For months, one of the numbers attached to Spartanburg’s new data center was 459,000 gallons of water per day. This week, NorthMark Compute and Cloud said the Valara facility at the former Kohler site on South Pine Street is projected to use an average of 1.2 million gallons per day at full buildout, with a maximum of 1.6 million.
The data center didn’t suddenly get larger over the weekend. The pipes didn’t widen on Sunday night. What changed was the public object.
NorthMark now says the 459,000-gallon figure came from a Phase I wastewater survey and didn’t represent full-buildout water demand. It also says it provided the larger figures to Spartanburg Water on February 11 as part of the project’s Large Water Meter Questionnaire. In its own public correction, the company put the matter plainly: people should be able to rely on the information it provides, and it is now revalidating other numbers it has shared publicly.
That deserves to be taken seriously without turning immediately into an accusation about motive. But it also makes the story harder than a simple case of one estimate replacing another. Different numbers were moving through different parts of the public process at the same time, and the smaller one remained the number many of us used to understand the project’s water demand.
By May, while Spartanburg County was in severe drought, Spartanburg Water was asking customers to voluntarily reduce non-essential use. In August, at the county fact-finding meeting that became the basis for my own sermon about this project, Spartanburg Water’s representatives still put the figure at roughly 459,000 gallons a day. The utility also said then that it could serve the project without harming current customers.
All of those facts belong together. The capacity question matters. So does the drought. So does the February questionnaire. So does the number that remained in public circulation months later.
The question isn’t simply whether one number was right and another wrong. It is how a project becomes publicly knowable through numbers, who has which numbers when, and what happens when the picture changes.
A Number I Carried Into the Pulpit
I have a particular reason for caring about this. A month ago, I stood in the pulpit at the Unitarian Universalist Church of Spartanburg and preached about this same project. The sermon was called What We Ask of the Water, and I used the figure Spartanburg Water’s own people had put on the record at the county’s August fact-finding meeting: about 459,000 gallons per day.
That morning, I was careful to say that 459,000 gallons wasn’t a scandal by itself. It was a number. The harder questions were what we were asking of the watershed, who was in the room when that request was made, and whether having the technical capacity to supply the water settled what we ought to ask of a place.
Now I know that the number I carried into the pulpit described something more limited than I understood it to describe. NorthMark says the larger full-buildout projection had already been supplied to Spartanburg Water in February. Yet in August, 459,000 gallons was still the number that became publicly visible to me and to anyone else listening to that meeting.
That changes how I understand what happened. It doesn’t require me to decide that somebody intended to mislead the public. It does require me to notice that the flow of information was broken somewhere, and that the break mattered.
The watershed didn’t change. My view of what we were asking of it did.
The Sunday before that sermon, in The First Home Is a Watershed, I had started with an even simpler idea that also runs through Spartanburg Watershed and What the Basin Was Aiming At. A property map and a watershed can describe the same acre differently. One tells us where ownership begins and ends. The other follows the slope and shows where the water goes.
Maybe numbers work like that too. A number isn’t the thing it measures. It makes some feature of the thing available to us while leaving other features outside the frame. Change the question and another number appears. Change the scale and another set of relationships comes into view.
The important question, then, isn’t only whether the number is accurate. We also need to know what kind of number it is, what question it answers, who has it, and what remains beyond it.
What Kind of Number Is It?
Large infrastructure projects arrive in public life through numbers: megawatts, gallons, acres, jobs, tax revenue, tons of emissions, decibels, setbacks, permit thresholds. But those numbers aren’t merely different views of the same thing. They are different kinds of claims.
The 1.2 million gallons is a projection, an estimate of what the completed facility is expected to require on an average day. A permit limit is different: it sets a boundary within which a facility is legally required to operate. An air-dispersion result is different again: it is a model output, a calculation of what pollutants are expected to do under specified assumptions and conditions. A regulatory threshold marks the point at which one legal category becomes another. And when a county writes 65 megawatts into an ordinance, that number is not a measurement. It is a line drawn through a public decision.
Valara’s newly approved air permit makes the differences unusually visible. The South Carolina Department of Environmental Services has approved eleven additional natural-gas turbines (five rated at 17 megawatts and six at 54 megawatts) with pollution-control systems. According to SCDES figures reported by WYFF, estimated uncontrolled formaldehyde emissions from the permitted configuration would be about 117.6 tons per year. With the required controls, the estimate is about 9.2 tons per year, below the 10-ton-per-year threshold for a single hazardous air pollutant used in the facility’s synthetic-minor classification. SCDES also says its dispersion modeling places the maximum modeled formaldehyde concentration below the applicable standard.
We need those distinctions. A projection tells us what someone expects. A permit tells us what someone is allowed to do. A model tells us what is expected to happen if its assumptions hold. A threshold tells us when a regulatory category changes. Confusing them makes the project less legible, not more.
