A chimney several hundred feet tall, a car radiator, and a camel: the physics of making 17 gigawatts in the Panhandle without draining it.
There is a saying, often attributed to Einstein, that if you cannot explain something simply, you do not understand it well enough. Our cooling system has lately been the subject of long reports full of borrowed tables. I will take the other route and explain it the way I would to my neighbors in Amarillo, because they are the ones who actually own this question.
Water comes first
In the Texas Panhandle everything starts with water. We are a Texas company. Our people ranch here, farm here, and raise kids here on top of the Ogallala aquifer. So when we designed the largest private energy campus in America, water was not a line item to optimize at the end. It was the first requirement we wrote down: build 17 gigawatts with minimal effect on the aquifer. The towers in this article serve the steam side of the campus: the combined cycle natural gas plants and, later, the reactors. Natural gas turbines running in simple cycle make no steam, so they use no cooling towers and next to no water. The solar and battery blocks that round out the 17 gigawatts need essentially no cooling water.
That one requirement settled our cooling technology before any consultant could weigh in. A conventional wet-cooled plant makes electricity by evaporating a river. The two newest reactors in Georgia are permitted to draw up to 43,000 gallons per minute from the Savannah River for cooling, and most of what they draw leaves as vapor. Scale that approach to our four reactors and you arrive at roughly 124 million gallons a day at peak. Permitting that in the Panhandle would be wrong even if it were possible, and it is not possible. The precise gallons belong in our permits and filings, where we publish them with their assumptions, and they will evolve as the campus grows toward 17 gigawatts. The shape of the answer will not change: on an average day the campus sips, on the hottest afternoons it drinks from its own reservoirs filled on mild days, and the draw on the region stays small and steady by design. For scale, wet cooling for the four reactors alone would demand roughly 124 million gallons a day at peak. The entire campus, reactors, natural gas plants, and data centers together, is designed to run on a small fraction of that. Keep that in mind when we get to cost, because the plant we supposedly should have priced against is a plant that cannot exist here.
The chimney: a machine whose moving part is air
Every steam plant on earth, whether it burns natural gas or splits atoms, has the same final step: after steam spins the turbine, you must cool it back into water and send it around again. The only question is where the heat goes. Wet towers dump it into evaporating water. Air-cooled condensers, called ACCs, push turbine steam through huge ducts and blast it with dozens of large fans, all day, every day. We chose something older and more elegant: a hyperbolic tower several hundred feet tall that lets the sky do the work.
The trick is simple. Warm air is lighter than cool air, so it rises. Build the tower tall enough and shape it correctly, and it becomes a chimney that pulls its own wind: Panhandle air rushes in at the bottom, sweeps past the heat, and floats out the top. No fans in the draft path. Nothing pressurized. The best fan is the one you never install, and the West Texas wind shows up for work free of charge. In fact, on most days our challenge is the opposite one, and motorized louvers around the base restrict the airflow so the tower does not overcool.
The radiator: why the tower does not drink
Now the part that makes it dry. Around the bottom hundred feet of the tower we wrap a ring of steel cooling grills. If you have ever opened the hood of your car, you already understand our design. Your engine is cooled by liquid running through a radiator in a sealed loop, and the breeze from driving carries the heat away. You do not pour in new coolant every morning. Filled once, reused for years.
Our tower is exactly that, at Texas scale. Warm water from the plant condenser runs through the radiator ring in a closed loop, the tower’s self-made wind blows across the fins, and the cooled water heads back for another lap. The loop does not evaporate and does not get topped up. On a normal day the tower provides gigawatts of cooling and drinks nothing. I have walked a tower whose grills have been in continuous service for forty years, original equipment. This is not exotic hardware. It is plumbing with excellent posture.
The hot day: sixteen helpers and a 95-degree rule
So when does water enter the picture? Picture driving across the Arizona desert in August. At 110 degrees, your radiator is trying to shed heat into air nearly as hot as the coolant, and it struggles. Every dry cooling system faces that limit: the hotter the air, the less heat it can carry away.
Amarillo hands us a limited number of afternoons like that each year. For those hours, each tower carries sixteen water-to-water heat exchangers, a wet section that sleeps most of the year. Below 95 degrees ambient, they stay off and the tower runs entirely dry, on a zero-water closed circuit. Above 95 degrees, they wake up, add evaporative cooling to the loop, and hold the plant at full performance through the heat of the afternoon. Then the sun drops, the Panhandle cools, and they go back to sleep. We expect them to work on the order of 35 days a year, a few hours at a time. That is the only time the system consumes water, through evaporation and the routine housekeeping of the wet loop. Even then, the tower is designed to show no visible plume. Drive by on the hottest day of the year and the only thing rising off the top is shimmer.
