A 200 MW U.S. facility in this data center cooling case study cut cooling makeup water use by 40%, saving 138 million gallons per year. A hybrid zero liquid discharge (ZLD) and reverse osmosis (RO) train raised cycles of concentration (COC) from 3 to 6.5 while recovering 95% of blowdown water. Annual water and sewer fees fell by $1.7 million against a $3.2 million capital outlay, and the plant stayed inside its NPDES discharge limits.
Key findings from this data center cooling case study
A hybrid RO plus ZLD system on a 200 MW data center recovered 95% of cooling tower blowdown. Makeup use fell 40%, or 138 million gallons per year. COC rose from 3 to 6.5. Annual water and sewer savings reached $1.7 million at $3.2 million CAPEX, with about 2.67 years payback after $0.5 million yearly OPEX.
Why cooling water costs and permits squeeze large campuses
Data centers often use 3 to 5 million gallons of water per megawatt each year, based on 2023 EPA figures cited in the original project brief. Evaporative towers usually take 80–90% of that volume on large campuses. High purchase and discharge fees push owners to raise COC before they add new wells or city capacity.
NPDES permits often set strict blowdown quality limits on TDS, metals, and temperature. Local groundwater caps, such as those reported around Quincy, WA, further constrain makeup. Corporate targets like Microsoft’s 2030 water-positive pledge add internal pressure to cut potable demand.
Blowdown commonly carries TDS at 500–2,000 mg/L and chlorides at 100–500 mg/L. Calcium and magnesium drive carbonate scale when LSI climbs. Biological growth, including Legionella risk, forces continuous biocide control even when mineral control is strong.
This campus relied on mineral-rich groundwater makeup. Rapid scale and corrosion held COC near 3, so daily blowdown volumes stayed high. Limited fresh supply left little room to grow IT load without reclaiming blowdown as high-quality makeup.
Utility bills rise fastest where both makeup and sewer are metered at industrial rates. On this campus, discharge fees tracked recovered volume closely, so every gallon reclaimed cut two line items at once. That double credit is why hybrid recovery cleared internal ROI gates faster than RO with off-site brine haul.
How do high-purity water systems support cooling?
High-purity water systems for data center cooling keep circulating TDS and chlorides low enough to raise COC to 5 or higher. Many owners also target Water Usage Effectiveness (WUE) below 1.2 L/kWh on the same metering boundary. Diagnosis starts with blowdown rate, makeup quality, Langelier Saturation Index (LSI), and microbial counts such as ATP.
Most plants we size for arid campuses run makeup TDS near the high end of the local well range. RO feed TDS was held below 1,500 mg/L for a high-efficiency RO system for cooling water purification. Loop chlorides were targeted below 200 mg/L to limit corrosion. LSI control ahead of the membranes curbed carbonate scale.
Makeup water, corrosion inhibitors, biocides, and airborne dust all add dissolved solids over time. Influent at this site showed TDS 1,800 mg/L and chloride 450 mg/L, which capped COC. Project goals were at least 90% TDS removal from blowdown, 40% less makeup, and COC of 6 or higher with stable uptime.
| Parameter | Makeup Water Quality (Influent) | RO Feed Target | Treated Effluent Goal |
|---|---|---|---|
| Total Dissolved Solids (TDS) | 1,800 mg/L | <1,500 mg/L | <100 mg/L (RO Permeate) |
| Chlorides (Cl⁻) | 450 mg/L | <400 mg/L | <20 mg/L (RO Permeate) |
| Hardness (CaCO₃) | 350 mg/L | <50 mg/L | <5 mg/L (RO Permeate) |
| Langelier Saturation Index (LSI) | +1.2 | <0.5 | N/A |
| Water Recovery Target | N/A | 75% (RO Stage) | 95% (Overall System) |
Hybrid ZLD and RO train for blowdown recovery
The selected train used pretreatment, a two-stage RO block, and a ZLD crystallizer on the RO brine. That sequence handles high mineral blowdown while returning permeate and distillate as makeup. Field teams favor this split when brine haul is costly or discharge is banned.
Pretreatment: Multi-media filtration removed suspended solids before ion exchange. Softening cut hardness from 350 mg/L to less than 50 mg/L. Antiscalant at 5–10 ppm and sulfuric acid held LSI below 0.5 ahead of the membranes.