The same is true of scale. Valara already had a permit for twenty-four permanent natural-gas generators, and the eleven newly approved turbines add roughly 409 megawatts of generation capacity. Taken together, the site’s permitted on-site generation is about 457 megawatts. So here a megawatt does not lead mainly toward some distant abstraction called “the grid.” It leads directly to natural gas, combustion equipment, pollution controls, permits, and the air over South Pine Street.
The numbers are real. The project is also more than any one of them.
The Problem of the Frame
I’ve been thinking a lot lately about what happens when something becomes visible, but the categories available to us are too small to receive it. That was part of the question in When Seeing Is Not Enough. Dust can be visible on a house while people disagree about whether it constitutes legally cognizable harm. Noise can be heard, light can cross a property line, water can be withdrawn from a shared system, and emissions can enter an atmosphere that belongs to no parcel at all. Seeing something and knowing what to do with it are different problems.
The water story adds another problem: what happens when information exists inside the process but doesn’t become part of the public object?
NorthMark says Spartanburg Water had the higher water projection on February 11. By May, the utility was asking customers across the Spartanburg system to voluntarily conserve water during severe drought. By August, its representatives were still publicly discussing Valara through the 459,000-gallon figure. Then, in September, NorthMark corrected its public statements and said it was revalidating other numbers.
That sequence doesn’t tell us why the information moved the way it did, and I’m not interested in assigning an intention the public record doesn’t establish. But it does tell us that “who’s in the room when the ask gets made,” the question I raised in What We Ask of the Water, is partly a question about information. Who knows what? When do they know it? Which version makes it into the meeting, the permit discussion, the newspaper story, the sermon?
Ecological attention has to be capable of revision at that level too. Replacing 459,000 with 1.2 million in a spreadsheet isn’t enough if we never ask how the smaller figure became the durable public description in the first place.
A parcel map answers one question and a watershed another. An air permit answers one question, and a person sitting on a porch across the road may be asking another. A water-demand questionnaire, a wastewater survey, a drought plan, and a public presentation each disclose something different. The work is not to choose one as the “real” account. It is to know what kind of account we’re looking at, what it can tell us, and where its edges are.
The closer we look at infrastructure, the harder it becomes to say where the thing itself begins and ends.
A County Tries to Draw a Line
Spartanburg County Council made its own attempt to sort among kinds of data centers Monday night, voting 5–1 for a one-year moratorium on new development. The ordinance doesn’t stop Valara, which is already permitted, and it includes a pathway for smaller facilities meeting specified conditions, including a ceiling of 65 megawatts and no routine on-site generation beyond emergency backup.
That distinction is easy to overlook, but it matters here. Valara is built around extensive on-site natural-gas generation. The county’s exception is drawing a line that would exclude precisely that model from the smaller class it is willing to consider during the moratorium.
The shape of that exception also emerged after months of public argument, including testimony from Lockhart Power and its parent company, Pacolet Milliken, that smaller grid-supplied facilities should not be treated as equivalent to hyperscale projects with extensive on-site generation. Residents and environmental advocates were making different arguments about water, air, noise, transparency, land use, and cumulative effects. Whatever one thinks the final ordinance should include, the 65-megawatt line did not fall from the sky. It came out of a public struggle over which differences ought to count.
That is one reason the moratorium interests me. The county hasn’t simply declared a data center either good or bad. It is trying to decide what kind of data center belongs where, at what scale, under what conditions, and in what relationship to the place around it. That is a harder question, and an ecological one.
Ecology rarely lets us evaluate a thing in isolation. It asks what feeds it, what it feeds, what it displaces, what passes through it, what becomes dependent upon it, and what remains after it arrives. Those questions become especially important with infrastructure because infrastructure is so good at disappearing from ordinary attention.
We notice the faucet, not the reservoir and treatment plant; the light switch, not the generating station and transmission corridor; the webpage, not the server rack, gas turbine, cooling system, electrical substation, fiber route, watershed, and tax agreement that allow it to appear instantly on a screen.
A data center may be one of the clearest examples of this strange modern condition. Its product feels almost immaterial. Its metabolism is anything but. That is the same infrastructure problem I’ve been following since Project Spero.
The More Clearly Something Appears
Every morning for the last few years here in Spartanburg, I’ve sat for a while beneath a black walnut tree in my backyard. One thing the practice has made increasingly difficult is the assumption that greater attention makes an object simpler. The opposite keeps happening.
At first there is a tree. Then there are leaves at different stages of yellowing, branches moving at different speeds in the wind, insects, birds, shade moving across the ground, soil, heat, rain from the night before, fallen limbs after a storm, and other plants growing differently because the walnut is there.
