Why Antananarivo Rewrites the Standard Data Center Water Balance
Antananarivo sits on the central highland plateau at roughly 1,276 m elevation, where ambient conditions are tropical but cooler and less humid than coastal tropical references. That shift moves the cycles-of-concentration (CoC) decision window: makeup water is not concentrated as aggressively for a given ambient wet-bulb, so a 4–6 CoC operating band is achievable without forcing blowdown total dissolved solids (TDS) into the upper end of the generic 1,200–6,000 mg/L range quoted in global references (S1, 2025-12).
JIRAMA is the dominant municipal supplier for the capital region, drawing surface water from the Ikopa River system and associated highland reservoirs, with groundwater as a secondary makeup source. There is no seawater salinity intrusion mechanism like the Manila Bay dry-season case that drives other tropical references toward higher baseline TDS, so JIRAMA source water generally starts with a lower TDS baseline than equivalent coastal-tropical intakes (S2, 2026).
Industrial discharge in Madagascar is overseen at the national level by the Ministère de l'Environnement et du Développement Durable through the Office National pour l'Environnement (ONE). Project-specific environmental permits — the étude d'impact and the arrêté d'autorisation — set site effluent ceilings for parameters including TDS, chloride, sulfate, and hydrocarbons; the actual permit text controls, so verify current site-specific limits with ONE before design freeze. Because municipal sewerage in Antananarivo is intermittent and industrial pretreatment is not uniformly enforced at the receiving end, on-site treatment to a reuse or zero liquid discharge (ZLD) endpoint is operationally safer than discharge-to-sewer for a hyperscale load.
The Four Wastewater Streams a Madagascar Data Center Actually Produces
Cooling-tower blowdown is the largest stream by volume: 25–30% of makeup water at 4 CoC, carrying TDS of 1,200–6,000 mg/L, suspended solids 10–50 mg/L, residual biocides, scale inhibitors, and corrosion products (S1, 2025-12; S2, 2026). RO reject from humidification and process-water polishing arrives at 20–30% of feed flow with elevated TDS concentrate that re-enters the same waste train; co-treating it with blowdown is the standard configuration.
Diesel generator cooling water, floor wash, and humidifier bleed are lower-volume but intermittent, often oily, and cannot be routed to the same membrane system as blowdown without oil-removal pretreatment upstream of any ultrafiltration pretreatment skid. Stormwater contacting outdoor equipment pads is seasonal and high-volume with suspended-solids spikes during the November–April wet season; it should be segregated from the blowdown treatment train and handled by a separate sedimentation/filtration chain before any reuse or discharge.
| Stream | Peak flow character | Typical chemistry / loadings | Co-treatment decision |
|---|---|---|---|
| Cooling-tower blowdown | 25–30% of makeup at 4 CoC; near-continuous | TDS 1,200–6,000 mg/L; SS 10–50 mg/L; biocides, scale inhibitors, corrosion products | Route to UF + BWRO train |
| RO reject (humidification / process polish) | 20–30% of RO feed; near-continuous | Elevated TDS concentrate; low SS | Co-treat with blowdown in same RO |
| Generator cooling water / floor wash / humidifier bleed | Intermittent, low volume; often oily | Hydrocarbons, SS, variable TDS | Oil removal upstream; route to UF + RO |
| Seasonal stormwater (Nov–Apr) | High volume, short duration, SS spikes | High SS, low TDS, debris | Segregate; separate sedimentation/filtration |
Cooling-Tower Chemistry and Reuse Targets for an Antananarivo Site
Cooling-tower makeup reuse should land at TDS 500–1,500 mg/L, suspended solids 10–25 mg/L, pH 6.5–8.5, and silica below 150 mg/L to keep scaling risk manageable at 4–6 CoC (S1, 2025-12; S2, 2026). Those are the numbers an engineer puts in the RFQ for the industrial RO system permeate spec and the cooling-tower makeup blending ratio.
