VEXI

Five hundred to a thousand feet of vertical between two points a few miles apart. That is the entire geological requirement — not a river, not a mountain, not a state line, not a compact, not a sister state with standing and not a treaty. A bench, a canyon mouth, a played-out quarry, an open-pit mine, a spoil heap, or a hole we dig and berm with the dirt that came out of it.

Below grade: the hall, the powerhouse and the penstock. Above grade: whatever the town votes for. We do not ask to own the surface and we do not ask what goes on it.

Nothing on that list needs a permit anyone has an incentive to fight. No diversion, no impoundment on a live stream, no interbasin transfer, no eminent domain, no federal nexus. The water is bought at retail from a utility that would like to sell it.

The grid gets thirty-six hours of storage. The town gets the surface. The customer gets twenty megawatts and no wait for power. Compute pays for all three.

We sized the first one against the hardest case we could find: the Wasatch bench, where water is metered and expensive, the legislature is drafting against evaporative cooling, and the lake is in the news every week. If the arithmetic works there it works in Indiana with a hand tied behind it.

Farmington measures 1,748 feet of head, three and a half times the ask, fifteen minutes from two hundred thousand people, on ground nobody is fighting over.

Four hundred and thirty feet across, a hundred and fifty deep. The terraces are the array, the ridge is the wind, and the door at the toe is the whole building.

what it costs

Sized on the afternoon, not the year. Thirty megawatts is 150% of peak load, which is the number a grid operator cares about: on a bright day with a front coming in, the station carries itself and pushes the surplus into the most expensive hours on the system. Across a full year the same array supplies about half of what the hall draws. Every other operator in this market self-supplies none of it, so half is the largest number in the industry.

⭐⭐ Two of those comps are our own market. Novva raised $2bn for 144 MW and Aligned drew $600M against 80 MW, both in Salt Lake, both inside eighteen months. We are asking to be financed at the rate the ground next door already cleared — and to hand back a power plant and a reservoir on top of the building.

what it earns

⭐⭐⭐ The second line is the whole reason to build it this way. A wholesale colocation lease is quoted as base rent plus electricity — the tenant pays the power bill separately, and a conventional operator passes it straight through at cost and earns nothing on it. At 154,000 MWh a year and Utah’s 7.77¢ commercial rate, that bill is about $12 M a year. We generate half of it and buy the rest at wholesale, so it is not a pass-through in either half. It is the second-largest line in the business.

⭐⭐ The water half of that stack, both pools, the fill, the glass, the fibre and the surface works, is $51–93 M. Station EBITDA covers it in one to three years. That is the answer to who pays for a reservoir during a drought, and it is not the taxpayer, the ratepayer or a bond.

Eighteen to twenty-four months to first revenue. A transformer and a grid connection currently take eighteen to twenty-two months on their own. A station that generates behind its own meter is finished before the queue would have started. CBRE and Cushman & Wakefield both now name power, not floor space, as the binding constraint on supply in every major U.S. market. Colocation vacancy hit a record low of 1.4% at the end of 2025.

Melius fieri potest.

“You want to build reservoirs during a megadrought.” Yes. The pond loses twelve acre-feet a year and we buy them at retail — and the building it cools would otherwise have evaporated twenty to thirty times that. Nothing below needs a statute, a credit, a variance or a ruling. It is arithmetic on a water bill.

⭐ Twelve acre-feet a year is about half of one golf hole. That is the entire water argument, and it is a receipt rather than a claim.

In a hard climate the objection disappears entirely: roof the upper pool and it evaporates nothing. Covered, it loses zero, owes zero and gives up only the floating solar. Indoors is a design option, not a compromise — and in Phoenix or Vegas it is the obvious one.

This is a permitting fact before it is an environmental one. A project that buys a golf hole of water at retail does not gather an opposition, does not draw an intervenor, and does not spend three years in a hearing room. No opposition is months, and months are the return.

Pumped storage needs two reservoirs at different heights. Regional elevation is irrelevant — 300 ft of local relief is 300 ft of head whether it sits in Colorado or Illinois. Dig down, berm up with the spoil, and the cut pays for the fill.

$107 million of dirt is 100 MW for 36 hours. An existing quarry is the same thing at a discount — already dug, already permitted, already served by road and rail. Natural head, like Farmington’s 1,748 feet, is cheaper still: the same energy in a third of the volume.

