Our View
The Power Plant Needs Water, Too
The closed-loop answer describes the servers. The turbines have never been described at all.
Brian O’Neill’s application for 900 Conshohocken Road states that the data center “will not draw water from the adjacent Schuylkill River nor from on-site wells,” and describes a closed-loop cooling system. That describes the server halls. The same proposal includes an on-site natural gas power plant, a separate thermal system with a separate water balance, and no public document describes it.
The single fact that would settle most of the question is whether the turbines are simple cycle or combined cycle. A simple-cycle plant has no steam to condense and needs very little water. A combined-cycle plant adds a steam turbine and a condenser, and if that condenser is served by an evaporative cooling tower, the annual figure is substantial. Air-cooled condensers avoid most of it at higher cost. Which configuration is proposed has not been disclosed, and neither has the megawatt total or the turbine count.
None of this requires assuming the project will consume a lot of water. It may consume very little. The point is narrower and harder to argue with: until the power-generation configuration is on the record, neither the Zoning Hearing Board nor the public can independently evaluate the facility’s total water demand.
What the record says
The May 2026 application states that the facility will not draw water from the adjacent Schuylkill River nor from on-site wells. On power, O’Neill is proposing to generate electricity on site using natural gas turbines, which he says means the facility would not draw from the grid, with backup generators in place to fill any gap when the system goes down.
What the record does not contain is the size or design of that power plant. The project’s total power draw is not public, and neither are its water and air-emissions studies. A closed-loop representation about the data halls has been offered as the complete answer on water, while the machine running continuously beside those halls has never been described well enough for anyone to check whether it needs water at all.
The question that decides most of the answer
Gas turbines by themselves are not thirsty. As the Union of Concerned Scientists explains, gas combustion turbines require no cooling because they have no steam to condense, which is why combined-cycle systems need much less water than traditional steam turbine technologies, and why dry cooling is more economical for gas plants than for other thermoelectric options.
A combined-cycle plant is a different proposition, because it captures exhaust heat to run a steam turbine, and that steam has to be condensed somewhere. Cooling water from an evaporative cooling tower is typically used to cool the condenser, and water consumption can be greatly reduced by replacing the evaporative tower with air-cooled condensers, though at higher capital cost. Federal data gives the scale: in 2021, natural gas combined-cycle generation averaged a water-withdrawal intensity of 2,803 gallons per megawatt-hour.
So the same phrase, on-site natural gas generation, covers designs that need almost no water and designs that need a great deal, and the difference turns on equipment selections the applicant has already made. Multiplied by an undisclosed megawatt figure running around the clock, the range is enormous, which is why the number belongs on the record rather than in an estimate assembled by residents.
The smaller variables
Turbines lose output as intake air warms, and some installations condition the inlet air to compensate. In evaporative coolers, water is sprayed over media and the latent heat of vaporization provides sensible cooling of the air, with water requirements ranging from 30 to 80 gallons per minute. Inlet fogging uses more still. Plenty of installations use neither.
Emissions control is similar. Some turbine designs still use water or steam injection to limit NOx formation, and that consumes treated water whenever the unit runs, while many current machines reach the same limits using dry low-emission combustion and no water at all. Both are ordinary. The public record does not say which is proposed here.
Closed loop describes two different machines
“Closed-loop” is also not a defined term. The Uptime Institute is explicit that water is consumed in the heat rejection loop, and that in open cooling towers as well as evaporative and adiabatic coolers, water absorbs heat from the data center as it changes to vapor. It then describes a large 25 MW data center using a closed-loop adiabatic evaporative system, which in cooler climates may rely on free cooling for 90 to 95 percent of the year, with heat rejection requiring water only on the hottest days, sometimes for just 4 to 12 hours per day over 5 to 20 days per year. That is a closed loop, and it evaporates water.
Fully dry designs exist and are marketed under the same phrase. Vantage Data Centers describes systems designed to run without the use of evaporation, coupling dry heat rejection with liquid-cooled servers to achieve near-zero water usage, and Oracle describes closed-loop, non-evaporative systems with no evaporation, blowdown, or continuous makeup water requirement.
