Why Cape Town data centres face a water problem that evaporative cooling cannot solve
Evaporative cooling in Cape Town is a permanent loss, not a recyclable draw. Dam storage recovered from the 2018 "Day Zero" crisis, but the political memory of standpipes and per-capola limits has not — and neither has the hydrological ceiling. The two data centres approved in 2026 in the airport industrial area total more than 120,000 m² of floor space, and city modelling projects the cluster could consume up to 34% of Cape Town's electricity at full build-out (EWN, 2026-07). On the water side, the figures have been noisier: a 12 ML/year claim and a 4.4-billion-litre claim have both circulated publicly (S3, S5). Engineering consensus for an evaporative hyperscale site in a Mediterranean climate sits in the 1.8–2.5 ML/MW-year band, so the 4.4-billion-litre figure is at least an order of magnitude high; the 12 ML figure is closer to a single 5–6 MW facility's annual blowdown plus evaporation. The New York State moratorium on new data centre construction, imposed to review environmental impact, is the international precedent city planners now cite when they say "no" (EWN, 2026-07). In 2026 Cape Town, the design constraint is therefore not whether to cool, but how to honour a social licence that treats every evaporated kilolitre as a withdrawn public resource.
Cooling-tower blowdown chemistry: what is actually in the wastewater
Blowdown is the source water concentrated by the cycle factor, plus every chemical the cooling-water programme has added to keep the loop operational. Cape Town municipal potable feed typically runs at TDS 150–280 mg/L, total hardness 60–120 mg/L as CaCO₃, and chloride 80–150 mg/L (CoCT water-quality data, 2025). At cycles of concentration (CoC) of 3×, blowdown TDS is roughly 450–840 mg/L; at 6× it climbs to 900–1,680 mg/L; at 8× it can exceed 2,200 mg/L — and chloride pushes the 304/316 stainless condenser tubes into pitting-corrosion territory above ~500 mg/L Cl⁻. On top of the concentrated background, operators dose oxidising biocides (Cl₂, ClO₂, bromine) at 0.5–5 mg/L residual, non-oxidising biocides in intermittent slugs that can spike to 50+ mg/L, plus phosphonate- or polymer-based antiscalants at 2–10 mg/L, and azole corrosion inhibitors at 1–5 mg/L. A February 2026 TNFD case study confirmed the load profile to design around: salts, copper and zinc stripped from condenser tubes, and residuals of the chemical programme (per the TNFD case study, 2026-02). The implication is that blowdown is not "concentrated drinking water" — it is a mixed industrial wastewater that pretreats poorly and discharges inconsistently if chemistry is left unmanaged.
Cape Town regulatory ceiling: Industrial Effluent By-law discharge limits
The City of Cape Town Industrial Effluent By-law (2014) is the binding compliance instrument for any data centre discharging cooling-tower blowdown to the municipal sewer. The headline limits: TDS ≤1,000 mg/L, pH 5.5–9.5, free chlorine ≤50 mg/L, sulphates ≤500 mg/L, and zero visible oil or grease. Metals ceilings are tight — Zn ≤5 mg/L, Cu ≤2 mg/L, Fe ≤5 mg/L — which makes corrosion-byproduct control at the condenser mandatory, not optional. Direct discharge to stormwater or to the Liesbeek, Black River, or Salt River catchments requires a Section 21 water-use licence from the Department of Water and Sanitation, a process that routinely runs 12–18 months. For 2026 builds, the City additionally requires a Water-Sensitive Design (WSD) submission alongside building plans, using the CoCT WSD Manual as the design checklist. In practice, this means the blowdown treatment train has to be specified in the planning application, not retrofitted after commissioning — a sequencing change that has caught more than one hyperscale developer off guard.
The treatment train: side-stream filtration, softening, RO, and disinfection in series
The defensible 2026 train for a Cape Town hyperscale site runs four stages in series, with a sidestream architecture that recycles 75–85% of the blowdown volume back to the cooling-tower basin as make-up.
