The same loop that moves a data center’s waste heat has to answer for the water, too — and the honest answer isn’t an adjective, it’s a meter. One water system, four climates, and a single published number. This is the case, in the open.
A co-located greenhouse is the one benefit that, framed carelessly, looks like it adds to a data center’s water burden. So this is the page where we are most exact about what we do and don’t claim.
A data center touches water twice: the water it evaporates on site, and the much larger volume used at the power plant that feeds it. We change neither — though one optional configuration, the Campus Catchment below, changes where the first water comes from, never how much of it there is. The industry’s own shift to liquid cooling is what moves that needle — a Microsoft-led life-cycle assessment found it cuts blue-water use 31–52% against air cooling — and that’s the operator’s win, not ours.
What the greenhouse can own is narrower and true: built to sip the town’s water, recycle most of what it draws, and report the net on a meter. Not “water-positive” as a pledge — that word carries an accounting baggage that’s the opposite of auditable. A number on a gauge, instead.
The honest answer on water isn’t a promise. It’s a meter — the water twin of the BTU meter we already publish on the heat loop.
The short version: the greenhouse recycles almost all of its water, catches rain, and even pulls clean water out of the air using the data center’s cold side. Whatever small amount of town water it still needs shows up on a meter anyone can check — so the water claim is a number you can verify, not a promise. The rest of this section is the detail behind that.
Four levers ship on every Intelligent Harvest greenhouse, in every climate. They’re what make this a system rather than a string of one-offs — and why the same sentence about water holds whether the site is in Lynchburg or Lisbon.
The recovery lever is the one that matters most. A sealed recirculating greenhouse that recaptures its transpired water has been estimated to use 70–90% less water than open-field farming at the same yield — the town gets local food without importing agriculture’s water footprint.
Recirculating CEA vs open-field, at equal yield (Dr. Greenhouse; industry CEA reporting). Illustrative range.
A third-party net-withdrawal meter on the make-up line records how much town or aquifer water the greenhouse actually draws, net of what it captures and recycles — and that number is published.
This is what separates a benefit from a billboard. The crops flex by climate. The cooling flexes. The water strategy flexes. The meter does not. A project whose entire pitch is “a benefit you can verify” has to be the first to submit the one thing a skeptic can read off a gauge — so we do, on the public record, rather than asking anyone to take a brochure on faith.
A claim you can audit is not greenwashing. A claim you can’t is.
Onto that shared spine each site bolts a climate configuration — changing only what it must. We keep two confidence levels visibly separate, so the desert’s ambition never bleeds onto Virginia’s certainty.
A 2.5-acre (108,900 ft²) flagship roof at the Mid-Atlantic NOAA normal of 46 inches, at a design collection coefficient of 0.85, returns on the order of 2.65 million gallons a year — about 2.5 million at the conservative 0.80 coefficient, and about 1.9 million in a 72%-of-normal dry year† — before any campus roof is added; the same roof in an arid 350 mm/yr climate still returns roughly 0.79 million. Design estimates from stated assumptions — the engineer of record sets the real numbers.
In the temperate config the greenhouse still draws a little town water to make up — on the order of a few percent of what it uses. The catchment above is sized to return at least that much: the small gross make-up is offset by the rain the site captures, so the net withdrawal — not the gross — targets zero. That net is the only number the third-party meter reports, and it’s the one we publish. Net-zero is the committed design target; the gauge, not the adjective, is what holds us to it.
It’s the one configuration we tier as a design goal rather than a commitment — and the reason is physical honesty.
Two things change. You cool without evaporating, because evaporative cooling is useless in the very heat where you’d reach for it — so the loop, dry coolers, and waste-heat-regenerated desiccant do the work, and the desiccant captures the humidity it pulls from the air. And you use the free heat to make supply: Sundrop Farms grows on twenty hectares of South Australian desert, running seawater through heat-driven distillation to make up to a million liters of fresh water a day. Swap concentrated solar for a data center’s reject heat and the logic is identical.
A data center’s reject heat is low-grade — roughly 35–50 °C — below the comfortable range for the distillation Sundrop uses. The matched processes are humidification–dehumidification and membrane distillation, and proving them here — with responsible brine handling — is the engineer’s job. So we name the floor and the aim, and refuse to print a figure before the math is stamped.
The threat isn’t scarcity — it’s surplus, intrusion, and sinking ground. So the coastal config’s headline is the exact opposite of the desert’s.
