What a New Territories data center actually has to treat
Three wastewater streams converge at a hyperscale site in Tsuen Wan, Sha Tin, Fanling or Yuen Long, and the spec should start by metering each one separately rather than treating "data center wastewater" as a single thing. Cooling-tower blowdown (CTBD) is the largest by both volume and dissolved load: an evaporative system concentrates make-up water until scale-forming ions and treatment chemicals reach their limit, and the operator purges a controlled fraction as blowdown. Process wastewater comes next — humidification bleed, chiller bleed, equipment washdown, and the periodic test water from diesel generator sets. Site run-off from rooftops, loading bays and paved areas rounds out the inventory, generally low in dissolved load but variable in suspended solids and any fuel or oil contamination from vehicle movement.
For a ~100 MW hyperscale facility, daily water use is on the order of 2 million litres, with a meaningful fraction leaving as CTBD (IDE Tech, 2026). Hong Kong specifics reshape that baseline. Many New Territories sites draw fresh water from the Dongjiang supply and reclaimed water where the Water Supplies Department makes it available, but sites in Tsuen Wan, Sha Tin and the northwest New Territories pull from coastal or estuarine sources, which lifts baseline chloride, bromide and TDS in make-up water and therefore in CTBD relative to inland sites. Discharge of industrial wastewater to the sewer is controlled under the Water Pollution Control Ordinance (Cap. 358) via an EPD discharge licence with site-specific effluent quality limits, and discharges to waters inside Victoria Harbour and to Deep Bay are tightly restricted, which is the first design driver for any blowdown strategy.
Cooling-tower blowdown chemistry and why cycles of concentration matter
Cycles of concentration (COC) is the ratio of dissolved solids in the circulating cooling water to the make-up water feeding it. Raising COC from 3 to 6 roughly halves blowdown volume, but the trade-off is that silica, calcium, magnesium and alkalinity climb sharply in the circulating water and so do the biocides, scale inhibitors and corrosion inhibitors added to keep the system in service. A cooling tower at 4 COC typically loses around 25–30% of make-up water to blowdown; for a facility using 10 million gallons per month that is 2.5–3 million gallons per month discharged (Genesis Water Tech, 2026).
CTBD chemistry is brackish. Total dissolved solids sit at 1,200–6,000 mg/L — roughly 4–8× the make-up water — with elevated calcium, magnesium, silica and alkalinity, accumulated treatment chemicals, suspended solids in the 10–50 mg/L range from corrosion products and biofilm fragments, and a steady background biological load (Genesis Water Tech, 2026). Side-stream filtration taking 1–5% of circulation flow is the standard pre-condition for higher COC: it cuts suspended solids, biological loading and corrosion products, and protects any downstream membrane system. On a New Territories site where ambient humidity is high and intake TDS is already elevated, the practical upper COC without blowdown treatment is often 4–5; pushing past 6 almost always requires a reuse train to stay inside the EPD discharge licence.
| Cooling-tower operating point | Approx. CTBD TDS (mg/L) | Blowdown as % of make-up | Scale risk | Treatment implication |
|---|---|---|---|---|
| COC 3 (baseline, no side-stream) | 900–2,400 | ~33% | Low to moderate | Discharge to sewer with basic filtration |
| COC 4–5 (typical operating window) | 1,200–3,000 | ~20–25% | Moderate; silica approaches limit | Side-stream filtration required; EPD licence feasibility study needed |
| COC 5–6 (with side-stream filtration) | 1,500–3,600 | ~17–20% | High; silica and calcium sulfate scaling | Membrane polishing or NF to permit higher COC |
| COC 6–8 (reuse-driven) | 1,800–6,000 | ~12–17% | Severe without RO permeate dilution | RO permeate blend, antiscalant program, conservative recovery |
The bands above are not project quotations; they are derived from the COC, blowdown-fraction and CTBD TDS ranges in Genesis Water Tech (2026) and are intended to bracket the operating envelope, not to specify a particular plant. The actual TDS for a given COC depends on make-up water chemistry, holding-time index, and the antiscalant program in service.
Discharge, reuse or zero liquid discharge: a Hong Kong decision framework
Three end-use strategies exist for CTBD — direct discharge under licence, cooling-tower make-up reuse, and zero liquid discharge — and each carries a different Hong Kong risk profile (Genesis Water Tech, 2026). The first question is not technical, it is regulatory: will the EPD issue a discharge licence for the proposed effluent quality and flow at this site, and will the receiving environment accept it? Discharges to the public sewer are charged and constrained by the licence's effluent quality limits on TSS, BOD, COD, total nitrogen and, where the receiving water is saline-influenced, salinity and chloride. Some jurisdictions cap discharge TDS below 1,500 mg/L and levy fees of $5–15 per 1,000 gal; "just send it to the drain" is rarely free in water-stressed regions (Genesis Water Tech, 2026). Hong Kong's specific added risk is that several New Territories sites sit near Deep Bay or along Rambler Channel, where the receiving-water sediment sensitivity raises the bar on any salinity, heavy-metal or nutrient load.
