Why Direct Wastewater Cannot Cool a Data Center
Meta treats wastewater for its data centers through an MBR-RO-UV train that converts municipal sewage into industrial-grade reclaimed water for evaporative cooling. In 2026, Meta committed at least $270 million to wastewater treatment infrastructure around its data centers, and Loudoun Water data show recycled water supplies about 43% of daily consumption across 250+ Loudoun County data centers — the remaining 57% still comes from potable sources (gate.com, 2026).
Untreated sewage cannot be piped straight into a cooling tower because it carries four contaminant classes, each of which attacks a different piece of equipment. Dissolved salts and hardness ions (calcium, magnesium, silica) deposit as scale on evaporative cooling fills and heat-exchanger tubes, reducing heat-transfer coefficients and forcing unscheduled shutdowns. Ammonia and urea break down into nitrifying bacteria that colonize tower fill and promote biofouling, which drops cooling capacity and elevates Legionella risk. Suspended organics and emulsified oils foul RO pre-filters and coat membrane surfaces, raising differential pressure across the train. Residual chloride and hydrogen sulfide attack copper and carbon-steel metallurgy in the cooling distribution loop, accelerating corrosion rates by 2–4× relative to potable makeup (gate.com, 2026).
Evaporative cooling makes the problem worse. The tower cycles water 4–8× through latent-heat transfer before discharging it as blowdown, so every non-volatile dissolved solid concentrates in the recirculating loop. A 200 mg/L TDS makeup becomes 800–1,600 mg/L at the point of blowdown, which is why the industry treats cooling-tower blowdown as a separate wastewater stream from the cooling makeup.
Two output streams have to be designed for: (1) the makeup water that feeds the cooling tower, and (2) the cooling-tower blowdown that must be discharged, reused, or evaporated. Tertiary treatment removes ~99% of impurities from wastewater, which is the engineering target any data-center makeup stream must clear (Water Air & Soil Pollution, 2021). Meeting that target is exactly what the MBR-RO-UV train is built to do.
The MBR-RO-UV Train: How Meta Converts Sewage to Cooling Makeup
Three unit operations in series — Membrane Bioreactor, Reverse Osmosis, and Ultraviolet disinfection — produce the reclaimed water that Loudoun Water and other reclaimed-water suppliers deliver to Meta's evaporative cooling loops. The train matches the process flow that gate.com attributes to Loudoun Water suppliers, and the same unit operations are what any project bidder would specify for a comparable reclaimed-water cooling plant.
Step 1 — Membrane Bioreactor (MBR). Submerged PVDF hollow-fiber membranes with a nominal pore size of ~0.1 μm replace the secondary clarifier in a conventional activated-sludge plant. Mixed-liquor suspended solids (MLSS) run at 8,000–12,000 mg/L, an order of magnitude higher than a clarifier-based system, which collapses the aeration tank footprint and drives BOD/COD removal above 95%. MBR effluent typically lands at <1 NTU turbidity and <5 mg/L TSS, a quality stable enough to feed RO membranes directly without media filtration (gate.com, 2026). A properly sized submerged MBR system for the first barrier of the reclaimed-water train is the unit operation that makes the rest of the train reliable.
Step 2 — Reverse Osmosis (RO). High-pressure pumps (10–30 bar) force MBR permeate through polyamide thin-film composite membranes that block monovalent and divalent ions, silica, and low-molecular-weight organics. Modern brackish-water RO runs at 80–95% recovery, with 85% the typical design point for data-center cooling makeup. At 95% recovery, an industrial RO unit delivering up to 95% recovery produces only ~5% concentrate by volume, but that concentrate carries the rejected salts and is the feed stream for the blowdown treatment covered later in this article. Permeate TDS lands at <50 mg/L with conductivity below 100 µS/cm, which is why RO is the only barrier that can drop hardness and silica below cooling-tower scaling thresholds.
Step 3 — Ultraviolet (UV) Disinfection. A final chemical-free UV disinfection for reclaimed cooling water at a typical dose of 40 mJ/cm² inactivates chlorine-resistant Cryptosporidium and Giardia to meet sanitation standards. With that polishing dose, the same train can reach the potable-grade ceiling that gate.com notes is achievable "with specific treatment processes." A conventional wastewater-reuse train — sedimentation, sand filter, chlorination — fails on data-center duty because residual hardness and silica still scale tower fills, while chlorination by-products (THMs, HAAs) attack copper heat-exchanger metallurgy.
