.80 kW per person, taking the upper bound — on the Moon temperature swings are almost 300 degrees, heating will be needed), plus industry with a 1.6 coefficient — total around 2.1 GW. Kilometer towers at the pole average 3 GW for 90% of the time. Nuclear fission sits in reserve: 3 GW reactor power requires 7.5 km² of radiators, but the darkness of polar craters with a view of black sky is an ideal place for radiating panels, and there are over four hundred suitable craters.
So energy is a weak constraint: there's enough sun and uranium for even several cities. The authors write directly that 3 GW of solar generation at the "peaks of eternal light" makes lunar data centers realistic too — each needs about a gigawatt. And here the audit pivots: if watts are fine, then the fate of lunar civilization is decided in liters.
The consumption norm is American, no breaks: 125 tons of water per person per year for drinking and hygiene. Breathing is unexpectedly cheap: conversion through molecular masses gives only 1.8 tons of water per year per person to replenish station atmosphere oxygen — a measly 1.4% of direct consumption. But food is the whale: by World Bank estimate, feeding one person takes 2–5 tons of water per day. Even optimistically squeezing this figure in half and believing in efficient agricultural technologies, the authors get 500 tons per person per year total consumption.
Add up for a million residents — and without any recycling the billion tons vanishes in 2.4 years. For scale perspective: a billion tons is a cubic kilometer of water, enough to flood Manhattan's Central Park to a depth of 120 meters, a third higher than the Statue of Liberty. Sounds solid — until you put a million-person city next to it, which drinks such volume faster than two Earth election cycles pass.
All hope lies in recycling. The authors' model is indecently simple: water is consumed linearly, part returns with efficiency e, and reserve lifetime equals T/(1−e). The entire fate of the lunar city is packed into a tiny fraction (1−e) — the percentage point that doesn't return to the tap.
The bar was set by ISS. The station's life support system until 2023 returned 94% of water — urine, sweat, breath moisture cycled back into drinking water. In 2023 they added the Brine Processor Assembly, squeezing out brines, and reached 98% — losses three times smaller. This is humanity's record for closure, and even it means 2% of water goes down the drain forever.
Next comes the table worth the whole paper. At 94% the million-person city lives 33 years, at 98% — a hundred years, at 99% — two hundred, at 99.9% — two thousand. Each tenth of a percent of efficiency is decades of city life, the fate of an entire generation of lunatics packed into decimal places of plumbing. Meanwhile a city of 100,000 at the same record 98% lasts a thousand years — now that looks like "sustainable."
There's also a brutal detail the authors spell out separately: the ISS record was set by people who had neither shower nor washing machine. A permanent Moon base will want normal showers, laundries, and toilets working in one-sixth Earth gravity — and all this luxury will drag efficiency down. For a million-person city to last a thousand years needs e > 99.75% — a tenfold improvement on the record which, in the authors' phrasing, is hard not only to achieve but to maintain even for a disciplined and motivated population.
The most generous estimate of one billion tons is already the authors' second act of goodwill (the first being consumption norms). Real measurements are more modest. Meta-analysis done on 65 major polar craters gives for the eight richest repositories a total of about 34 million tons — twenty times less than the first 2009 radar estimates. Fresh 2026 work from ShadowCam imaging confirms ice in only 3.5% of cold traps — though where it exists, it could be up to 30% of regolith mass. Even estimates for Shackleton crater alone vary threefold: from 1.6 to 4.5 million tons.
And that's not all the discounts. Ice is mixed with regolith, and extracting it completely is impossible: for reference, oil fields yield 33–50% of reserves, and the lunar extraction coefficient nobody will know until they've drilled. The main blind spot is depth: spectrometers see microns, neutron methods see a meter or two, radars 10–20 meters, while regolith rubble goes tens of meters down where no orbital instrument today looks.
Summing all discounts, the press release authors write: today's best estimates are roughly 30 times lower than a billion tons. The arithmetic scales ruthlessly: time to drying shrinks by the same factor, and even a small city gets maybe ten-twenty years of water, if that. A hundred years of life for a million-person city is, strictly speaking, the upper bound of optimism, not the baseline scenario.