The camel: where hot-day water comes from
Even those few wet afternoons do not reach into anyone’s well. On the mild days, which is most of the year, our supply lines quietly fill large on-site reservoirs. On hot afternoons the wet sections drink from that stored water, the way a camel draws on its hump. The draw on outside supply stays flat and boring by design, and we are engineering the storage ponds to be covered so evaporation does not steal the savings account. The reservoirs ride the peaks, and the aquifer barely notices the weather.
Does it cost us power?
Remarkably little, and this is the point most analyses miss. Vendor engineering for a 1,250 MW nuclear unit shows that a properly sized dry-wet tower delivers about 99.9 percent of the annual generation of a conventional all-wet tower while using about 15 percent of the water. That is the entire trade: give up roughly a tenth of one percent of annual output, keep 85 percent of the water in the ground. The wet section exists precisely so that hot days do not cost megawatts. We are also interconnecting the towers across the site, so cooling capacity freed by any unit that is down for maintenance or running in simple cycle can help its neighbors on a scorcher. This system was sized for August in Texas, not for a brochure.
About that briefing note
A July 2026 briefing note from IEEFA argues our cooling will be ruinously expensive compared to wet cooling, and that our reactors face a novel licensing barrier. Both claims dissolve on contact with the details.
On cost, the note prices our towers against a full wet cooling system. As covered above, a wet system for four reactors needs on the order of 124 million gallons a day at peak. That water does not exist here to permit, which the note itself concedes when it calls our region water-stressed. You cannot declare the Panhandle out of water and then bill us for the water-cooled plant we refused to build. Our actual engineering choice, the one on our books, was between the two dry-first options: ACC or hybrid tower. We studied both with vendors and with our FEED partner Hyundai E&C, against the same old Vogtle ACC study the note leans on. The ACC needs dozens of large fans running around the clock, a permanent tax on our own electricity many times larger than the tower’s. It forces a redesign of the turbine hall, and it stresses the turbine on the hottest afternoons, exactly when Texans need every megawatt. And it is a one-way street: an ACC can never learn to use water on the worst days, while our tower can flex. Up front the two options cost about the same. Over the life of the plant, the tower wins on any honest math. Choosing the better machine at a comparable price is not a cautionary tale. It is procurement.
It is also worth saying plainly what the note’s own sources show: the studies it cites against us evaluated direct air-cooled condensers, a different machine from our towers. In our system the turbine never meets the air. It exhausts into a conventional water-cooled condenser, the same interface it has at Vogtle, and the closed water loop carries the heat to the tower. The famous conclusion that air cooling was “not technically feasible” at Vogtle was a verdict on ACCs, rendered nearly two decades ago. It says nothing about a Heller-type hybrid tower, and its cost tables say even less.
On licensing, the note’s subtitle announces that the AP1000 “is not licensed to operate with hybrid cooling,” which sounds ominous and means very little. The reactor, the nuclear island, and the safety systems are standardized, and we are copying the operating unit at Vogtle as faithfully as we can. We did not design the risk in; we copied it out. But heat rejection has always been site-specific on every nuclear plant ever built. Vogtle cools with wet towers drawing on the Savannah. The AP1000s operating in China cool with seawater. Same certified reactor, different sites. Our cooling design is reviewed by the NRC within the combined license application we already have under active review, the first new one in fifteen years. That is the normal path, not a new gate. And one more fact the note never mentions: the AP1000’s safety cooling is passive. Gravity, natural circulation, and a large water tank above the containment protect the reactor. Plant safety does not depend on our towers at all. The towers are how we make money, not how we stay safe.
Proven where water is precious
Is the technology new? It was proven before I was born. Closed-loop indirect dry cooling has been in commercial service since the 1950s, and it runs today at hundreds of installations across Asia, the Middle East, and Europe. Among them: the world’s largest dry-cooled combined cycle plant in Türkiye at 2,310 MW, coal units of 660 to 1,000 MW across China, and towers throughout Europe, where water use is watched the way we watch it here. Engineers call it the Heller system, after the professor who invented it, and it is the technology at the heart of our December agreement with MVM EGI, the company that has designed and refined these towers for decades. We did not stumble onto any of this. We went looking for the most water-frugal cooling technology operating anywhere in the world, walked its towers in person, and brought it home to Texas. I have personally walked inside a hyperbolic hybrid tower serving a 1.2 GW natural gas plant in Türkiye. In Germany, the 1,400 MW Neckarwestheim Unit 2 reactor ran for 34 years on a hybrid wet-dry cooling tower until the country retired its nuclear fleet in 2023. In South Africa, the coal-fired Kendal station generates over 4,000 MW with indirect dry cooling, the largest such plant on earth, using a small fraction of the water of a wet-cooled plant. Even the tiny Arctic reactor that critics cite as the only air-cooled nuclear plant in history runs a Heller-family system, and on warm days even it sips a little spray water. The exception knows the rule.
America is late to this not because our engineers could not do it, but because our geography spoiled us. We had mighty rivers and cheap water, so we boiled them. Water is not cheap anymore, anywhere. The physics is proven at scale around the world, and now it goes to work in Texas.