| Unit Operation | Primary Function | Typical Output | Data-Center Role |
|---|---|---|---|
| MBR (PVDF, ~0.1 μm) | Biological oxidation + physical separation | <1 NTU, <5 mg/L TSS, BOD/COD removal >95% | RO feed protection |
| RO (TFC polyamide) | Dissolved solids removal | 80–95% recovery, permeate TDS <50 mg/L | Scale and silica barrier |
| UV (low-pressure Hg) | Pathogen inactivation | 40 mJ/cm² dose, >99.9% inactivation of Cryptosporidium | Chemical-free final polish |
| Conventional train (sed/sand/Cl₂) | Basic reuse | Residual hardness, chlorinated organics | Fails on tower duty |
Meta's Site Footprint and the $270 Million Infrastructure Pledge
Meta's reclaimed-water strategy is best read site by site, because each U.S. data-center hub has a different supplier and a different reuse mix. Loudoun County, Virginia is the densest data-center cluster globally, with more than 250 facilities drawing on Loudoun Water for cooling makeup; recycled water supplies 43% of that daily consumption while the remaining 57% still comes from municipal potable sources (gate.com, 2026). At least 24 additional data centers are planned in the county, so the 43% share will not rise on its own — additional reclamation capacity has to be financed and built first.
That financing is what Meta's 2026 commitment of at least $270 million in wastewater treatment infrastructure is designed to unlock. The structure is the anchor-tenant model that Michael Obradovitch describes: hyperscaler funding lets a utility expand its reclamation plant once, then serve many data centers, rather than each operator building a private plant. The funds flow to off-site utilities such as Loudoun Water, not to on-site Meta treatment plants, which is why this is a water-utility capital story as much as a data-center one.
Cheyenne, Wyoming is the cautionary case. The Cheyenne Board of Public Utilities (BOPU) traced a rare-bacteria discharge in 2025 to a contractor working on a Meta data-center site, an event that landed in municipal drinking-water coverage (wyomingnews.com, 2025). The engineering lesson is that infrastructure funding does not by itself guarantee operational safety: discharge permits, monitoring wells, and contractor oversight must travel with every dollar that flows into a new reclamation train. The incident is a useful internal benchmark for any operator weighing the difference between capital committed and effluent actually kept in compliance.
Cooling Blowdown: The Second Wastewater Stream Most Plans Overlook
Designing the makeup train is only half of a reclaimed-water cooling project. A cooling-tower loop concentrates dissolved solids 3–5× relative to the makeup water, and the resulting blowdown carries corrosion inhibitors, biocides, and Legionella-control chemicals that make it the harder stream to permit. At 95% RO recovery upstream, a 10,000 m³/day cooling plant still produces roughly 500 m³/day of brine concentrate that needs a discharge path or a near-ZLD finish before it can be released.
Sidestream options that hyperscalers and their utility partners are evaluating in 2026: a second-pass RO at 70–80% recovery to cut brine volume by another 50–70%; a brine concentrator or mechanical vapor compression (MVC) evaporator for sites targeting zero liquid discharge; and routing the concentrate to lower-quality reuse applications catalogued in the wastewater-reuse literature, including agriculture, golf-course irrigation, vehicle washing, firefighting, toilet flush, and building construction (Water Air & Soil Pollution, 2021). Pretreatment matters here too — a DAF pretreatment ahead of the MBR is the standard way to strip fats, oils, and floating solids before they reach the biological stage.
Direct discharge permits around Loudoun, Phoenix, and the Dallas–Fort Worth corridor have tightened steadily through 2025, which is why hyperscaler interest in ZLD is rising even though full ZLD still imposes a heavy energy penalty. The practical target for most 2026 projects is high-recovery RO plus an industrial reuse outlet, with a brine concentrator held as a future option rather than a day-one spec.
Decision Framework: Specifying a Reclaimed-Water Cooling Train in 2026
The following rules translate the prior sections into a spec checklist a B2B engineer can lift into a project.
Decision rule 1 — End use dictates the train. If the goal is industrial cooling makeup only, MBR-RO-UV at 80–90% RO recovery is sufficient and matches Meta's reference train. If the goal is direct drinking reuse, add a second-pass RO plus a remineralization stage. If the goal is ZLD, add an evaporator/crystallizer on the concentrate.
Decision rule 2 — Confirm the upstream hydraulic capacity first. Dr. Greta Zornes of CDM Smith flagged that rural data-center sites often have no wastewater plant large enough nearby, and that the real bottleneck is construction time, not capital (gate.com, 2026). A site with 5 MGD of demand but a 2 MGD upstream plant is a 36–48 month build, not a 12-month one.