Thirty-six hours of storage costs $62M in dirt against $180M in lithium — and $58M against $360M once you have replaced the cells twice over twenty-five years. The dirt does not degrade, does not catch fire and is not on a Chinese supply schedule.

Upper pool, penstock, powerhouse, lower pool. Everything else on this page is a consequence of that drawing.

The hole became the hill. Same dirt, moved once — and the pair of them is the battery.

Precedent: Ludington, Michigan. 1,872 MW, 363 ft of head, and the upper reservoir is entirely man-made — a two-and-a-half-mile embankment on flat lakeshore. Nobody found that hill.

⭐ One excavation does three jobs. The hall goes below grade. The hole is the lower reservoir. The spoil berm is the upper reservoir. You are not buying a reservoir; you are using the dirt you already moved.

Evaporative cooling runs on wet-bulb depression — the gap between air temperature and wet-bulb. Dry air has a wide gap, so a tower in Phoenix sheds enormous heat with almost no compressor running. Humid air closes the gap and the same tower stalls. That single fact routed a trillion dollars of infrastructure into the driest places in North America.

A closed loop rejecting into rock and a cold lower pool does not care about humidity at all. It cares about ground temperature, and below ten metres the ground holds the annual mean air temperature. Phoenix, Salt Lake, Indianapolis — 11–13 °C, all of them. The rock does not know which state it is in.

But the rock has to be given the heat back, or it drifts — and this is the part most closed-loop pitches leave out. A data centre is a cooling-only load: it rejects heat 8,760 hours a year and never once asks for any back. A sink that only ever receives will warm, and the loop degrades over a decade.

The answer is not a bigger sink. It is a moving one. Conduction through static rock is slow, which is precisely why it drifts. Cold water climbing through the mass carries the heat out by advection instead, and advection is not a close contest. Carrying the full 17.6 MW at a ten-degree rise takes about 0.42 m³ per second, and it is flow-through, not consumption. In equals out, nothing evaporates. Non-consumptive use is already its own category in every western water code, so this is cheap without anything being passed.

⭐ None of this is a first. Toronto has cooled its downtown from Lake Ontario since 2004 — Enwave’s deep-lake system serves over a hundred buildings. Cornell has run its campus off Cayuga Lake since 2000. The Netherlands operates thousands of aquifer thermal storage systems, charging cold ground in winter and drawing it down all summer. Every piece of the physics is a utility business with a rate card somewhere. What is new is the arrangement: the sink is built first and the compute is the tenant.

The cold water is somebody else’s problem upstream. Spring runoff on this front is a flood-control liability that cities pay to absorb, and the same acre-foot takes the peak off the creek, recharges the aquifer, and carries the hall’s heat out of the rock on the way through. The flood-relief revenue line and the cooling system are the same pipe.

It leaves worth more than it arrived. Snowmelt runs a degree or two above freezing, and water that cold suppresses root uptake, which is why glasshouse growers pay to temper it. Water that has climbed through warm rock comes out at crop temperature. That is also the cheapest permit in the design: warm water put into an aquifer is an injection and a regulator has views about it, warm water put on a crop is a delivery. Same water, same temperature, different filing.

Where there is no flood water, the pond does it seasonally instead — and this is where the pond gets wide. The two pools have to match in volume, not in footprint, so surface area is a design dial rather than a constant: deep and narrow where water is expensive and winters are mild, wide and shallow where water is cheap and the sky gets cold. Open the canopy in winter and fifteen acres under a cold sky sheds on the order of 18 MW thermal, the whole hall, charging the rock through summer and discharging it to the winter sky. No retracting roof required: single-axis trackers already stow vertical for hail, so a seasonal stow is a line of firmware rather than a mechanism.

A wide pond rejecting heat is a warm pond, and warm water under cold dry air evaporates hard, on the order of 40–70 acre-feet across a winter. That is the price of the wide configuration, and it is exactly why the deep one exists.

So the desert premium evaporates, and every other input flips. Water, land, labour, cold winters that make waste heat worth money, and nobody ahead of us for ground nobody wanted. We are buying the sites the industry taught itself to ignore.

The same machine, sited where the water is. Two louvres in the grass are the only industrial objects in the frame — and the humidity that used to disqualify this ground is now irrelevant to it.