Pushing from 4 to 6+ CoC cuts blowdown volume but raises scaling risk and pushes blowdown TDS toward the 6,000 mg/L upper bound. Biocide and corrosion-inhibitor residuals accumulate proportionally with CoC, so the downstream membrane pretreatment cannot rely on the same chemical program used in the open loop — the chemistry has to be switched to formulations compatible with the membrane skid. Conventional brackish-water reverse osmosis (BWRO) caps at 75–80% recovery before silica, calcium carbonate, or calcium sulfate scaling becomes unmanageable (S4, 2026; S5, 2026); above that, a different design philosophy — controlled salt precipitation or dynamic RO operation — is required, not just more antiscalant. For a 100 MW hyperscale build, water demand can reach approximately 2 million L/day (S4, 2026; S5, 2026), which sets the upper bound on blowdown volumetric design.
Four Treatment Trains from Minimum Compliance to Zero Liquid Discharge
Option 1 — side-stream filtration only — uses a 10–25 micron self-cleaning filter on 1–5% of circulation flow, holds blowdown suspended solids below 15 mg/L, and enables higher CoC without reducing TDS. Capital runs USD 50,000–200,000 for typical data center flow rates (S1, 2025-12; S2, 2026). It is the minimum train for any Antananarivo site drawing from JIRAMA but does not by itself solve a strict ONE TDS exceedance.
Option 2 — UF + BWRO — is the default Madagascar train: ultrafiltration pretreatment ahead of an industrial RO skid, recovering 50–85% of blowdown at 10–50 mg/L TDS permeate, with operating cost of USD 1.50–3.00 per 1,000 gallons (S1, 2025-12; S2, 2026). This is the train that brings an Antananarivo site inside typical industrial discharge TDS limits in most cases, and it is the baseline against which the local ONE arrêté should be checked.
Option 3 — UF + RO + nanofiltration (NF) polish — targets sites where hardness rather than total TDS is the binding constraint; NF recovery sits at 70–85% and the NF permeate blends with RO permeate to raise overall reuse volume (S1, 2025-12). The NF step is sized to soften rather than fully demineralize, and it preserves chloride passage so the closed loop does not accumulate total dissolved solids beyond what the cooling tower can tolerate at the chosen CoC.
Option 4 — UF + RO + mechanical vapor compression (MVC) brine concentrator — is the ZLD path: 95–98% recovery, distillate below 10 mg/L TDS, 15–25 kWh per 1,000 gallons (S1, 2025-12; S2, 2026); justified when ONE clearance for site discharge is restricted or a hyperscale tenant contractually requires zero liquid discharge. The MVC distillate is high-purity and returns to the cooling-tower makeup line, while the residual brine at 20–30% dissolved solids is the only stream that still requires handling as a waste.
| Option | Train | Recovery | Permeate / distillate quality | CapEx band (50,000 GPD class) | OpEx band | When to select |
|---|---|---|---|---|---|---|
| 1 | Side-stream filtration only | No volume reduction; enables higher CoC | Same TDS as blowdown, SS <15 mg/L | USD 50,000–200,000 | Minimal | JIRAMA source, ONE permit allows current TDS |
| 2 | UF + BWRO | 50–85% | 10–50 mg/L TDS permeate | USD 250,000–500,000 | USD 1.50–3.00 / 1,000 gal | Default Madagascar train |
| 3 | UF + RO + NF polish | 70–85% (NF); blended overall higher | 30–50% of feed TDS (NF permeate) | Above Option 2 | Between Options 2 and 4 | Hardness-bound, not TDS-bound |
| 4 | UF + RO + MVC | 95–98% | <10 mg/L TDS distillate; 20–30% solids brine | USD 3–8 million (ZLD overlay) | USD 5–15 / 1,000 gal | ONE discharge restricted, tenant ZLD required |
Fouling Control on a Warm, Biologically Active Antananarivo Loop
Membrane fouling is the dominant failure mode on tropical sites due to warm, humid, biologically active feed water. Self-cleaning side-stream filtration at 10–25 micron keeps blowdown suspended solids below 15 mg/L before the UF pretreatment skid (S2, 2026). The side-stream filter is the lowest-cost insurance for the rest of the train.
Bio-organic flocculant dosing aggregates colloidal solids and biofilm fragments that bypass conventional filtration, which is a real problem in 28–32°C loop water (S1, 2025-12; S2, 2026). Antiscalant chemistry must be tuned for silica and calcium sulfate: conventional BWRO caps at 75–80% recovery before scaling becomes unmanageable (S4, 2026; S5, 2026), so an automatic antiscalant and biocide dosing skid sized for the JIRAMA source-water profile is what pushes recovery toward 85%.