Decision rule 3 — Pretreatment sequencing. Bar screen → grit removal or DAF → equalization → MBR. A rotary mechanical bar screen at the head of the train protects downstream membranes from rags and grit that municipal plants do not always polish out.
Decision rule 4 — Disinfection choice. UV for chemical-free compliance on the makeup line, chlorine dioxide for residual protection in long cooling distribution loops where biofilm control matters most, ozone for odor and color control on stored reclaimed water. Each chemistry has a different by-product profile and a different compatibility story with copper heat exchangers.
Decision rule 5 — Sludge handling. Specify a plate-and-frame filter press for MBR and DAF sludge so the concentrate stream does not become a landfill liability. MBR waste-activated sludge and DAF float combined typically run 2–4% dry solids; a filter press lifts that to 22–28% cake, which is the disposal-ready range.
| Decision Rule | If Condition | Specify |
|---|---|---|
| End use | Cooling makeup only | MBR-RO-UV, 80–90% RO recovery |
| End use | Direct drinking reuse | Add second-pass RO + remineralization |
| End use | Near-ZLD target | Add brine concentrator or MVC evaporator |
| Site hydraulics | Rural, small upstream WWTP | Phase project 36–48 months |
| Pretreatment | High FOG or grit in influent | Bar screen → DAF → equalization → MBR |
| Disinfection | Long distribution loop | UV + ClO₂ residual |
| Sludge | 2–4% dry solids WAS + float | Plate-and-frame filter press to 22–28% cake |
For Latin American projects sizing similar trains, the Rosario data center cooling blowdown treatment guide and the Córdoba data center blowdown treatment guide walk through blowdown-specific configurations. Breweries run analogous reclaim trains at smaller scale; the Heineken brewery wastewater treatment process uses an MBR-RO sequence on a similar feed-water profile.
The Hyperscaler-Water-Utility Model: Why Meta Funds Off-Site Treatment
Meta's $270 million flows to local utilities, not to private on-site plants, because the anchor-tenant model concentrates capital where it can serve many customers. One expanded reclamation plant, sized to Loudoun Water's full demand curve, replaces the dozen private plants each data center would otherwise build — and it consolidates discharge permitting under a single utility that already holds the NPDES permit. Federal policy is moving in the same direction: legislation sponsored by Rep. Pigott would provide a 30% tax credit to accelerate corporate investment in expanding recycled water infrastructure, which would shorten the payback on the anchor-tenant model and pull more utilities into the same construction cycle (gate.com, 2026).
The Cheyenne BOPU incident is the risk-allocation caveat. Even with hyperscaler funding, the utility remains the discharge permit holder and is on the hook for bacteria, Legionella, and any permit excursion caused by a data-center contractor's discharge. Contracts have to specify monitoring, indemnity, and operator-of-record responsibilities, not just a capital commitment. As AI compute grows through 2026 and 2027, the same anchor-tenant model will migrate to Phoenix, Columbus, and the Dallas–Fort Worth corridor, and the MBR-RO-UV train will become a standard reference design across new U.S. data-center hubs.
Frequently Asked Questions
What treatment train does Meta use for data-center cooling water?
Meta's reclaimed-water suppliers run an MBR-RO-UV train: a submerged MBR (~0.1 μm PVDF) for biological and solids removal, RO for salt and silica rejection at 80–95% recovery, and UV at ~40 mJ/cm² for chemical-free pathogen inactivation (gate.com, 2026).
How much has Meta committed to wastewater treatment infrastructure?
Meta announced at least $270 million in funding for wastewater treatment infrastructure around its data centers in 2026, structured as anchor-tenant commitments to local utilities such as Loudoun Water rather than on-site treatment plants (gate.com, 2026).
Why does Loudoun County still use 57% potable water for data centers?
Loudoun Water data show recycled water supplies 43% of daily consumption across 250+ data centers, with the remaining 57% still drawn from municipal potable sources because reclamation capacity has not kept pace with demand; at least 24 additional facilities are planned (gate.com, 2026).
Can wastewater be reused directly for evaporative cooling without treatment?
No. Untreated wastewater contains salts, urea, bacteria, and organic compounds that scale cooling fills, drive biofouling, and corrode heat-exchanger metallurgy, which is why tertiary treatment targeting ~99% impurity removal is the engineering baseline (gate.com, 2026; Water Air & Soil Pollution, 2021).
What happened with the Cheyenne BOPU discharge incident?
Cheyenne BOPU traced a rare-bacteria discharge in 2025 to a contractor working on a Meta data-center site, demonstrating that infrastructure funding alone does not guarantee permit compliance and that monitoring must travel with capital (wyomingnews.com, 2025).