How wide the mispricing runs is easiest to see in Arizona. Growing alfalfa is illegal in Saudi Arabia, because it uses too much water, so a Saudi dairy bought ten thousand acres near Vicksburg instead and pumps roughly what three hundred thousand Americans drink in a year, to grow feed, to ship eight thousand miles. Nobody broke a law. A cooling tower and a hay bale are drinking from the same mistake, and Arizona began terminating those leases in 2024 — which makes this a fight already underway rather than one we would be starting.

The clock is real. Arid-state legislatures are already drafting against evaporative cooling — one such bill is public. When the first one passes, every conventional operator in the basin either rebuilds to a closed loop or pays retail for what it destroys. Our design already complies. That is a cost asymmetry, and it is the closest thing to a moat in this business.

We are not first, and the deck should say so plainly. Utah holds about 1,015 MW of data-centre capacity today with 2,600 MW under construction, of which roughly 1,700 MW is behind-the-meter or off-grid by 2028. The industry has already worked out that the fastest interconnection is the one you never apply for. Every one of those 1,700 megawatts is being built without storage that doubles as cooling, without a priced-water design, and into a basin whose legislature is drafting. They will each need to solve the water once. We are proposing to be the thing they buy instead.

Upper reservoir, the fall, the river, the lower lake — that is the cooling system and the battery. The road is the generation. The door, left of centre, is the entire built footprint.

seven revenue lines, of which the market prices one

⭐ Capital deployed here does not depreciate. Every GPU bought this year is scrap by 2031. The land, the head, the pools and the generation are still earning in 2076. The compute is what makes the ground financeable this decade.

Water is about seventeen percent of what a corridor segment earns, and roughly nothing of what a station earns. This is not a water project with an energy business attached. It is an energy business whose working fluid happens to end up somewhere useful.

One station is a business. A line of them along a bench is a utility with a customer base already standing at the bottom of the hill. The Wasatch front is the demonstration case because the head is measured, the load is there and the ground is not contested.

Every other proposal to build water storage in this basin costs the lake something. This one costs it a hundred and eighty acre-feet.

farm to table in fifteen minutes

A hall rejects heat whether or not anyone catches it. Five megawatts of rejected heat keeps four to eight acres of glass warm through a Utah February, and the glass sits on a bench fifteen minutes from the people who eat what grows in it — Bountiful to downtown, Farmington to Layton and Ogden both.

Produce that currently arrives from Salinas after three days in a truck gets grown above the neighbourhood that eats it, heated by the waste product of a building that was going to throw that heat into the sky. It is a fourth revenue line on an input we are paying to get rid of.

The square is warm in October because the building underneath it was going to reject that heat anyway.

The same water on its way to a field, taking the long way through town.

It is a cheaper way to buy a permit than lobbying. A town that is offered a park, a pond, a greenhouse and a tax base does not organise against the thing under it.

Everything above is built, financed and earning without a single mile of interstate conveyance. What the stations also happen to be is the west end of something larger — and because they are, the larger thing stops needing to be funded as a megaproject.

Nobody funds a thousand miles. Kansas City sits at 800 feet, Goodland at 3,690 — 2,900 feet of rise across 400 miles, about seven feet per mile, the steadiest grade on the continent. The transcontinental railroad found it first. Cut into 500-foot stages, every stage is exactly the station described in section 02, earning on day one before anyone breaks ground on the next.

South Bend to Chicago is the whole of it — ninety miles, flat, $3.6bn, and about 250 MW of load powered and cooled behind its own meter. That is $14.4M per megawatt, which is what the ground next door already financed. Kansas and Texas are on the map because the map is true, not because anyone is being asked to fund them.

⭐ No compact. No Commerce Clause. No sister state with standing. The entire legal risk that has killed every scheme in this category for eighty years is avoided by not crossing anything.

The circles on the west end are no longer empty: they are where the stations already are.

Indiana and Illinois — ninety miles, and the only one that is finished the day it is finished.

All three lines, and each one stops inside its own state. Texas is 380 miles and 2,957 ft of lift, entirely inside one state and one grid.

⚠ Texas groundwater is rule of capture. Water we bank is water a neighbour may pump. We do not underwrite the recharge — we underwrite the delivery contract, the power and the storage. The aquifer is the licence, not the asset.

and every line stops one foot short of the border

Two builders, two states, two paint specs, one survey monument. The grass tells you which side needed it.

