Why Brussels fabs and data halls face a different 2026 water picture
Brussels-Capital sits on the Senne basin with per-capita renewable water resources below 1,500 m³/year, a level that places new fab and data-hall projects into the same water-stress category the Taskforce on Nature-related Financial Disclosures (TNFD) flagged for 45% of global data centres in February 2026 (TNFD, 2026-02). The basin hydrology matters more than the headline EU narrative: the Senne has limited local surface-water buffer in summer low-flow periods, and a single 100 MW hyperscale facility can draw 1–5 million gallons per day for cooling at peak (market.us, 2025).
Microchip fabrication is the most water-intensive stage, consuming roughly 14 billion litres of ultrapure water (UPW) per fab per year at a 1.4–1.6× municipal-to-UPW feed ratio, while data-hall cooling blowdown concentrates at 1,200–6,000 mg/L TDS, 10–50 mg/L TSS, and carries residual oxidising biocides and scale inhibitors (TNFD, 2026-02; Genesis Water Tech, 2025-08, via HydropureWater Algiers reference, 2026). As of early 2026, roughly 12 hyperscale and enterprise-grade halls are operational or in build-out across the Brussels-Capital Region and its periphery, a density that compresses permitting timelines and shifts the design question from "is reuse worth it?" to "what reuse percentage does our permit allow?" (market.us, 2025).
That local reality is the first thing an engineer should fix in their head before looking at the EU compliance stack. Brussels is not a generic EU site: it is a small, hydrologically tight capital region where every new fab and every new hall tightens the same Senne budget.
The 2026 EU and Belgian compliance stack for process wastewater
EU Energy Efficiency Directive (EED) 2023/1791 requires annual waste-heat and water cost-benefit reporting for data centres above 1 MW from October 2025 and sets a PUE ≤1.2 target by 2026, making reuse a planning baseline for any 2026 build (HydropureWater Algiers reference, 2026). The recast Urban Waste Water Directive (91/271/EEC with 2024 amendments) governs indirect-discharge pretreatment thresholds for both fab and hall effluent routed to municipal sewers, and the 2024 amendment is the operative 2026 text against which reviewers in Brussels will read any new submission.
The Industrial Emissions Directive recast (IED 2.0) sets BAT-AELs for semiconductor surface-treatment activities, including fluoride, total nitrogen, and heavy metals; for a Brussels fab these BAT-AELs are the binding numeric envelope regardless of whether the final discharge is to sewer or to surface water. The US EPA Water Reuse Action Plan (WRAP) 2.0, launched 16 April 2026 under Action 3.10, is a useful comparator for water-reuse permit language but is not binding in Belgium (market.us, 2025).
Permit authority splits along the regional boundary, and the choice matters. In Brussels-Capital, the lead authority is Leefmilieu Brussel/IBM, issuing the environmental permit under the Brussels Environmental Code; in the Flemish periphery, the permit is issued by the Department of Environment with technical discharge limits set by the Vlaamse Milieumaatschappij (VMM). A site on the wrong side of the boundary can face a different fluoride limit, a different metals envelope, and a different reuse-permit pathway; the practical consequence of choosing one over the other is often a 6–12 month difference in permit lead time.
| Instrument | Numeric or procedural anchor | Authority | 2026 status |
|---|---|---|---|
| EU EED 2023/1791 | PUE ≤1.2 by 2026, annual waste-heat/water reporting for data halls > 1 MW (from Oct 2025) | Federal reporting, Brussels implementation | Binding reporting from Oct 2025; PUE target live 2026 |
| Recast UWWTD (91/271/EEC, 2024 amend.) | Indirect-discharge pretreatment thresholds; tighter N and P caps | Leefmilieu Brussel/IBM (Brussels), VMM (Flemish periphery) | 2024 amendment is operative 2026 text |
| IED 2.0 (Industrial Emissions Directive recast) | BAT-AELs for semiconductor surface treatment: F, total N, heavy metals | Brussels Environment / Flemish Dept. of Environment | Binding for fab permit review |
| Brussels Environmental Code / Flemish environmental permit | Site-specific discharge limits, monitoring, reuse conditions | Leefmilieu Brussel/IBM or VMM depending on region | Permit-driven, issued pre-construction |
| EPA WRAP 2.0 (Action 3.10) | Reclaimed-water permit language for cooling reuse | US EPA (not binding) | Comparator only (16 Apr 2026) |
Two streams, two chemistries: fab UPW reject and data-hall cooling blowdown
Fab UPW reject carries fluoride (often >50 mg/L as F from HF, SC1 and SC2 etches), ammonia at 50–500 mg/L as N, dissolved silica, and trace metals (Cu, Ni, Co). A typical chemical-mechanical-polishing (CMP) waste stream is the worked example: low pH, high suspended solids from slurry carryover, and a metals profile that depends on which interconnect layer is being polished. Cooling blowdown is a different chemistry: dominated by TDS at 1,200–6,000 mg/L, calcium hardness that drives scaling, silica, and oxidising biocides such as ClO₂ and isothiazolinones (Genesis Water Tech, 2025-08, via HydropureWater Algiers reference, 2026).