Cooling-tower make-up reuse is the highest-value path for a Hong Kong site. It lifts COC, cuts fresh water withdrawal from the Dongjiang supply, and reduces both sewer volume and any marine discharge, but it requires RO or NF quality permeate at 50–85% recovery and a real commitment to managing the concentrate stream (Genesis Water Tech, 2026). Zero liquid discharge, with 95–99% overall recovery and the cost bands typical of $3–$8M CAPEX and $5–$15 per 1,000 gal OPEX, is hard to justify in Hong Kong where marine discharge under licence is usually available, unless the site is in a no-discharge zone, has an aggressive landlord WUE cap, or faces a receiving-water body that the EPD will not licence (Genesis Water Tech, 2026). On the carbon side, the Open Engineering LCA (2026) finds that reuse adds about 2× the GWP of freshwater, with treatment energy accounting for roughly 80% of the difference; on a decarbonised grid the carbon penalty becomes negligible, so the freshwater-saving case carries the decision. A useful additional reference point for non-Hong Kong permitting and discharge-fee context is the data center cooling blowdown treatment in New York City guide, which runs a similar decision tree against a different receiving-water regime.
| End-use strategy | Typical recovery | 2026 USD CAPEX band | 2026 USD OPEX band | Hong Kong fit |
|---|---|---|---|---|
| Discharge compliance (sewer or marine outfall under EPD licence) | n/a — single pass | Lowest; basic filtration and chemistry trim | Discharge fees and any neutralisation chemicals | Default where receiving water is not Deep Bay/Rambler Channel and limits are achievable |
| Cooling-tower make-up reuse (NF/RO permeate blend) | 50–85% | RO skid $250k–$500k for 50,000 GPD; add UF pretreatment | $1.50–$3.00 per 1,000 gal (RO) | Highest-value path for hyperscale sites on Dongjiang or coastal make-up |
| Zero liquid discharge (RO + MVC + crystalliser) | 95–99% | $3M–$8M system | $5–$15 per 1,000 gal | Only where discharge is blocked by EPD or landlord cap; rarely economic in Hong Kong |
Selecting the treatment train: side-stream filtration, UF, NF, RO and MVC
The pretreatment floor for any membrane-based reuse train is side-stream filtration at 10–25 micron on the cooling-water recirculation loop, sized at 1–5% of circulation flow (Genesis Water Tech, 2026). When CTBD is sent to membranes, an UF pretreatment skid at 0.01–0.1 micron and around 90–95% recovery is the standard choice, running at low pressure (10–30 psi), backwashing on permeate, and tolerating Hong Kong's high-humidity feed without extensive chemical conditioning. UF alone is the right answer when the only objective is biological and particulate control ahead of discharge.
Reverse osmosis is the workhorse for high-recovery reuse. RO delivers 95–99% dissolved-solid removal with permeate TDS of 10–50 mg/L, which is well suited to high-silica Hong Kong make-up mixes, and lets the operator lift COC aggressively by blending RO permeate back into the tower. Recovery on CTBD is limited to 50–85% by scaling potential as concentrate TDS climbs, and antiscalant injection is mandatory; a 50,000 GPD RO treating blowdown installs for $250,000–$500,000 with operating costs of $1.50–$3.00 per 1,000 gal (Genesis Water Tech, 2026). For a higher-recovery reuse train with stable concentrate chemistry, an industrial RO system configured at conservative local recovery with controlled salt precipitation and dynamic cross-flow operation can push overall recovery beyond 85% on silica- and calcium-rich CTBD (IDE Tech, 2026).
Nanofiltration sits between UF and RO. NF at 75–150 psi removes most hardness, sulfate and a fraction of dissolved solids while letting chloride pass, gives 70–85% recovery, and produces permeate at roughly 30–50% of feed TDS — a fit for New Territories sites where the EPD discharge limit is set on hardness rather than total TDS (Genesis Water Tech, 2026). Mechanical vapour compression is the thermal step that pushes overall system recovery to 85–95%: MVC on RO concentrate gives 95–98% recovery with distillate below 10 mg/L TDS, but consumes 15–25 kWh per 1,000 gal and costs $1–$3M for a 10,000–30,000 GPD unit (Genesis Water Tech, 2026). A common New Territories baseline is therefore: side-stream filtration on the cooling loop → UF → brackish RO at conservative recovery → RO permeate blended back as cooling-tower make-up → RO concentrate to sewer under EPD licence, with MVC added only where the licence or the receiving water will not accept the concentrate. For context on how broader water-use reduction interacts with this train, the reduce water usage in manufacturing guide covers the WUE and PUE trade-offs in more detail, and the AI in wastewater treatment 2026 trends piece covers where closed-loop control and membrane-flux optimisation are moving.