Stage 1 — Side-stream filtration. Automatic self-cleaning strainers at 50–100 μm ahead of the cooling-tower basin remove particulate carryover (leaf debris, airborne dust, scale chips) and cut RO fouling load by 60–80%. A multi-media filter polisher ahead of the RO reduces SDI below 3, which is the working ceiling for spiral-wound elements on Cape Town feed.
Stage 2 — Softening. Once CoC exceeds 5×, lime/soda or weak-acid cation (WAC) is required to drop Ca²⁺/Mg²⁺ before the RO. A 100–300 mg/L Ca(OH)₂ dose typically takes hardness from 600 mg/L as CaCO₃ down to <50 mg/L. Sludge is handled in a lamella clarifier operating at 3–5 m/h surface overflow rate, producing a thickened underflow for the press.
Stage 3 — Side-stream RO. A slipstream of 10–20% of total blowdown is processed through an industrial RO system at 75–85% recovery. Permeate is recycled to cooling-tower make-up, dropping freshwater draw by 40–60% versus a baseline 3-cycle system. Concentrate (~20–25% of slipstream flow) goes to brine management or, on constrained sites, to a downstream evaporator. A PLC-controlled chemical dosing system meters antiscalant proportional to RO feed flow and trims pH to 7.0–7.5 with sulphuric acid to keep Langelier Saturation Index in the −0.2 to +0.2 band.
Stage 4 — Disinfection. An on-site chlorine dioxide generator dosing 0.5–2 mg/L ClO₂ achieves 99.9% microbial kill in <15 min contact time and does not form trihalomethanes the way NaOCl does. ClO₂ is the specifier's choice for the final effluent because it is effective at the high-pH and high-TDS window the concentrate stream sits in. Sludge from Stages 2 and 3 is dewatered in a plate-and-frame filter press to 22–28% dry solids for off-site disposal to a licensed hazardous landfill.
| Stage | Equipment | Inlet spec | Outlet spec | Removal / key parameter |
|---|---|---|---|---|
| 1. Side-stream filtration | Auto strainer + multi-media filter | TSS 20–80 mg/L | TSS <5 mg/L, SDI <3 | 60–80% TSS cut |
| 2. Softening | Lime/soda or WAC + lamella clarifier | Hardness 400–600 mg/L as CaCO₃ | <50 mg/L as CaCO₃ | >90% hardness removal |
| 3. Side-stream RO | Industrial RO unit, 75–85% recovery | TDS 900–2,200 mg/L | Permeate <50 mg/L TDS; concentrate 4,000–6,000 mg/L | 95–98% TDS rejection |
| 4. Disinfection | ClO₂ generator, 0.5–2 mg/L | Heterotrophic count 10⁴–10⁶ CFU/mL | <100 CFU/mL | 99.9% kill in <15 min |
| 5. Sludge handling | Lamella clarifier + plate-and-frame press | Softening sludge 2–4% DS | Cake 22–28% DS | Cake volume reduction ~90% |
For projects where the City requires nutrient reduction in the sewer discharge (typically where multiple hyperscale sites share a sub-catchment), an MBR polishing step can be added between the RO concentrate stream and the sewer — a configuration we covered in the electrodialysis for salt removal guide and the Dammam data centre cooling blowdown guide.
Cooling-strategy comparison: evaporative, adiabatic, and hybrid liquid-to-liquid for Cape Town
Choosing a cooling topology in 2026 is no longer a thermal-engineering decision in isolation — it is a water-and-power co-optimisation problem driven by the 34% grid-share projection and the public-perception cost of evaporation.
