Why Belo Horizonte Is a Distinct 2026 Wastewater Envelope
Belo Horizonte is not a generic Brazilian or generic Latin American site for fab and hyperscale design. The Velhas and Paraopeba basins that supply the metropolitan area are under allocation pressure from mining, steel, sanitation (COPASA), and urban demand, and any 2026 hyperscale freshwater withdrawal triggers formal review by the basin committees (CBH das Velhas, CBH do Paraopeba). The mining legacy of the Iron Quadrangle — active tailings management and historic drainage from the Rio das Velhas basin — establishes a baseline metal and TDS profile that any new discharge must be designed against, not a greenfield freshwater envelope. Globally, 45% of data centres sit in basins at high risk of water-availability disruption (TNFD, 2026-02), and the Velhas/Paraopeba system falls in that band.
COPAM 2026 review practice for water-stressed ICT projects mirrors the 80% reuse-floor pattern that has closed permit economics in comparable 2026 jurisdictions, including the Baku semiconductor and data hall 2026 guide analogue anchored on the Caspian/Maraza basin. Designers should treat that floor as a hard constraint from the first mass balance, not a permit-negotiated variable. AI rack densities in 2026 are pushing BH-bound hyperscale halls toward direct-to-chip cooling and closed-loop water, which raises the local reuse floor above the global hyperscale average (UltraFacility, 2026). The result is a design envelope where 80% recovery is the permit floor, 95% is the engineering ceiling, and basin allocation sets both.
CONAMA, COPAM, and the 2026 Effluent Envelope
CONAMA Resolução 357/2005 sets receiving-water quality classes and the framework for effluent standards in Brazil. CONAMA Resolução 430/2011 sets the conditions and standards for effluent discharge, and is the binding federal instrument for fab and hyperscale-hall effluents. Minas Gerais COPAM and NORMAM norms add state-level criteria and the licensing procedure, including the LO (Licença de Operação) and the RCA/PCA EIA package that hyperscale projects must submit before commissioning. Any 2026 BH fab or hall must demonstrate compliance against the CONAMA 430 discharge standards and the COPAM state addenda; the two are layered, not alternative, and the COPAM conditions are typically more restrictive on heavy metals, fluoride, and temperature ΔT.
The binding contaminants for a semiconductor envelope are F⁻ (HF-etch at 50–500 mg/L per Baku guide S2), Cu (Cu-CMP up to 100 mg/L per S2), TMAH in the developer stream, nitrate, total nitrogen, and trace heavy metals. Temperature ΔT and pH excursion control matter as much as the chemical columns because chiller trips on hyperscale sites cause 1–5 pH swings and 30–40 °C blowdown pulses (per Baku guide S2). For a hyperscale hall the dominant parameters are TSS, BOD/COD, oils, total phosphorus, pH, and ΔT; for a co-located fab the F, Cu, TMAH, and trace-metal columns dominate. Belo Horizonte discharge to the Paraopeba/Velhas system or to the COPASA sewage network is constrained at hyperscale flow because the receiving works are not sized for fab chemistry or for cooling-tower blowdown volumes in the 200–1,000 m³/day band.
| Parameter | CONAMA 430 typical VMP band | Hall CTBD influent (typical) | Co-located fab influent (typical) |
|---|---|---|---|
| pH | 5–9 (verify COPAM state addendum) | 6.5–8.5 | 1–5 excursions during HF events |
| TSS (mg/L) | ≤100 (verify COPAM) | 20–80 | 50–200 (CMP slurry) |
| F⁻ (mg/L) | ≤10 (verify COPAM) | <1 | 50–500 (HF-etch) |
| Cu (mg/L) | ≤1.0 (verify COPAM) | <0.1 | up to 100 (Cu-CMP) |
| TMAH (mg/L) | Verify COPAM | negligible | 10–100 (developer stream) |
| Temperature ΔT (°C) | <3 from receiving body | 30–40 blowdown | 25–35 |
| Total N (mg/L) | ≤20 (verify COPAM) | <5 | 20–60 (TMAH + nitrate) |
The right column is the design boundary the EPC has to negotiate down; the centre column is the permit ceiling after treatment.
Stand-Alone Hyperscale Hall vs Co-Located Fab: Two Different Design Problems
The envelope you design for depends entirely on whether a fab is co-located, and the two cases drive materially different trains and CAPEX bands. A stand-alone BH hyperscale hall sees cooling-tower blowdown at 4–6 cycles of concentration with TDS up to 2,000 ppm and 30–40 °C blowdown, plus AHU condensate below 50 mg/L TDS (per Baku guide S2). The blowdown formula B = E/(COC−1) converts evaporation rate E into blowdown volume: 25% blowdown at 4 COC, 20% at 6 COC, and that ratio drives the RO train size and the downstream brine volume. AI and direct-to-chip cooling push the hall side of the envelope toward higher COC and tighter water budgets in 2026, shrinking the blowdown fraction but raising its TDS (per UltraFacility, 2026).