Reverse osmosis: Pretreated blowdown fed a 2-stage high-efficiency RO system for cooling water purification. Elements were 8-inch Dow Filmtec BW30-400 spiral-wound thin-film composites. Design recovery was 75% at about 200 psi average pressure, with 95% TDS rejection and 10–15 gfd flux.
Loop hygiene used an on-site ClO₂ generator for microbial control in cooling loops at 0.5–1 ppm. That dose controlled biofilm without sending free chlorine residuals that would damage polyamide membranes.
ZLD: RO reject went to a SaltMaker evaporative crystallizer (Saltworks). That stage recovered about 90% of brine water as distillate. The residual solids left as a dry salt cake for non-hazardous landfill disposal. Overall system recovery reached 95% when permeate and distillate both returned to the towers.
Controls: A PLC with SCADA tracked TDS, pH, conductivity, flow, and pressure across skids. Automated dosing and alarms supported 99% uptime after commissioning. Operators adjusted setpoints from conductivity and LSI trends rather than fixed calendar dosing alone.
Skid hydraulics kept RO concentrate above crystallizer minimum flow during low IT load nights. A small concentrate tank buffered those dips so the evaporator did not cycle off on every load swing. That buffer was a late pilot lesson written into the full-scale P&ID.
| Component | Key Specification | Operating Parameter |
|---|---|---|
| Pretreatment (Softener) | Ion Exchange Resin (Strong Acid Cation) | Hardness Reduction: >95% |
| RO Membranes | Dow Filmtec BW30-400 (8-inch spiral-wound) | TDS Rejection: 95% |
| RO Operating Pressure | 2-stage configuration | 200 psi (average) |
| RO Flux Rate | Optimized for longevity | 10-15 gfd |
| RO Water Recovery | Stage-specific | 75% (RO stage) |
| ZLD Evaporative Crystallizer | SaltMaker (Saltworks) | Brine Water Recovery: 90% |
| Chemical Dosing (Antiscalant) | Proprietary polymer blend | 5-10 ppm |
| Chemical Dosing (Biocide) | Chlorine Dioxide | 0.5-1 ppm |
| pH Adjustment | Sulfuric Acid | Maintain LSI < 0.5 |
Pilot, modular build, and startup fixes
Successful industrial water projects usually prove the chemistry at reduced flow before locking CAPEX. A 3-month pilot treated about 10% of blowdown flow, near 50 m³/h. The pilot confirmed COC movement from 3 toward 6 and permeate TDS below 500 mg/L before full spend approval.
Full-scale units arrived on skids as prefabricated treatment systems for rapid deployment. Off-site fabrication cut on-site mechanical and electrical work from about 12 weeks to 8 weeks. That 40% shorter install kept the tower plant online during changeover. Design through commissioning took roughly 6 months.
Early RO fouling from residual organics and colloidal silica responded to citric acid cleans and antiscalant tweaks. ZLD heat-exchanger scale eased after tighter pH and antiscalant control. Cooling-tower microbial spikes fell once on-site ClO₂ generators for microbial control in cooling loops dosing was optimized.
Commissioning sequenced softener regeneration, RO permeate quality, then crystallizer heat balance. Teams held tower COC near the old setpoint until permeate conductivity stabilized for several days. Only then did operators step COC from 3 toward 6.5 in controlled increments while watching LSI and chloride.
Those startup notes matter for owners copying this data center cooling case study layout. Budget CIP chemical, spare membranes, and crystallizer inspection ports in the first-year OPEX, not as afterthoughts.
Keep spare antiscalant and acid totes on site for the first quarter. Lead times on specialty polymers can stretch past two weeks and stall recovery if LSI drifts.
How effective is data center water recycling?
Data center water recycling at this site cut makeup demand 40%, from 345 to 207 million gallons per year. COC rose from 3 to 6.5 while the package reclaimed 95% of blowdown. That is the measured effectiveness for this hybrid RO-ZLD train under the stated groundwater makeup quality.
Water and sewer fees dropped $1.7 million per year. Against $3.2 million CAPEX, the early target was about 2 years simple payback. The tracked calculation using ($1.7M − $0.5M OPEX) / $3.2M gave about 2.67 years. Cooling-tower chemical use fell 30%, and uptime rose from 95% to 99%.
