Why Microsoft Data Center Wastewater Is Different From a Typical Industrial Discharge
Data center cooling blowdown is fundamentally a mineral-concentration problem, not an organic-load problem. After 2–5 cycles through an evaporative cooling tower, the bleed stream carries elevated calcium, magnesium, and silica pulled from the makeup water; the TDS climbs with each cycle until the heat-transfer surfaces start scaling (per the Microsoft local blog, confirmed in the EPA Quincy case study). At Quincy, Washington, that makeup historically came from the Columbia Plateau basalt aquifer up to 1,000 ft deep — water already high in dissolved solids before any cooling cycle (per EPA case study, 2021). Discharging that stream back to a conventional POTW like Quincy's MWRF pushed influent TDS well above the Washington Department of Ecology's 500 mg/L groundwater recharge guideline, forced extra fees on the data centers, and produced a visible mineral cake on the recharge basins in summer (per EPA case study).
For most of the 20th century, the default answer was "send it to the municipal plant." That answer breaks in two places in 2026: when the receiving POTW's discharge permit caps TDS below what the blowdown carries, and when the source aquifer is already over-drawn. Hyperscalers now bypass both failure modes by co-financing purpose-built industrial reuse plants. The canonical example is the Quincy Water Reuse Utility (QWRU), where Microsoft financed $31 million of capital under a 30-year operating agreement with the City of Quincy (per EPA case study, 2021). The contrast campus — Mount Pleasant, Wisconsin — sidesteps the same physics with zero-water closed-loop liquid-to-chip cooling, which generates no blowdown to treat (per Wisconsin Public Radio, 2024).
Inside the Quincy Water Reuse Utility: Microsoft's Closed-Loop Treatment Train
The QWRU is a 10-system treatment train connected by more than 30 miles of pipe across Quincy's industrial area, operational since June 30, 2021 (per EPA case study). Microsoft financed the $31 million capital cost and pays a monthly operations fee; the City of Quincy owns and operates the plant. The blowdown stream follows a defined sequence: cooling-tower bleed → lime softening (calcium hydroxide for Ca/Mg hardness) → ultrafiltration → split between a blend tank and a hot-process softening (HES) unit → reverse osmosis → permeate buffer tank (~8 hours of storage) → return to the cooling tower (per EPA case study, 2021). Conductivity is blended online to a 300–350 µS/cm target, with the HES/RO and UF streams mixed at varying ratios through the day to hold that band.
Two brine streams leave the train. HES brine carries the rejected calcium and magnesium; RO brine carries the rejected silica. Both pipe to lined brine ponds for maximum evaporation, with solid residuals disposed of separately on a 2–3 year cycle (per EPA case study). The system is topped up with 260 million gal/yr of Columbia Basin Project irrigation-canal water — about 95% of makeup — plus 5% softened potable groundwater. The only discharge is to the brine ponds; there is no surface- or groundwater discharge, which is why the QWRU qualifies as a closed-loop industrial reuse plant rather than a zero-liquid-discharge system in the strict sense.
| Parameter | Value | Source |
|---|---|---|
| Cooling cycles before blowdown | 2–5 cycles | Microsoft local blog; EPA case study |
| Conductivity target (blended permeate) | 300–350 µS/cm | EPA case study (2021) |
| Permeate buffer storage | ~8 hours | EPA case study (2021) |
| Distinct treatment systems | 10 | EPA case study (2021) |
| Length of interconnecting pipe | 30+ miles (48 km) | EPA case study (2021) |
| Makeup from Columbia Basin Project canal | 260 million gal/yr (95%) | EPA case study (2021) |
| Makeup from softened potable groundwater | 5% of total | EPA case study (2021) |
| Potable water demand offset | 138 million gal/yr (522 million L/yr) | EPA case study (2021) |
| Brine/solid residual disposal interval | Every 2–3 years | EPA case study (2021) |
| Capital cost (Microsoft-financed) | $31 million | EPA case study (2021) |
| Operating agreement term | 30 years | EPA case study (2021) |
Mount Pleasant, Wisconsin: The 2026 Zero-Water Cooling Campus

Microsoft's Mount Pleasant campus is the design counterpoint to Quincy. Rather than treat blowdown after it is generated, the campus is engineered so most of the cooling load never produces blowdown. Most of the site runs on zero-water closed-loop liquid-to-chip cooling; the campus reserves a peak of 350,000 gpd of municipal water for the hottest May–September days, with non-cooling loads dominating the rest of the year (per WPR, 2024). The published water balance for Phase 1 is 234,000 gpd intake against 81,000 gpd discharge (~2 million gal/yr), scaling at full build-out to 702,000 gpd intake and 243,000 gpd discharge (~6 million gal/yr) (per WPR, 2024). All intake water comes from Racine's existing Lake Michigan allocation of up to 7 million gpd, approved by the Wisconsin Department of Natural Resources in 2018 under the Great Lakes Compact for the straddling-community diversion. Phase 1 construction wraps in early 2026, and the $3.3 billion investment is layered on land originally optioned for the Foxconn LCD plant.