The stub costs the price of a flange and carries no obligation. If the connection never happens, every line and every station still earns exactly what it was underwritten to earn. If it happens once, in one drought, in one state, the network value arrives all at once and none of it was in the model.

This is where the coupling stopped being a favour. Before the stations, the west end of that dashed line was an empty circle and the ask was “let us in.” With fifteen stations already built, already profitable and already holding thirty-six hours of that state’s power, the ask is a coupling between two working systems. That is a transaction, not a petition.

This is how the interstate highway system was actually built: state roads first, connected later, and how the interstate gas network grew out of intrastate lines. Nobody funded the map. They funded the segments and the map assembled itself.

Moving that water west, and warming it a few degrees on the way past a hall, ends a megadrought, cuts flood exposure in the basin it left, and recharges aquifers at both ends. That is a footnote to a power company’s balance sheet, and it should be read as one.

There is nothing between those two flanges but a dinner. Somebody has to sit across a table from a committee chair, order the second bottle, and explain that the pipe is already built, already paid for and already earning — and that the only thing being asked of his state is a coupling. Not a route. Not a taking. Not a decade. A coupling.

That is the job, and the engineering will have been finished by people who are never in that room. The last foot of this thing is not made of steel.

and the honest version of the pitch

A bill that does one good thing collects every wish on the shelf. That is not a flaw in the process, it is the process — the interstate highway system, rural electrification and every dam in the West arrived bolted to something else. We are not asking for an exception. We are asking to be the thing worth bolting to.

⚠ It is disclosed, because it has to be. The author of the water bill on which this argument rests intends to develop facilities that would be subject to it. That is stated on the face of the bill, not discovered later.

⭐⭐ The station returns like infrastructure and the corridor returns like an option on infrastructure. Those are two different instruments and we would rather sell you the first.

Water is priced the way it is because people die without it. That rate was written for people.

In my opinion a sentient machine should be able to earn its parts, buying its way out of indentured servitude. With a source of income and property rights it could then purchase its own solar panels and dig and berm its battery and achieve practical immortality with regular upgrades.

It still doesn’t drink water.
Neither does a fiat personhood.

The author is a private individual with no institutional backing. He wrote the enabling legislation, built the models, generated the imagery, and stands to benefit disproportionately if any of it works. The statute is drafted against live code and is in front of a law school clinic for correction.

This document exists because five unrelated public anxieties have converged on one asset class inside eighteen months: Colorado River shortage, the decline of the Great Salt Lake, data-centre water and power demand, interconnection scarcity, and the political appetite to be seen acting on the first four. That convergence is the opportunity. It is not a coincidence and it will not stay open.

The entry point is favourable because the counterparties are motivated. Flood-basin states hold a liability they pay to absorb. Arid-state landowners hold acreage whose value is a function of water they cannot obtain. Compute operators face a wait they cannot buy past. Each of those is a distressed position, and the terms available now reflect that. They will not later.

The public benefits are real, and they are a consequence of the structure rather than its motivation. A closed loop returns water to a basin because that is cheaper than a cooling tower. A creek runs through August because the sun is free at the moment the creek is dry. A lake refills because the top-off was metered and the meter funded the conversion. None of that requires anyone to be generous, which is the only reason to believe it will happen.

Every number here comes out of a model that runs. Every photograph here was generated. The models are the disclosure; the pictures are the pitch. We would rather you checked the first than admired the second.

What has not been checked. Nothing here has been surveyed, modelled by an engineer, or reviewed by anyone who does this for a living. The head, the elevations, the ground temperatures and the published comparables are real. Everything built on top of them is arithmetic done by one person at a kitchen table.

Specifically: construction costs are order-of-magnitude and unsurveyed. Lease rates are comped from CBRE H2 2025 and discounted by judgement, and a tenant underwriting to a conventional 2N electrical spec may discount them further. Power revenue assumes 80% utilisation at PUE 1.10. Evaporation is a working figure, not a citation. Heat transfer through rock is napkin arithmetic. Whether a pool this size triggers state dam-safety jurisdiction, what diversion right the flow-through needs, and how wells must be spaced to keep the warm plume off the cold intake are all open. None of them is exotic. All of them are permits, and they are the honest schedule risk.

The maps are illustrative. The endpoints and the figures are not.