A co-located fab in BH changes the problem qualitatively. HF-etch fluoride runs 50–500 mg/L, TMAH appears in the developer stream, CMP nanoparticles (silica, ceria, alumina) enter with the slurry, and Cu can reach 100 mg/L in untreated CMP effluent (per S2, Lai & Lin 2004). For that envelope, full-stream ZLD is the 2026 default because the combined HF + nanoparticle + TMAH load cannot be negotiated down to the CONAMA 430 receiving-water band. Fab and hall streams must be segregated from the first mass balance, with the AHU condensate on its own line and the fab process drains on a separate train — the design principles match the Vienna semiconductor and data hall 2026 guide framing but with a Brazilian regulatory overlay.
Reference Treatment Train for a BH Hyperscale Hall (2026)
The six-stage train below is what a 2026 BH tender will most likely evaluate for a stand-alone hyperscale hall, with parameters and equipment choices common to high-reuse ICT projects.
Stage 1 — Equalisation. A dedicated EQ tank with 4–8 h HRT and online pH/conductivity dampens the 1–5 pH excursions that follow chiller trips, which are routine on a hyperscale site. AHU condensate (typically <50 mg/L TDS) is segregated on its own line because glycol from coil leaks requires stripping, not blending into the main RO loop.
Stage 2 — DAF and multi-media filtration. A DAF unit in the 4–300 m³/h class floats oils, biofilm, and metal-hydroxide floc, followed by a multi-media filter that drops SDI below 3 and protects the RO from Cu, Fe, and Zn fouling.
Stage 3 — Softening and antiscalant dosing. A twin-tank softener (1–45 T/h class) targets hardness <50 mg/L as CaCO₃ and silica <10 mg/L as SiO₂, with PLC-controlled antiscalant dosing tied to RO feed flow to handle the residual scaling potential that softening cannot reach. On the Velhas/Paraopeba intake, the softener needs to track seasonal silica swings rather than hold a fixed setpoint.
Stage 4 — MBR polish (optional, stand-alone hall only). A submerged PVDF MBR polishing system with 0.1 µm membranes delivers <1 NTU and <10 mg/L COD, allowing direct RO feed without media filtration. Required only if sanitary load is co-mingled; fab streams replace this stage with a dedicated HF-removal and metals-precipitation step upstream of UF.
Stage 5 — Two-pass RO at 80–95% recovery. A two-pass industrial RO system delivers permeate to cooling-tower make-up spec. First pass runs at 150–250 psi (10–17 bar) for bulk salts; second pass polishes to TDS <200 mg/L and Cl⁻ <100 mg/L. Above 95% recovery, silica scaling on the second-pass membranes drives CIP frequency up sharply, and 80–95% is the practical operating window.
Stage 6 — MVC brine concentration. Mechanical vapor recompression at 25–40 kWh/m³ of brine concentrated. Reserve MVC for RO brine on a stand-alone hall; full-stream MVC is the non-negotiable baseline for any co-located fab. A side-stream filtration step at 1–5% of total circulation using 10–25 µm self-cleaning units drops suspended solids, and a ClO₂ generator upstream of the RO controls biofouling without the isothiazolinone residual load that shortens RO membrane life (per Baku guide S2). Use RO/UF membrane elements rated for high-silica feed to control replacement frequency.
When a BH Co-Located Fab Forces Full-Stream ZLD
The trigger condition for full-stream ZLD on a 2026 BH project is well defined: HF-etch fluoride, CMP nanoparticles, and TMAH co-occur in the wastewater envelope. The combined load cannot be negotiated down to the CONAMA 430 receiving-water band, and COPAM 2026 practice treats full-stream ZLD as the default for any co-located fab or fab-hall hybrid.
CMP and HF streams must be segregated upstream of the main hall train; an HF-removal and metals-precipitation step replaces the MBR polish and feeds the UF step. A hollow-fiber UF system upstream of DAF improves colloid removal on Cu-CMP and stabilises downstream RO performance (per Baku guide S2). Above 95% RO recovery, silica scaling drives CIP frequency up sharply; full-stream ZLD is the way to push past 95% without paying the OPEX penalty in membrane replacement. MVC electricity at 25–40 kWh/m³ of brine is the dominant OPEX line and is the line item most sensitive to the CEMIG tariff structure in Minas Gerais — none of the top-ranking 2026 ICT wastewater guides address this overlay, and it is the single biggest cost driver a Brazilian EPC will see on the OPEX line. EV-1: Full-stream ZLD is over-specified for a stand-alone hall; reserve MVC for RO brine only. EV-2: For any co-located fab or fab-hall hybrid, full-stream ZLD is the 2026 default.