The implication for a 2026 designer is that hyperscaler-grade cooling comes in two non-overlapping archetypes. In hot, dry, mineral-rich source-water regions, the Quincy model — treat all blowdown in a purpose-built reuse loop — is the reference design. In temperate regions where closed-loop liquid-to-chip is technically and economically viable, the Mount Pleasant model — avoid blowdown at the source — is the blueprint. These are not interchangeable; the equipment list, brine-handling obligations, and regulatory profile differ sharply.
| Metric | Phase 1 | Full build-out | Source |
|---|---|---|---|
| Daily water intake | 234,000 gpd | 702,000 gpd | WPR (2024) |
| Daily wastewater discharge | 81,000 gpd | 243,000 gpd | WPR (2024) |
| Annual water intake | ~2.8 million gal/yr | ~8.4 million gal/yr | WPR (2024) |
| Annual discharge | ~2 million gal/yr | ~6 million gal/yr | WPR (2024) |
| Closed-loop cooling water use | 0 gpd (most of campus) | 0 gpd (most of campus) | Microsoft spokesperson via WPR (2024) |
| Peak non-cooling reserve (May–Sep) | 350,000 gpd | 350,000 gpd | WPR (2024) |
| Lake Michigan diversion cap | 7,000,000 gpd | 7,000,000 gpd | WI DNR 2018 allocation (per WPR, 2024) |
| Phase 1 construction completion | Early 2026 | — | WPR (2024) |
Reclaimed Water, Rainwater and the Reuse Strategy Around the World
Microsoft treats wastewater handling as a portfolio, not a single plant. Reclaimed and recycled water is in use at sites in Texas, Washington, California, and Singapore, with expansion flagged in 2023 onward (per Microsoft local blog). Quincy sits inside that portfolio: the QWRU cuts potable water use at the campus by 97% and returns roughly 1.5 million m³/yr to community drinking supply through the indirect potable reuse path (per Microsoft local blog; cross-checked against the EPA case study demand-offset figure of 138 million gal/yr).
On the alternative-source side, rainwater capture is operational in Middenmeer (Netherlands) and Ireland, with announced expansion into Canada, the United Kingdom, Finland, Italy, South Africa, India, and Austria (per Microsoft local blog). In practice, campuses that still rely on municipal water do not add chemicals for cooling unless source-water hardness or bacterial control demands it; closed-loop heat-transfer loops that use propylene glycol are emptied only rarely, and the spent glycol is collected and hauled off for regulated disposal rather than sewered (per Microsoft local blog). For a 2026 buyer comparing regions, the takeaway is that hyperscaler water sourcing is matched to climate, source-water TDS, and regulatory ceiling — not picked from a global template.
Engineering Trade-offs: When to Reuse, When to Avoid, When to Discharge

The Quincy and Mount Pleasant campuses encode two non-overlapping design philosophies, and the choice between them is set by local physics and regulation rather than corporate preference. The reuse model wins when source-water TDS is already high, when receiving-water TDS limits are tight (the Washington Department of Ecology's 500 mg/L groundwater guideline is the textbook threshold), and when a municipal co-finance partner can be assembled for a 30-year operating agreement. The avoidance model wins when air-side economization covers the bulk of the year and when AI / liquid-to-chip cooling is technically viable at the rack level, eliminating the blowdown stream entirely (per Microsoft local blog; WPR, 2024).
Conventional discharge is still the right answer at sites where the local POTW has TDS headroom and where hyperscaler water-positivity commitments are met through offset projects — rainwater harvesting, community replenishment, and stormwater capture — rather than in-plant treatment (per Microsoft local blog). Microsoft's own 2022–2030 targets give a 2026 benchmark for any spec: 18% water-intensity reduction already achieved, 40% improvement goal by 2030 across owned datacenters (per Microsoft local blog). The four archetypes below are the decision matrix a campus designer should walk before specifying equipment.