NPDES compliance held without liquid brine discharge. Legionella counts stayed below 10 CFU/mL per ASHRAE 188-2021. Cleaner circulating water also reduced pumping energy and slowed corrosion on chillers and exchangers.
Secondary benefits showed up in maintenance tickets. Heat-exchanger pull intervals lengthened once circulating chloride and hardness fell. Operators also reported fewer unplanned biocide shocks after ClO₂ dosing stabilized in the 0.5–1 ppm band.
From a permitting view, eliminating liquid brine removed a recurring NPDES negotiation point. Solid salt cake handling still needed landfill acceptance paperwork, but that path was simpler than seeking a higher blowdown mass loading.
| Metric | Pre-Treatment Performance | Post-Treatment Performance | Improvement |
|---|---|---|---|
| Cooling Water Consumption | 345 million gallons/year | 207 million gallons/year | 40% Reduction (138M gallons/year) |
| Cycles of Concentration (COC) | 3 | 6.5 | +117% |
| Annual Water/Sewer Savings | $0 | $1.7 million | N/A (New Savings) |
| System Uptime | 95% | 99% | +4% |
| Chemical Usage (Cooling Tower) | Baseline (high) | 30% Reduction | 30% |
| NPDES Compliance | Challenging | Full Compliance | Achieved |
What treatment fits closed-loop cooling towers?
Reverse osmosis suits moderate-salinity blowdown below about 2,000 mg/L TDS when a brine outlet still exists. RO commonly delivers about 95% TDS removal at lower energy than full evaporation. Membranes still foul without solid pretreatment and hardness control.
Stand-alone ZLD can push overall recovery toward 99% and handle influent above 3,000 mg/L TDS. Capital cost and energy run much higher than RO alone. MBR systems for wastewater reuse in data centers remove organics and solids for non-potable reuse. They do not replace high-TDS blowdown desalting by themselves.
Hybrid RO plus ZLD uses RO for bulk desalting, then crystallizes the brine. Overall water savings typically land in the 85–95% band. Indicative unit costs in the original analysis were RO CAPEX $0.50–$1.20/m³ and ZLD $2.00–$4.00/m³. Hybrid packages sat near $1.50–$3.00/m³ treated with OPEX about $0.80–$1.50/m³.
Closed-loop cooling towers still evaporate water; “closed loop” here means the treatment plant returns purified blowdown instead of sending it to sewer. Owners sometimes confuse that with adiabatic dry coolers that avoid evaporative loss entirely. Those dry systems cut water use further but shift load to power and are outside this case boundary.
| Technology | Pros | Cons | Best Use Case for Data Centers | Typical CAPEX/OPEX Range |
|---|---|---|---|---|
| Reverse Osmosis (RO) | 95% TDS removal, relatively low energy, reliable | Brine disposal needed, membrane fouling risk | Moderate TDS cooling blowdown (<2,000 mg/L), high-quality makeup water production | CAPEX: $0.50-$1.20/m³ OPEX: $0.20-$0.50/m³ |
| Zero Liquid Discharge (ZLD) | 99% water recovery, no liquid discharge, handles high TDS | High capital cost, energy-intensive, complex operation | High TDS blowdown (>3,000 mg/L), zero-discharge mandates, severe water scarcity | CAPEX: $2.00-$4.00/m³ OPEX: $1.00-$2.50/m³ |
| Membrane Bioreactor (MBR) | Effective organic removal, compact footprint, high effluent quality for reuse | Higher OPEX, limited to organic contaminants, not ideal for high TDS | Treating domestic/process wastewater for non-potable reuse (e.g., irrigation, toilet flushing) | CAPEX: $0.80-$1.50/m³ OPEX: $0.40-$0.80/m³ |
| Hybrid (RO + ZLD) | Balances cost/recovery, 85-95% water savings, robust for complex contaminants | Higher initial investment than RO alone | High TDS blowdown where maximum water recovery and minimal discharge are critical | CAPEX: $1.50-$3.00/m³ OPEX: $0.80-$1.50/m³ |
Cost drivers and payback for hybrid recovery
The $3.2 million CAPEX broke down as $2.5 million equipment, $500,000 installation, and $200,000 engineering. Annual OPEX was about $500,000. Energy ran near $0.30/m³ ($1.14/1000 gallons) of treated water. Chemicals were about $0.15/m³, labor $0.10/m³, and maintenance $0.20/m³.