CAPEX, OPEX, and Payback for a 2026 BH Hyperscale Hall
CAPEX bands for a 2026 BH hyperscale hall (engineering estimates, HydropureWater field data, 2026): a small hall under 200 m³/day with package plant plus haul-off runs $150–300/m³/day installed; mid-size 200–1,000 m³/day MBR + RO runs $400–700/m³/day installed, which is where the typical 5–20 MW BH hall sits; hyperscale or co-located fab with brine ZLD runs $800–1,200/m³/day installed (per Baku guide S2). Use the high-silica-rated RO elements to hold the OPEX curve flat across the recovery window.
OPEX lines that push cost up: membrane CIP frequency above 95% recovery on the second pass due to silica scaling; isothiazolinone biocide residuals shortening RO life, which is the case for a ClO₂ side-loop upstream of the RO rather than dosing in the tower; MVC electricity at 25–40 kWh/m³ of brine concentrated. The CEMIG regulated-tariff structure in Minas Gerais interacts directly with the MVC energy demand, and energy cost is a bigger OPEX driver than in jurisdictions with cheaper grid power; the project-specific tariff band should be requested from CEMIG at the kickoff meeting, not at commissioning. Avoided-discharge math at $5–15/kgal: 100 m³/day of untreated blowdown at the upper end is ~USD 400/day, so an 80% recovery RO pays back inside ~24 months at hyperscale flow (HydropureWater field data, 2026).
| Plant size (m³/day) | Configuration | CAPEX (USD/m³/day installed) | Dominant OPEX lines |
|---|---|---|---|
| <200 | Package + haul-off, no RO | $150–300 | Brine haul-off, softener regeneration |
| 200–1,000 | MBR + two-pass RO, 80–95% recovery | $400–700 | RO CIP, antiscalant, ClO₂ residual, membrane replacement |
| 200–1,000 (fab or hybrid) | Full-stream ZLD with MVC | $800–1,200 | MVC electricity 25–40 kWh/m³, high-silica RO replacement, CEMIG-tariff sensitivity |
Designers should size the RO train against the actual Velhas/Paraopeba intake silica profile rather than a generic 50 mg/L assumption, and validate the 80–95% recovery window against the RO design specifications 2026 reference before final equipment selection.
Frequently Asked Questions
What regulatory framework applies to a 2026 semiconductor or hyperscale data-hall wastewater discharge in Belo Horizonte?
CONAMA Resolução 357/2005 sets receiving-water quality classes, and CONAMA Resolução 430/2011 sets the binding effluent discharge standards for fab and hall effluents. Minas Gerais COPAM and NORMAM norms add state-level criteria and the LO (Licença de Operação) procedure, including the RCA/PCA EIA package for hyperscale projects. The COPAM state addenda are typically more restrictive on F, Cu, trace metals, and ΔT than the federal floor, and the EIA review for hyperscale projects runs on a multi-month clock that should start at kickoff, not at submission.
What internal water-reuse percentage should a 2026 BH fab or hyperscale hall target?
80% is the practical floor that closes the avoided-discharge payback inside ~24 months at hyperscale flow and that matches the basin-committee review pattern in comparable 2026 water-stressed ICT jurisdictions. 95% is the engineering ceiling; above 95% recovery on the second pass, silica scaling drives CIP frequency up sharply and erodes the OPEX gain. The 80–95% window is the design range; the 80% line is non-negotiable for permit economics, the 95% line is non-negotiable for OPEX discipline.
When is full-stream ZLD non-negotiable for a 2026 BH semiconductor project?
Whenever HF-etch fluoride (50–500 mg/L), CMP nanoparticles, and TMAH co-occur in the wastewater envelope. The combined load cannot be negotiated down to the CONAMA 430 receiving-water band, and COPAM 2026 practice treats full-stream ZLD as the default for any co-located fab or fab-hall hybrid. For a stand-alone hall without fab chemistry, full-stream ZLD is over-specified and MVC on the RO brine is the right envelope.
What CAPEX should a 200–1,000 m³/day hyperscale hall in BH budget for in 2026?
A mid-size 200–1,000 m³/day MBR + RO train runs $400–700/m³/day installed, which is where the typical 5–20 MW BH hyperscale hall sits. A small hall under 200 m³/day with package plant plus haul-off runs $150–300/m³/day installed; a hyperscale or co-located fab with brine ZLD runs $800–1,200/m³/day installed (HydropureWater field data, 2026). These are engineering estimates, not firm quotes, and they should be validated against site-specific influent characterisation and the final equipment proposal.
Can a BH hyperscale hall discharge its treated effluent to the municipal COPASA sewer?
Not at hyperscale flow. COPASA's receiving works in the BH metro are not sized for fab chemistry or for cooling-tower blowdown volumes in the 200–1,000 m³/day band, and COPAM 2026 review practice treats municipal-sewer discharge as a non-starter for ICT projects above the small-hall threshold. The two practical discharge paths are the Velhas/Paraopeba receiving water with treatment to the CONAMA 430 + COPAM band, or full-stream ZLD with brine concentration and solids handling on site.