| Design model | Trigger conditions | Reference campus | Discharge profile |
|---|---|---|---|
| Closed-loop reuse (treat all blowdown) | Source water high in TDS; receiving-water TDS limit <500 mg/L; municipal co-finance feasible | Quincy, WA — QWRU | None to surface/groundwater; brine to lined ponds every 2–3 yr |
| Avoidance (zero-water closed-loop) | Temperate climate; liquid-to-chip cooling viable; >85% air-side economization hours | Mount Pleasant, WI (2026) | Minor non-cooling discharge; no cooling blowdown |
| Conventional discharge to POTW | POTW has TDS headroom; offset projects cover water-positivity commitment | Most U.S. legacy campuses | Blowdown to sewer after 2–5 cycles |
| Reclaimed/rainwater substitution | Alternative source available; reduces potable demand without new discharge permit | San Antonio, TX; Middenmeer, NL | Same as baseline; lower potable draw |
What a Hyperscaler-Grade Reuse Plant Needs in 2026 Equipment Terms
The QWRU train maps cleanly to a buyer-facing equipment list. Pretreatment starts with a multi-media filter ahead of any membrane, sized to drop turbidity low enough to keep the downstream UF from fouling on TSS excursions. Softening is a lime-dosing skid with hot-process softening followed by a PLC-controlled chemical dosing system to hold Ca/Mg removal on setpoint across flow variations. The UF stage ahead of RO is typically a 0.03 µm PVDF membrane, which protects the RO from particulates and microbiological fouling; the RO itself runs at 70–85% recovery to hit QWRU-style conductivity targets, and a membrane bioreactor is worth shortlisting where the upstream stream carries any organic load. The brine residuals — softening sludge and RO concentrate — need a plate and frame filter press for the clarifier underflow and a lined brine pond or mechanical evaporator for the RO concentrate; for polish, an industrial RO system sets the recovery/conductivity envelope and a UF system feeds it.
Any makeup stream entering a closed loop should pass through UV or chlorine dioxide for bacterial control, matching the chemistry stance in the Microsoft datacenter FAQ. Comparable reference designs for tropical or water-stressed sites in Latin America and Africa are covered in the Rosario data center blowdown guide, the Córdoba data center blowdown guide, and the Luanda data center blowdown guide; the QWRU parameter bands carry across those climates once source-water TDS and discharge-permit limits are matched.
| QWRU unit operation | 2026 equipment category | Design note |
|---|---|---|
| TSS / turbidity reduction upstream of softening | Lamella clarifier or DAF | Protects lime reactor from particulate overload |
| Multi-media polish | Multi-media filter | Pre-UF guard; sized to feed water <1 NTU |
| Hardness removal (Ca/Mg) | Lime dosing + HES skid | Ca(OH)₂ reactor followed by hot-process softener |
| Reagent control | PLC-controlled chemical dosing system | Trim on effluent Ca and pH |
| Particulate / microbial barrier | UF system, 0.03 µm PVDF | SDI reduction ahead of RO |
| Dissolved-solids polish | Industrial RO system, 70–85% recovery | Permeate to 300–350 µS/cm |
| Softening clarifier sludge | Plate and frame filter press | Cake to lined disposal; filtrate returned to headworks |
| RO concentrate | Lined brine pond or mechanical evaporator | Brine to pond; solids hauled every 2–3 yr |
| Makeup-loop disinfection | UV or chlorine dioxide | Matches Microsoft closed-loop chemistry stance |
Frequently Asked Questions
How does Microsoft treat wastewater at its data center campus plant?
At Quincy, Washington, blowdown is treated in the Quincy Water Reuse Utility through lime softening, ultrafiltration, hot-process softening, and reverse osmosis, then blended to 300–350 µS/cm and returned to the cooling tower (per EPA case study, 2021). The system is closed-loop: no industrial wastewater is discharged to the municipal plant, and brine is piped to lined evaporation ponds with solids disposed every 2–3 years.
What process units make up Microsoft's Quincy Water Reuse Utility?
Ten distinct treatment systems connected by 30+ miles of pipe: lime softening, ultrafiltration, a blend tank, hot-process softening, reverse osmosis, and an 8-hour permeate buffer tank, plus brine handling and chemical dosing skids (per EPA case study, 2021). The plant became operational on June 30, 2021, financed by a $31 million capital contribution from Microsoft under a 30-year operating agreement with the City of Quincy.
How much potable water does the QWRU save?
About 138 million gallons per year (522 million L/yr) of potable groundwater offset, equivalent to a 97% reduction in potable demand at the connected data center campus (per EPA case study, 2021; Microsoft local blog). Makeup water is 95% Columbia Basin Project irrigation-canal water at 260 million gal/yr, with 5% softened potable groundwater.
What is the 2026 design direction for new Microsoft campuses?
Zero-water closed-loop liquid-to-chip cooling, with most of the campus generating no blowdown at all and only a peak 350,000 gpd reserve for the hottest May–September days (per WPR, 2024; Microsoft local blog). Mount Pleasant, Wisconsin is the first campus built to this standard, with Phase 1 construction completing in early 2026 under Racine's existing 7 million gpd Lake Michigan allocation.