But even that description contains different kinds of knowing. Some things are directly visible. Some belong to memory. Some are anticipated. Some are inferred. Some I know because biology has taught me how to recognize them. Others remain inaccessible no matter how long I sit there. The tree becomes more particular as those relations appear, and at the same time harder to imagine as an isolated thing.
I think infrastructure deserves that same quality of attention. Water use isn’t air pollution, noise isn’t tax revenue, and a neighbor’s experience isn’t an engineering model. An engineering model isn’t invalid because it cannot contain a neighbor’s experience either. The discipline is to keep those differences intact without pretending they have nothing to do with one another.
That is harder than choosing a side, but it is closer to what living in a place actually requires.
What the Numbers Can’t Hold
We should want the numbers to be right. If full-buildout water demand is projected at 1.2 million gallons per day, then say so. If the maximum could reach 1.6 million, that matters too. Spartanburg Water’s assurance that the system has capacity to serve the project belongs in the public account, just as the drought conditions under which the utility asked the rest of us to conserve water belong there. The same is true of air: pollution controls, enforceable permit conditions, modeled concentrations, and the state’s conclusion that the facility can operate within applicable standards are all relevant facts.
Ecological attention can’t mean collecting only the facts that confirm what we already wanted to believe. It also can’t mean allowing a technically correct number to stand in for a place.
That may be what this strange year of data centers is teaching me. Greater precision doesn’t necessarily make the story more reassuring or more alarming. It makes more relationships visible. Water leads toward treatment capacity, wastewater, rainfall, rivers, drought planning, future customers, and growth. On-site generation leads toward natural gas, combustion, permits, monitoring, and the air moving across neighborhoods. A regulatory threshold leads toward the legal categories built around it. A 5–1 vote points to the people, testimony, interests, and arguments that drew the line.
The question, then, is larger than whether 459,000 was the wrong number and 1.2 or 1.6 million the right one. I want to know what kind of number I am looking at, what it allows me to see, and what appears when I look again.
In When Harm Has to Mean It, I wrote that responsibility begins by allowing more of the world to appear. In When Seeing Is Not Enough, I added that seeing is not enough if what appears is never allowed to count. This week has added another discipline to that sequence:
We have to remain willing to see again.
That includes me. I carried 459,000 gallons into a pulpit because it was the public number available to me. I would rather acknowledge that the picture has changed than defend the picture because I once spoke from it.
Ecological attention cannot simply produce a clearer version of the world we already thought was there. Sometimes greater attention changes the object we thought we knew. Sometimes it reveals that a projection was heard as a measurement, that a threshold was heard as a description of safety, or that one institution held information that had not yet become part of the public picture.
Here in Spartanburg, what remains outside any single number is not abstract. It is water and air, electricity and land, neighborhoods and industry, tax revenue and public trust, people already living here and people who will live here later.
The numbers matter, and we should get them right. But none of them is the place.
The place is what has to hold all of it.
Sources
NorthMark Compute and Cloud. “Water Needs Update.” September 21, 2026. NorthMark’s statement revising projected full-buildout water demand to an average of 1.2 million gallons per day, with a projected maximum of 1.6 million gallons per day, and stating that the higher figures were provided to Spartanburg Water on February 11, 2026.
Alex Land, WYFF 4. “Water demand for Spartanburg data center more than doubles original estimate.” September 21, 2026. Reporting on NorthMark’s revised water-use projections, the February 11 Large Water Meter Questionnaire, and Spartanburg Water’s assessment that the system has capacity to serve the project.
Spartanburg Water. “Spartanburg Water Customers Asked to Follow Voluntary Water Restrictions.” May 21, 2026. Spartanburg Water’s announcement of voluntary conservation measures after Spartanburg County entered severe drought conditions.
South Carolina Department of Environmental Services. “Valara Holdings High Performance Compute Center.” Updated September 18, 2026. SCDES project page containing the permit decision, public-hearing materials, air-modeling information, permit documents, and responses to public comments for the Valara facility.
South Carolina Department of Environmental Services. “Air Quality Department Decisions.” September 18, 2026. State records for the Valara permit, including approval of five 17 MW and six 54 MW natural-gas-fired turbines with pollution-control systems.
Alex Land, WYFF 4. “NorthMark attorney explains air pollution controls and enforcement for Spartanburg data center.” September 21, 2026. Reporting on Valara’s synthetic-minor classification, formaldehyde estimates, pollution controls, air-dispersion modeling, monitoring, and enforcement.
Isabel Martin and Ripley Simone, WSPA 7NEWS. “Spartanburg County Council passes data center moratorium.” September 21, 2026. Reporting on County Council’s 5–1 vote for a one-year moratorium on new data-center development, the exceptions adopted by council, and the treatment of already-permitted projects, including Valara.