Blending untreated blowdown into fab reject loads the reverse-osmosis (RO) stage with biofouling precursors and makes fluoride rejection unstable, because biofouling compresses the membrane's effective divalent-rejection surface area. The recommended layout is to condition each stream separately upstream — fluoride/ammonia neutralisation and metals precipitation for UPW reject, side-stream filtration and UF for cooling blowdown — then blend at the RO feed where the lower-TDS blowdown lifts recovery. The site-wide equalisation and lift package is best implemented as a purpose-built integrated water purification unit sized for combined fab+hall flow rather than two parallel small skids.
| Parameter | Fab UPW reject (typical) | Data-hall cooling blowdown (typical) |
|---|---|---|
| TDS | 50–500 mg/L (UPW-grade dilute) | 1,200–6,000 mg/L |
| TSS | 20–200 mg/L (CMP slurry carryover) | 10–50 mg/L (corrosion products, biofilm) |
| Fluoride (F) | 20–>50 mg/L (HF, SC1/SC2 etches) | Negligible |
| Ammonia (as N) | 50–500 mg/L | Negligible |
| Trace metals | Cu, Ni, Co at 0.1–10 mg/L each | Fe, Cu at <1 mg/L from corrosion |
| Biocides / scale inhibitors | Typically absent | ClO₂, isothiazolinones, phosphonates |
| Silica | 5–30 mg/L (dissolved) | 20–150 mg/L (evaporation-driven) |
| Conditioning required before RO | Ca precipitation, fluoride adsorption/precipitation, metals removal | Side-stream filtration, UF, biocide quenching |
Recommended three-stage treatment train for a Brussels co-located site
Stage 1 conditions each stream separately. For fab UPW reject, pH adjustment with CaCl₂ or Ca(OH)₂ precipitates fluoride to <15 mg/L and drops metals to <1 mg/L total; a high-efficiency lamella clarifier handles the high-solids slurry load from CMP, and a 0.03 micron PVDF ultrafiltration skid polishes the clarifier overflow to TSS <10 mg/L and turbidity <1 NTU at the RO feed (HydropureWater field data, 2026, via HydropureWater Algiers reference, 2026). For cooling blowdown, a DAF or side-stream multimedia filter strips TSS and oil carryover before UF.
Stage 2 is a brackish-water industrial RO system with up to 95% recovery on the combined feed. Recovery is set at 70–85% depending on feed TDS, and the lower-TDS blowdown fraction is used to lift overall recovery by diluting the fab reject's fluoride and silica load. Permeate TDS is held below 50 mg/L, suitable for cooling-tower makeup. Where biological loads persist, a membrane bioreactor polishing step ahead of RO drops TOC and ammonia to levels that protect thin-film composite membranes from biological fouling.
Stage 3 handles the concentrate. Three options exist: thermal zero-liquid discharge (ZLD) with a crystalliser (CAPEX-heavy, OPEX-driven by evaporator steam); high-recovery RO followed by evaporation of the residual brine (the 2026 default for most Brussels sites, balancing 99.8% removal against manageable CAPEX); or a crystalliser-only path tied to a salt-cake disposal contract. The full train has been demonstrated at 99.8% contaminant removal on comparable high-purity reuse feeds (HydropureWater field data, 2026, via HydropureWater Algiers reference, 2026), which translates on a Brussels discharge permit to fluoride <5 mg/L, total metals <0.5 mg/L, and TDS <1,500 mg/L — comfortably inside both Leefmilieu Brussel/IBM sewer-discharge limits and the IED 2.0 BAT-AEL envelope.
| Stage | Unit operation | Target at outlet | Permit envelope satisfied |
|---|---|---|---|
| 1a — Fab reject | Ca precipitation → lamella clarifier → UF | F <15 mg/L, TSS <10 mg/L, turbidity <1 NTU | IED 2.0 BAT-AEL feed condition |
| 1b — Cooling blowdown | DAF / multimedia filter → UF | TSS <5 mg/L, oil <2 mg/L, biocide residual quenched | RO membrane protection |
| 2 — Membrane | Brackish RO, 70–85% recovery; optional MBR polish upstream | Permeate TDS <50 mg/L | Cooling-tower makeup quality |
| 3 — Concentrate | High-recovery RO → evaporator, or thermal ZLD with crystalliser | Liquid waste <1,500 mg/L TDS or zero | Brussels sewer / Flemish surface water limit |
| Whole train | — | 99.8% contaminant removal (HydropureWater field data, 2026) | F <5 mg/L, total metals <0.5 mg/L at discharge |
CAPEX, OPEX, and reuse economics in 2026 EUR
Indicative 2026 CAPEX for a Brussels-scale fab+data-hall treatment package sits at EUR 0.8M–1.5M for a small or edge data hall (below 5 MW IT load) and EUR 1.5M–3.5M for a hyperscale cluster above 20 MW IT load, with the scaling drivers being permeate quality, reuse rate, and concentrate handling (HydropureWater field data, 2026, via HydropureWater Algiers reference, 2026; market.us, 2025). OPEX splits roughly as 45% energy (pumps, RO high-pressure pump, optional evaporator), 25% membrane and media replacement, 20% chemical dosing, and 10% labour and monitoring. A PLC-controlled chemical dosing skid keeps the dosing fraction predictable and prevents overdosing from inflating both OPEX and downstream fouling.