| Unit process | Removal role | Recovery | Permeate / product quality | Operating pressure or energy | Indicative 2026 USD cost |
|---|---|---|---|---|---|
| Side-stream filtration (10–25 micron) | SS, corrosion products, biofilm fragments | n/a (bleed) | Cooling loop cleanliness, not a permeate stream | Low; mechanical scraping | $50,000–$200,000 typical data center installation |
| UF (0.01–0.1 micron) | SS, bacteria, viruses, high-MW organics | 90–95% | RO feed quality | 10–30 psi | Add to RO package; modest incremental CAPEX |
| NF | Hardness, sulfate, partial TDS | 70–85% | Permeate TDS ~30–50% of feed | 75–150 psi | Lower than RO at equivalent flow; specific to skid |
| BWRO | 95–99% dissolved solids, silica, hardness | 50–85% on CTBD | Permeate 10–50 mg/L TDS | 150–400 psi | $250k–$500k installed for 50,000 GPD; OPEX $1.50–$3.00 per 1,000 gal |
| MVC | 95–98% recovery on RO concentrate | Drives overall system to 85–95% | Distillate <10 mg/L TDS | 15–25 kWh per 1,000 gal | $1M–$3M for 10,000–30,000 GPD |
Putting it together: a New Territories sizing and supplier checklist
Before asking a supplier to quote, the specifier needs a defined site data package: full make-up water analysis (TDS, chloride, bromide, silica, calcium, magnesium, alkalinity, TOC), peak ambient wet-bulb for the cooling-tower design case, target COC and WUE/WUI, intended EPD discharge route, and the landlord's water budget cap. The train should be specified by end-use first — define permeate TDS, target COC, and maximum discharge volume, then match UF, NF, RO and MVC to that envelope, asking suppliers for guaranteed recovery and antiscalant regime rather than nominal flux. The CAPEX bands the spec should anchor against in 2026 USD are: side-stream filtration $50,000–$200,000 for a typical data center installation; RO skid $250,000–$500,000 for 50,000 GPD; MVC $1–$3M for 10,000–30,000 GPD; full ZLD $3M–$8M (Genesis Water Tech, 2026). Lifecycle energy, membrane replacement and antiscalant chemicals typically dominate OPEX, so the request for quotation should ask for an itemised OPEX rather than a single per-1,000-gal figure. Hong Kong-specific items the supplier must confirm: prior EPD discharge licence experience in the New Territories, a local agent for membrane elements and chemicals, and a credible lead time for RO elements during a hyperscale build-out, since global RO element supply is the realistic schedule risk.
Frequently Asked Questions
How does the EPD discharge licence pathway work for cooling-tower blowdown under the Water Pollution Control Ordinance?
Discharge of industrial wastewater to the public sewer or to waters is controlled under the Water Pollution Control Ordinance (Cap. 358) and requires an EPD discharge licence with site-specific effluent quality limits on parameters such as TSS, BOD, COD, total nitrogen, heavy metals, and where the receiving water is saline-influenced, salinity and chloride. The first step is a pre-application meeting with EPD to confirm that the proposed discharge route and quality are licensable; this should happen before the treatment train is frozen, because the answer drives both the COC target and the concentrate disposal path. Cap. 358 and the EPD's Technical Memorandum on standards are the controlling instruments, and discharges to waters inside Victoria Harbour and to Deep Bay are tightly restricted, which is often the binding constraint for northwest New Territories sites.
What CAPEX and OPEX bands should we anchor our budget to in 2026?
For 2026 USD, a 50,000 GPD brackish RO skid for CTBD reuse installs for $250,000–$500,000 with OPEX of $1.50–$3.00 per 1,000 gal treated, an MVC evaporator for 10,000–30,000 GPD sits at $1M–$3M, and a full ZLD system lands in the $3M–$8M range with $5–$15 per 1,000 gal OPEX (Genesis Water Tech, 2026). Side-stream filtration on the cooling loop typically costs $50,000–$200,000. These are bracketing bands, not project quotations; request an itemised OPEX that separates energy, membrane replacement, antiscalant and labour so you can stress-test the lifecycle cost against your landlord WUE cap.
How do we choose between RO and nanofiltration for our New Territories source water?
Use RO when silica is the binding scale risk and when you need permeate in the 10–50 mg/L TDS range to lift COC aggressively; use NF when the discharge or reuse constraint is set on hardness rather than total TDS, and when the lower operating pressure (75–150 psi vs RO's 150–400 psi) and higher recovery (70–85%) make the case on energy and concentrate volume (Genesis Water Tech, 2026). For most New Territories sites with saline-influenced make-up and an EPD licence sensitive to chloride, RO is the safer default and NF is a partial-softening pre-step rather than a stand-alone solution.
What supplier-selection criteria matter most for a Hong Kong hyperscale build?
Confirm that the supplier has prior EPD discharge licence experience in Hong Kong, can name a local agent for membrane elements and antiscalant chemicals, and can commit to a credible lead time for RO elements during the construction window — global RO element supply is the realistic schedule risk on a hyperscale build. Ask for guaranteed recovery and antiscalant regime, not just nominal flux, and for an itemised OPEX that lets you model membrane replacement and chemical cost separately. The supplier should also be able to reference a New Territories or Pearl River Delta site where their CTBD train has been running long enough to validate the recovery and antiscalant claims.