A data center is acres of impervious surface — on this coast, a flooding liability. Paired with a designed catchment and retention pond, that liability becomes managed capacity, capturing intense rainfall instead of shedding it — in step with Virginia Beach’s own Sea Level Wise strategy. And because the greenhouse runs on rain and condensate, it never pumps the stressed aquifer, which is the region’s number-one water priority.
Running on rain and condensate and never drawing the local aquifer, the coastal config is aquifer-neutral — that one we can stand behind today. The aim is one step further: route the stormwater the campus captures toward managed aquifer recharge — the same intervention utilities across the country already use to push back saltwater intrusion and slow subsidence. That would make the campus aquifer-positive — putting more clean water back than it ever takes. We hold that as the design aim, under validation, because the recharge and water-quality path is the engineer of record’s to confirm. The neutral claim is committed; the positive one we’ll earn on the meter.
This does not stop sea-level rise or reverse subsidence — recharging an aquifer is a different intervention entirely. Our claim is narrow and strong: doesn’t draw the aquifer, absorbs stormwater, recycles its own water, metered. Seawater desalination is deliberately left out — brine into the Chesapeake watershed is a non-starter.
Three of the four configs ship a committed claim — proven gear in a proven climate, defensible at a podium today. One, the desert, ships a design target under validation. Keeping those two levels visibly separate is the whole safeguard.
That’s what makes the architecture global: one spine, four configs, one meter — each carrying only the claim its evidence supports. The crops change, the cooling changes, the water strategy changes by climate. The honesty, and the gauge that enforces it, do not.
One spine. Four climates. One meter.
Everything above stands on its own: a greenhouse that keeps its water, sips the town’s, and reports the net on a meter. That is the floor, and it is the whole deal wherever a town or an operator prefers it — self-sustaining, metered, done. The Campus Catchment is the configuration on top of the floor, offered where the operator signs one document.
The document is a roof tie-in: a drainage easement and runoff license, an exhibit to the ground lease. It lets us meet the campus’s rain at the curb — at grade, downstream of the drainage code already requires, through one precast vault with a first-flush diverter, a screen, and a meter. The clean flow banks in a covered wet cell. From there, a one-way, air-gapped, backflow-prevented line supplies three things back: the data center’s commissioning fill (≈400,000 gallons per building — the operator’s own public figure), its hottest-100-hours misting water, and a dedicated fire reserve no drought is allowed to borrow. Heat crosses the fence one way. Water crosses back the other. Both meters are published.
Greenhouse roofs alone cannot fund these promises — the canonical math shows the worst dry-year corner goes negative at campus scale without the data-center roofs, and clears only marginally even at the smallest build. So the offer travels with the easement, at every size. The design also ships two cells — a covered wet cell that wants to be full, and a dry detention cell that must sit empty — because one pond cannot serve both masters; and anything that will ever be misted passes filtration, UV, and residual treatment first.
Our system can only fail into their system working exactly as designed. The diversion weir is passive: if our valve jams, our pond fills, or Intelligent Harvest disappears, water continues over the weir to the operator’s detention and outfall — the path their stamped permit drawings already show. Nothing of ours sits in the critical path of their cooling, their drainage, or their compliance; and the value-of-water disclosure is made on purpose: one fill is worth about $2,800 at municipal rates†. The case was never the water bill — it is time, trust, and a town that can read the gauge.
The floor stands alone. The Catchment is offered, never required.
Sources & notes. CEA water efficiency — Dr. Greenhouse; industry CEA reporting. Liquid-cooling blue water — Microsoft-led life-cycle assessment, Nature, 2025. Seawater greenhouse precedent — Sundrop Farms, South Australia (independent; not affiliated). Coastal subsidence — Ohenhen et al., Nature (2024); Nature Cities (2025); NASA Earth Observatory. Managed aquifer recharge — USGS; EPA. Campus Catchment supply figures — internal model on stated assumptions (Canonical Water Math v1.2); first-fill and misting practice — public operator reporting (TeraWulf, Boyd/Hancock Co. KY, 2026). Greenhouse water, yield, and catchment figures are internal design targets or models with stated assumptions, not measured results. Environmental claims are scoped to reusing heat that is otherwise rejected, to metered net withdrawal, and — where the Campus Catchment is configured — to metered supply; nothing here claims the greenhouse reduces a data center’s own power or water use, halts sea-level rise, or reverses subsidence.