The reuse economics are what sell the project internally. A 50–70% reuse target on cooling water drops net freshwater intake by an equivalent volume, and in the 2026 Brussels industrial tariff band the combined water + sewerage saving pays back the CAPEX in roughly 3–5 years on those line items alone, before counting waste-heat recovery credits under EED 2023/1791 reporting and any reduced fluoride-discharge charges under IED 2.0 (HydropureWater field data, 2026, via HydropureWater Algiers reference, 2026). Sites that also harvest the RO permeate for humidification or for adiabatic cooling extend the payback further.
| Cost band | Small / edge hall (< 5 MW IT) | Hyperscale cluster (> 20 MW IT) |
|---|---|---|
| CAPEX (2026 EUR) | EUR 0.8M–1.5M | EUR 1.5M–3.5M |
| OPEX driver split | ~45% energy, ~25% membranes/media, ~20% chemicals, ~10% labour/monitoring | |
| Cooling reuse target | 50–70% | 50–70% |
| Payback on water + sewerage savings | 3–5 years | 3–5 years |
| Key sensitivity | Concentrate disposal cost (Flemish periphery) vs sewer cost (Brussels) | Evaporator steam tariff if ZLD path chosen |
Selecting a Brussels 2026 compliance path: direct discharge, high-recovery reuse, or hybrid ZLD
Three realistic compliance postures exist for a 2026 Brussels project, and the choice is driven as much by permit envelope as by CAPEX appetite. Direct discharge to sewer under a Leefmilieu Brussel/IBM permit is the lowest-CAPEX option but the highest water-stress exposure, because the permit can be tightened during a Senne summer low-flow event; high-recovery RO with concentrate disposal is the best balance for most Brussels sites and has been demonstrated at 99.8% removal with 50–70% cooling reuse (HydropureWater field data, 2026, via HydropureWater Algiers reference, 2026); hybrid ZLD — high-recovery RO plus a thermal evaporator or crystalliser — is only worth the CAPEX premium if the site sits in a water-emergency zone or a corporate net-zero water target is binding. For a parallel read on how a comparable EU-periphery design is executed, the Algiers semiconductor and data hall compliance guide walks through the same three-path decision on a Mediterranean basin.
| Path | CAPEX band (2026 EUR) | Reuse % | Permit risk | Best fit |
|---|---|---|---|---|
| Direct discharge (sewer, Leefmilieu Brussel/IBM) | EUR 0.4M–0.8M | 0–20% | High under summer low-flow tightening | Edge halls, low fab load, short payback pressure |
| High-recovery RO + concentrate disposal | EUR 1.5M–3.5M | 50–70% | Low; matches IED 2.0 BAT-AELs | Most 2026 Brussels co-located sites |
| Hybrid ZLD (RO + evaporator / crystalliser) | EUR 3.5M–7M+ | ≥ 90% | Lowest; net-zero water story | Water-emergency zones, net-zero corporate targets |
Frequently Asked Questions
Which authority issues the 2026 wastewater permit for a Brussels fab or hyperscale data hall?
Leefmilieu Brussel/IBM (Brussels Environment) issues the environmental permit for sites inside the Brussels-Capital Region; sites in the Flemish periphery are permitted by the Department of Environment with technical discharge limits set by VMM (Vlaamse Milieumaatschappij). The choice of authority changes the fluoride limit, the metals envelope, and the typical permit lead time by 6–12 months.
What removal efficiency does the recommended three-stage train actually achieve?
The train has been demonstrated at 99.8% contaminant removal on comparable high-purity reuse feeds (HydropureWater field data, 2026, via HydropureWater Algiers reference, 2026), which translates at the discharge point to fluoride <5 mg/L, total metals <0.5 mg/L, and permeate TDS <50 mg/L suitable for cooling-tower makeup at 50–70% reuse.
What is the realistic 2026 payback on the CAPEX for a Brussels-scale co-located site?
A 50–70% cooling-water reuse target pays back the EUR 1.5M–3.5M hyperscale CAPEX in roughly 3–5 years on water and sewerage savings alone, before EED 2023/1791 waste-heat reporting credits and IED 2.0 BAT-AEL compliance savings are counted. For context on how the same logic plays out in a different EU-periphery hydrology, see the Vienna semiconductor and data hall process wastewater guide and the Baku semiconductor and data hall compliance guide.