Recovered water avoided about $0.80/m³ ($3.03/1000 gallons) in combined purchase and discharge fees. That line alone totaled $1.7 million per year. Tower chemical optimization added roughly $0.10/m³, or about $0.1 million more. Net annual benefit after OPEX supported the ~2.67 year payback shown below.
Energy dominates hybrid OPEX once the crystallizer is online. At $0.30/m³ treated in this dataset, a 10% rise in power price stretches payback by roughly a quarter year if water tariffs stay flat. That sensitivity is why the pilot logged steam or electric duty hours before the full-scale purchase order.
Use this selection checklist before you freeze scope. Measure makeup TDS, chloride, and hardness under peak heat load. Confirm the COC ceiling before scale or corrosion limits bind. Price local water and sewer tariffs, including peak discharge fees. Verify NPDES or zero-liquid rules that apply to the outfall.
Pilot at about 10% flow for 8–12 weeks before full purchase. Compare RO-only brine haul against an on-site crystallizer using real trucking quotes. Model evaporator energy at the actual brine duty, not brochure averages. For deeper ZLD cost ranges, see the cost analysis for ZLD and brine treatment systems.
| Category | Item | Case Study Value |
|---|---|---|
| Capital Expenditure (CAPEX) | Equipment | $2.5 Million |
| Installation | $0.5 Million | |
| Engineering | $0.2 Million | |
| Total CAPEX | $3.2 Million | |
| Operational Expenditure (OPEX) | Energy (per m³ treated) | $0.30 |
| Chemicals (per m³ treated) | $0.15 | |
| Labor (per m³ treated) | $0.10 | |
| Maintenance (per m³ treated) | $0.20 | |
| Total Annual OPEX | ~$500,000 | |
| Annual Savings | Water/Sewer Fees Reduction | $1.7 Million |
| Chemical Usage Reduction | $0.1 Million | |
| Total Annual Savings | $1.8 Million | |
| Financial Metrics | Calculated Payback Period | ~2.67 Years |
Who this is for and next step
This approach fits campuses with makeup TDS near 1,800 mg/L, COC stuck near 3, and rising water or discharge fees. Sites with cheap sewer capacity and low tariffs may stay on RO-only with brine haul. Owners chasing closed-loop cooling with zero liquid discharge should stress-test energy cost before they skip the RO stage.
If you are sizing a similar hybrid package, share blowdown flow, makeup analysis, and tariff data through our request-quote form for cooling water recovery.
Frequently Asked Questions
How does RO brine disposal work in a ZLD system?
RO reject feeds an evaporative crystallizer that boils off remaining water and condenses distillate for reuse. Dissolved solids leave as a dry salt cake. That cake is typically sent to a non-hazardous landfill, so no liquid brine leaves the site. On this project the ZLD stage recovered about 90% of brine water before solids handling.
When should a hybrid RO-ZLD beat stand-alone RO or ZLD?
Choose hybrid RO-ZLD when you need 85–95% recovery, face zero-liquid rules, or pay very high water and disposal rates. Stand-alone RO is cheaper at moderate salinity but still makes brine. Full ZLD handles TDS above 3,000 mg/L well, yet capital and energy run higher. The hybrid keeps RO on the bulk flow and limits crystallizer duty to the concentrate.
How does recycling change Water Usage Effectiveness (WUE)?
Higher COC and blowdown reuse cut fresh makeup per kilowatt-hour of IT load, which lowers WUE. This facility’s 40% makeup cut, from 345 to 207 million gallons per year at COC 6.5, is the direct driver. Exact WUE still depends on IT load and free-cooling hours, so trend WUE against the same metering boundary before and after startup.
What maintenance does a hybrid RO-ZLD package need?
Plan CIP of RO membranes with citric or alkaline cleaners, dosing-pump calibration, and crystallizer heat-exchanger inspection. SCADA alarms on differential pressure, conductivity, and LSI help schedule work before trips. Routine lab checks on TDS, chloride, hardness, and Legionella guide antiscalant and ClO₂ setpoints and protect the 99% uptime target.