Starting now, not someday
This is not a promise about the 2030s. The first hyperbolic hybrid towers at Matador will serve our large F-class combined cycle natural gas units, running dry the overwhelming majority of the year from first fire. The reactors follow on the same architecture. And I am confident the same closed-loop, air-first principle will end up cooling data centers as well, for the same reason: compute, like steam, is ultimately a heat problem, and the sky is the largest heat sink we own.
A suggestion for the American energy industry
Here is the part that is bigger than our project. If you are building a thermal power plant anywhere in the United States, even beside a lake or a great river like the Savannah, consider the hybrid tower anyway. The Vogtle engineers, to whom every American nuclear builder owes a debt, were handed a menu in the late 2000s that read “wet tower or ACC,” and they chose correctly from that menu. The trouble is that the winning option was barely on it. Beside a river, you might install 24 heat exchanger cells instead of our 16 and let the wet section work more often. You would give up a rounding error of annual output and return billions of gallons a year to the river, the fish, the farmers, and the state. Cooling has been treated as the boring last chapter of power plant design for a century. In a country that intends to lead the world in AI and advanced manufacturing, water is strategic infrastructure, and the boring chapter just became a headline.
There is a quiet bonus for the nation, too. A plant that barely needs water can be built in far more places. You stop competing with cities and agriculture for riverbanks, and you start putting gigawatts where the land is affordable, the wind is free, and the transmission makes sense. Places like the Panhandle. Keep the lakes for bass and weekends.
The pledge
Fermi America is named for Enrico Fermi, who built the first reactor under a Chicago squash court with graphite bricks and nerve. Our contribution to that lineage is more modest and very Texan: we intend to cool the atom with wind, keep the Ogallala where it belongs, and sell the megawatts to the people building America’s future. We will keep publishing our numbers as the engineering matures, and our neighbors are welcome to hold us to every one of them.
If you cannot explain it simply, you do not understand it. If you cannot defend it simply, you should not build it. We can do both.
Technical Annex
For readers who want the engineering and regulatory detail behind the plain-English explanation above: the claims made in the July 2026 IEEFA briefing note, answered one by one, followed by the codes, standards, and regulatory framework that govern this work. Sources cited are public unless noted.
A. The briefing note’s claims, answered
B. Codes, standards, and regulatory framework
Nuclear licensing: 10 CFR Part 52. Project Matador Nuclear Units 1-4 are the subject of a combined license application accepted for NRC review in September 2025, the first new large-reactor COLA under active review in fifteen years. The site-specific Final Safety Analysis Report content, including the circulating water system in Chapter 10, is reviewed against NUREG-0800 (Standard Review Plan), Section 10.4.5.
Departures from the certified design: 10 CFR Part 52, Appendix D, Section VIII. Where the site-specific design departs from the AP1000 design control document, the established departure and exemption process applies. This is the routine mechanism by which every COL applicant adapts balance-of-plant systems to its site.
Environmental review: 10 CFR Part 51. Cooling water use, consumption, and thermal effects are evaluated in the Environmental Report and the NRC’s environmental impact statement. Fermi is executing the NRC’s first applicant-prepared EIS pilot, and cooling system water use is addressed within it.
Safety classification. Safety-related structures, systems, and components of the nuclear island are procured and constructed under 10 CFR 50 Appendix B and ASME NQA-1 quality assurance programs. The cooling towers and circulating water system are non-safety-related power-conversion equipment; the AP1000’s licensed safety case rests on its passive containment cooling system and does not credit the towers.
Structural design of the towers. Reinforced concrete hyperbolic shells are designed to ACI 318 with the guidance of ACI 334.2R (reinforced concrete cooling tower shells) and the international cooling tower guideline VGB-R 610, with wind and seismic loads per ASCE 7 and the applicable building code.
Thermal performance verification. Tower thermal performance is verified by acceptance testing per Cooling Technology Institute ATC-105 and ASME PTC 23 (atmospheric water cooling equipment).
Aviation: 14 CFR Part 77. Structures exceeding 200 feet require notice to the FAA for obstruction evaluation and marking/lighting determinations, a routine filing for tall stacks and towers nationwide.
Texas water. Groundwater use is permitted through the applicable groundwater conservation district, and any discharge is regulated by the Texas Commission on Environmental Quality. On-site reservoir storage is sized so that peak-day cooling demand is met from storage rather than increased instantaneous draw.
About Fermi America
Fermi Inc. (d/b/a Fermi America) (Nasdaq: FRMI), in partnership with the Texas Tech University System, is developing Project Matador, a private energy campus in Carson County, Texas now projected at 17 GW, combining natural gas generation, four Westinghouse AP1000 reactors currently under combined license review by the U.S. Nuclear Regulatory Commission, small modular reactors, solar generation with battery storage, and data center infrastructure.
Media Contact: Brian Magallanes | Fermi America | Brian@fermiamerica.com