I'll start with the most non-obvious twist, which makes the paper worth reading. For years lunar water was sold to the public as fuel for the space age: split H₂O into hydrogen and oxygen — and refuel interplanetary ships. The audit fully vindicates this scenario: Starship's second stage carries 1200 tons of propellant, and a billion tons of water is enough to fuel almost a million rockets — three launches a day for a thousand years. The refueling business on lunar water pencils out. The municipal one doesn't: what breaks the budget isn't rockets, but residents with showers, laundry, and dinner with fresh vegetables.
The authors see four loopholes out of the water dead end. First — squeeze recycling tenfold to 99.75%, which is dubious technologically and even more dubious organizationally: one breakthrough accident in the loop eats such savings for years. Second — spend less: vertical farms demonstrate minus 97% water consumption for strawberries, and eating habits are harder to change than any engine — beef requires six times more water per kilogram than soy; the average American's "water footprint" is 2.8 tons a year versus 1.1 for an Indian with low-meat diet. The lunar city goes vegan not from ethics but from water budget.
Third loophole — import: a million-person city at record 98% needs to land 10 million tons of water per year, that's 27 landings of thousand-ton tankers daily (while a thousand tons is already ten times the cargo capacity of Starship's lunar lander). The authors politely call such a schedule "formidable but not unimaginable"; the bottleneck is the supply of accessible low-velocity water-carrier asteroids. And fourth, the only one the authors themselves bet on — find more water: all methods see only the first meters, but real ice may lie deeper, and only on-site drilling can prove it.
The paper's final conclusion sounds like a sentence on megacities: permanent habitation on the Moon above a hundred thousand people is ruled out unless one of the four loopholes works. A settlement of a thousand inhabitants — the authors compare it to Antarctic station winter crews — can live for centuries without thinking about water. A hundred thousand — time to start economizing. A million — can't sustain longer than a century. And for lunar operations to make sense for Earth's economy, they add, Bezos and Musk actually need about 80 million residents — eight hundred times more than the water balance supports.
The most piquant thing is the chronology of the water hunt unfolding right around the publication. NASA assembled the VIPER rover, which was supposed to directly measure ice at the south pole for the first time — and in July 2024 canceled the project due to cost overruns: the assembled vehicle never left the clean room, they offered it for free, then dismantled it for instruments for future missions. Humanity's main staff "water meter" died without getting a drink.
And that same August 2026 China's Chang'e-7 departed for the south pole — the most complex Chinese lunar craft to date: module, rover, and hopping probe that on six articulated legs descends into the black Shackleton crater with a water molecule analyzer, makes minimum three hops, and returns to the sunlit rim to recharge. Landing expected by November. A queue lined up nearby: the Indians already landed close by in 2023, the European PROSPECT drill and Japanese-Indian LUPEX rover are preparing to follow. Whoever first maps the ice becomes the Moon's water authority — and here the paper warns separately: water distribution can't be left to "first come, first served" rule, or conflict will flare before the craters dry up.
An honest lunar economy emerges symmetrically: data centers on "peaks of eternal light" with radiators in permanently shadowed regions — gigawatt after gigawatt, almost no water footprint. Water is needed by people and cabbage, but servers need only dark sky for cooling. Perhaps the Moon's first truly large industry will be completely dry.
This work's value lies in its demonstrative lack of hype. No new physics, no breakthroughs: budget references, consumption norms, and a two-line equation. And precisely such arithmetic does to space optimism what neither skeptics nor comedians achieved: it doesn't argue with the dream, it hands it the utility bill.
I like three levels in this story. First technical: the bottleneck for conquering the Moon isn't rockets or radiation, but a percentage point in the water loop; every decimal place in recycling efficiency is worth decades of city life, and "plumbing percentage" is the most undervalued metric of civilization's readiness for expansion. Second — economic: by water budget the Moon is an Antarctic station with a server farm, not Manhattan with kiosks, and "self-sustaining cities" rhetoric is investment round language, not accounting. Third — philosophical: the same T/(1−e) equation the authors themselves extend to interstellar generation ships; any "humanity's womb ship" project will die in the same drain as the lunar city, only the scale of disaster is stellar.
And instinct suggests the main thing: the real test of space age maturity won't happen on the launch pad, but in the room where lunar water gets divided. If humanity repeats the "first come, first served" scheme there, we deserve not an eighth continent but the most expensive drought in history. Meanwhile as Chang'e-7 hops toward Shackleton crater — for precise lunar water meter figures we'll apparently be going to Beijing.