On-site disinfection with a chemical-free UV sterilizer or chlorine dioxide controls planktonic bacteria and biofilm precursors in the warm loop without adding residuals that would accumulate at high CoC. Skipping any one of these four steps — side-stream filtration, bio-organic flocculation, antiscalant/biocide dosing, UV disinfection — is what drives membrane replacement intervals from 3–5 years down to 12–18 months in tropical service (S2, 2026). A multi-media pre-filter ahead of the UF stage is the lowest-cost insurance for the Scope 3 water-withdrawal number surviving audit.
CapEx, OpEx and ESG Payback for a Madagascar Hyperscale Build
An RO train sized at 50,000 GPD runs USD 250,000–500,000 CapEx and USD 1.50–3.00 per 1,000 gallons OpEx; a ZLD overlay adds USD 3–8 million CapEx and USD 5–15 per 1,000 gallons OpEx (S1, 2025-12; S2, 2026). Avoided JIRAMA commercial purchases plus avoided discharge fees — typically USD 5–15 per 1,000 gallons in water-stressed basins (S1, 2025-12; S2, 2026) — recover an RO-only CapEx in roughly 3–5 years for a Madagascar-scale facility, with hyperscale tenants at 100 MW drawing up to approximately 2 million L/day (S4, 2026; S5, 2026) reaching payback faster through sheer volume. Cross-check those numbers against the data center water reuse cost and standards 2026 reference and the data center cooling water sustainability comparison for tropical benchmarks before locking in a budget.
A 50–85% blowdown recovery rate translates directly into a Scope 3 water-withdrawal reduction that hyperscale tenants report under CDP and IFRS S2 (S2, 2026), and a multi-media pre-filter ahead of the UF stage is the lowest-cost insurance for that ESG number holding up to audit. On the compliance side, the binding instruments are the ONE environmental permit (arrêté d'autorisation) for site effluent quality, the MININEE framework for industrial water use and discharge, and the PEOS/EIS-equivalent impact study for hyperscale builds — confirm current thresholds with the actual permit before design freeze. For a comparable tropical highland context, the Konya data center blowdown guide applies the same four-train architecture to a different climate, which is useful for pressure-testing the Antananarivo selection against a peer benchmark.
Frequently Asked Questions
What CapEx envelope should we put in front of finance for a Madagascar hyperscale data center?
A baseline UF + BWRO train at 50,000 GPD sits in the USD 250,000–500,000 CapEx band with USD 1.50–3.00 per 1,000 gallons OpEx; a ZLD overlay adds USD 3–8 million CapEx and USD 5–15 per 1,000 gallons OpEx (S1, 2025-12; S2, 2026). Hyperscale tenants at 100 MW draw up to approximately 2 million L/day (S4, 2026; S5, 2026), so size the train to the tenant's contractual peak flow rather than to the 50,000 GPD reference.
How do we select a treatment supplier who can actually deliver in Madagascar?
Confirm the supplier has shipped UF, RO and MVC skids into sub-Saharan or island-African contexts before, and ask for the lead-time band on the long-lead items (membrane elements, MVC compressor, high-pressure pumps) for the JIRAMA source-water profile. Request a reference list of operating sites at similar CoC and recovery targets, and verify the local service partner for Antananarivo before signing.
What compliance risk should we price into the design before the ONE permit is issued?
The binding instruments are the ONE arrêté d'autorisation for site effluent quality, the MININEE framework for industrial water use and discharge, and the étude d'impact for hyperscale builds (S2, 2026). Because municipal sewerage in Antananarivo is intermittent and industrial pretreatment is not uniformly enforced downstream, design the train to the strictest plausible TDS, chloride and sulfate ceilings and confirm the actual values against the permit text before design freeze.
How do we keep membrane replacement on a 3–5 year interval instead of 12–18 months?
Run side-stream filtration at 10–25 micron to keep blowdown suspended solids below 15 mg/L, dose a bio-organic flocculant for colloidal aggregation, tune the antiscalant chemistry for silica and calcium sulfate, and install chemical-free UV disinfection ahead of the UF (S1, 2025-12; S2, 2026). Skipping any one of those four steps is the documented cause of tropical membrane-replacement intervals compressing from 3–5 years to 12–18 months (S2, 2026).