The useful question is whether this system ever switches into evaporative operation, and if so, how many gallons are added each year to replace what evaporates and what is drained during maintenance.
If not the river and not wells, then where
Ruling out the Schuylkill and on-site wells leaves a public supplier. Aqua Pennsylvania serves parts of Montgomery County, drawing on sources that include the Schuylkill and Delaware rivers, the Upper Merion Quarry, and more than 100 deep wells. Water arriving through a pipe is still water taken from somewhere.
The standard check is a will-serve letter, in which the utility confirms it can supply the projected volume without degrading service to existing customers. Lowhill Township now requires a water feasibility study demonstrating adequate capacity and supply to serve the facility. Plymouth has no such requirement, and no such letter has been made public.
Where the water goes afterward
Evaporative operation produces a discharge stream as well as a demand. Concentrated water has to be bled off, and the loop chemistry travels with it. In its own briefing on data centers, the Delaware River Basin Commission flagged discharges with increased temperatures, increased chlorides and salinity, and the presence of antimicrobial and anti-fouling agents in closed loop systems that might require pretreatment if conveyed to municipal wastewater systems.
That is not a hypothetical concern here. Plymouth Township Council’s conditions included a ban on withdrawing from or discharging into the Schuylkill River. Provo, Utah handled the same issue by requiring that water from the closed-loop system be removed by a third party and not discharged into city systems. Either approach requires knowing the volume first.
Thresholds mark disclosure points
Three regulatory triggers apply, and none of them establishes that this project exceeds anything. They establish when an applicant has to disclose quantities and seek approval.
DRBC maintains a distinct application for electric generating or cogenerating facilities designed to consumptively use in excess of 100,000 gallons per day during any consecutive 30-day period, which attaches to the power plant regardless of where the water comes from. Groundwater here is held to a stricter standard than the rest of the basin, because DRBC requires that new or expanded well water projects inside the Southeastern Pennsylvania Ground Water Protected Area involving an average withdrawal of more than 10,000 gallons per day must obtain a Protected Area Permit, a stricter review threshold than the 100,000 gpd that applies in the remainder of the Delaware River Basin, and it lists Montgomery County in its entirety inside that area. State registration under 25 Pa. Code Chapter 110 applies to any total withdrawal exceeding an average of 10,000 gallons per day in any 30-day period, and separately to water obtained through an interconnection with another person exceeding an average of 100,000 gallons per day in any 30-day period.
If projected use crosses those lines, filings should already identify the quantities. If no filing is required yet, the applicant can still produce the engineering estimates the design was built on, because those estimates exist either way.
Testimony is not a condition
A representation from the witness table binds nobody unless the Board writes it into the decision with a definition, a measured quantity, a reporting obligation, and a consequence for exceeding it. Provo’s planning commission did this by recommending approval on conditions that would not allow onsite primary power generation, require a closed-loop water system, and guarantee water usage does not exceed 4,380,000 gallons per year. A number, not an adjective. Plymouth Township Council attempted something in that spirit and got nowhere, and Council says O’Neill rejected all 43 conditions and then filed a legal challenge to the township’s zoning ordinance itself.
Why this reaches past water
Cooling selection drives more than gallons. It determines equipment footprint and site layout, the size and acoustic character of the air-moving equipment, whether visible plume forms above the buildings, operating efficiency, and which permits are required. Without the thermal design, none of that can be evaluated independently, which matters for a site where homes sit roughly 200 feet from the fence line.
The questions for August 6
- How many megawatts of on-site generation, across how many units?
- Simple cycle or combined cycle?
- If combined cycle, is the condenser air-cooled or served by a cooling tower?
- Do any units use evaporative inlet cooling, inlet fogging, or water or steam injection?
- What is the expected annual water use of the power plant, stated separately from the data halls?
- Does the server cooling system ever switch into evaporative operation, and on how many days a year?
- How many gallons per year are expected to be added to replace evaporation and maintenance losses?
- Where does the discharge go, and who has confirmed capacity to supply and to receive it?
None of these calls for speculation. Every commercial power plant is designed around specific equipment selections, a water balance, and a set of operating assumptions, and those calculations already exist in the applicant’s engineering file. The open question is whether they will be disclosed before an approval is granted.