| Strategy | WUE (L/kWh) | PUE penalty | CAPEX (relative) | Water draw | Cape Town fit (2026) |
|---|---|---|---|---|---|
| Evaporative cooling tower | 1.0–2.0 | +0.1–0.2 | Lowest | Highest | Misaligned with Day Zero narrative; default for legacy builds |
| Adiabatic (dry + wet assist) | 0.3–0.5 | +0.3–0.4 | Moderate | Moderate | Suited to Mediterranean climate; only fires wet mode on ~30% of hours |
| Hybrid L2L (rear-door HEX / cold plate + dry cooler) | 0.1–0.3 | +0.5–0.8 | Highest | Lowest | Preferred where grid capacity exists and water allocation is capped |
For sites under 30 MW IT load facing the 2026–2028 Eskom-constrained grid, pair adiabatic ambient cooling with side-stream RO on blowdown — this combination cuts water draw by ~60% versus a 3-cycle evaporative baseline while keeping PUE penalty below 0.4. For >50 MW sites where the City's water allocation is the binding constraint, specify hybrid L2L with closed-loop chemistry (glycol-water or dielectric fluid) and reserve the RO slipstream for make-up polishing. The South African data centre cooling market is forecast to grow from USD 32.6M in 2025 to USD 61.5M by 2030 at a 9.5% CAGR (MarketsandMarkets, 2025), driven largely by the same water-and-power co-optimisation pressure this comparison is built around. To see how these constraints translate into a working process train for a comparable Middle East site, the Dammam data centre cooling blowdown guide is a useful parallel read.
ZLD feasibility and ROI under Cape Town's energy and water prices
Zero liquid discharge is not free, and for Cape Town it is not always necessary. The economic driver is the gap between water-in and sewer-out tariffs: industrial potable water at R45–55/kL (CoCT 2026 tariff schedule) plus effluent disposal at R15–20/kL gives a combined avoided cost of R60–75/kL for every kilolitre reused on-site. Mechanical vapour recompression (MVR) crystallisers evaporate 15–25 kWh/m³ of brine and run R12–18M in CAPEX for a 5 m³/hr unit — a 6–9 year payback on blowdown-only brine at current tariffs, and roughly 4–6 years if the site already needs softened water for process. Cape Town's >3,000 sunshine hours/year permit solar-thermal preheating of the RO concentrate, cutting MVR electrical load by 30–40% and trimming OPEX by R1.2–1.8M/year on a 5 m³/hr system. The realistic design rule: where sewer allocation is reliable and CoCT TDS ≤1,000 mg/L is achievable, side-stream RO at 80% recovery with sewer disposal of concentrate is the lowest lifecycle cost. Reserve ZLD for water-stressed sub-districts (northern fringes, Philippi aquifers) or where sewer discharge TDS is locally restricted. For broader context on brine concentration, the electrodialysis for salt removal guide covers the pre-crystalliser options in detail.
Frequently Asked Questions
What is the binding sewer discharge limit for a Cape Town data centre?
The City of Cape Town Industrial Effluent By-law (2014) sets TDS ≤1,000 mg/L, pH 5.5–9.5, free chlorine ≤50 mg/L, sulphates ≤500 mg/L, and metals ceilings of Zn ≤5 mg/L, Cu ≤2 mg/L, Fe ≤5 mg/L. Any discharge above these limits requires trade-effluent consent and may trigger a Section 21 water-use licence process.
How much water does a hyperscale data centre in Cape Town actually use?
Engineering consensus for evaporative hyperscale sites in a Mediterranean climate sits at 1.8–2.5 ML/MW-year, including both evaporation and blowdown. The 4.4-billion-litre figure circulating publicly is at least an order of magnitude too high; the 12 ML/year figure corresponds to a single 5–6 MW facility. Adiabatic or hybrid L2L topologies can cut this to 0.3–0.5 ML/MW-year or 0.1–0.3 ML/MW-year respectively.
Is zero liquid discharge required for new 2026 builds?
Not universally. ZLD is required where the municipal water allocation is capped, where the sub-district sewer cannot accept the concentrate TDS, or where the developer wants to bypass the public-perception cost of evaporation. Otherwise, side-stream RO at 75–85% recovery with sewer disposal of concentrate is the lowest-lifecycle-cost compliant path.
Why is chlorine dioxide preferred over NaOCl for the final disinfection step?
ClO₂ at 0.5–2 mg/L achieves 99.9% microbial kill in under 15 minutes and does not form trihalomethanes, which is the regulated by-product that pushes NaOCl dosing into compliance risk on high-TDS, high-organic feed. The automated chemical dosing for wastewater guide covers the control-loop design for either biocide programme.
