Why Mombasa is a Hydrology-First Site, Not a Datasheet-First Site
A 2026 process-wastewater design for a semiconductor or data-hall facility in Mombasa starts with three documents, not a vendor cut-sheet: a National Environment Management Authority (NEMA) Environmental Impact Assessment scoping report, a Water Resources Authority (WRA) water permit, and a Coast Water Works Development Agency (CWWDA) allocation letter. The TNFD February 2026 case study on water dependency in the technology sector makes the same point at the global level: treatment selection follows the dependency pathway, not the reverse (TNFD Feb 2026). In Mombasa the dependency pathway is administered by NEMA and the WRA, and the allocation pathway is administered by CWWDA. Engineers who size a clarifier, an ultrafiltration system, or an industrial RO system before those documents are in hand will tender the wrong plant.
The supplied research contains no NEMA effluent limits and no WRA numeric discharge conditions for a fab or data-hall site. The engineer must request the current limit values from NEMA at EIA scoping and from the WRA at the permit application stage rather than carry a generic specification into the design. The risk is asymmetric: a generic specification tends to over-design on safety and under-design on whatever the local regulator will reject at commissioning, because Mombasa's coastal hydrology and tidal outfall constraints are not the same envelope a temperate-inland datasheet assumes.
Mombasa's coastal hydrology, intermittent municipal supply, and tidal outfall put a Kenya coastal site in a comparable water-stress risk profile to the global figure that 45% of data centres sit in basins at high risk of water-availability disruption (TNFD Feb 2026, citing Hajonides et al. 2025). The comparison is a prompt, not a proxy: the design must be evaluated on its own hydrology study, not on a global average. The TNFD Feb 2026 case study also notes that 40% of existing fabs and over 40% of new fabs announced since 2021 are projected to be in basins with high or extremely high water-stress risk by 2030 (TNFD Feb 2026, citing Lepawsky 2024), which is the same logic that makes Mombasa a hydrology-first site.
The practical order of work is therefore: stream segregation first, treatment selection second, intake allocation third. This is the order the TNFD Feb 2026 case study implies for fab-side pollution, and it is the order a NEMA reviewer and a WRA officer will expect in 2026. A more detailed walk-through of the segregation logic is in our UPW scale-up guide, and the ZLD framing that follows from this intake-constrained logic is covered in the ZLD adoption outlook for 2026.
Global Water Anchors Translated to a Mombasa Envelope
The TNFD February 2026 case study gives an EPC four numeric anchors a Mombasa-bound engineer can cite at a permit meeting, all attributable inline to that source. Worldwide semiconductor water use doubled between 2012 and 2022 (TNFD Feb 2026, citing Marcello 2024). A single fab benchmark sits at around 14 billion litres of UPW per year (TNFD Feb 2026, citing WEF 2025). For every unit of UPW, 1.4–1.6 units of municipal water are used (TNFD Feb 2026, citing IDE Technologies 2024). On the data-centre side, a typical facility draws 25 million to 770 million litres per year depending on size, hyperscale facilities may exceed 2 billion litres per year, and the U.S. national data-centre draw was around 66 billion litres in 2023 (TNFD Feb 2026, citing Ceres 2025, Hines Research 2025, USGS 2015).
Worked Mombasa envelope (upper bound, not a target): a 5,000 m³/day UPW envelope at the 1.4–1.6× municipal-to-UPW ratio implies 7,000–8,000 m³/day of municipal input. That is the figure the CWWDA allocation letter must be able to support before any equalisation tank is sized. Samsung's reported 344,000 tonnes of water per day, equivalent to 344 million litres per day, is included only as a global benchmark to calibrate the Mombasa envelope downward (TNFD Feb 2026, citing Samsung Water Stewardship); the supplied research contains no Kenya-specific cost figure, so the engineer must request a budget from the EPC sized to the CWWDA allocation, not to a generic vendor list price.
A Tier-III data hall in a hot, humid coastal climate sits at the upper end of the 25 million–770 million litres/year range, because once-through and evaporative cooling both perform more poorly above 28 °C wet-bulb. For a 1,000 m³/day Tier-III envelope, the cooling-tower blowdown stream is typically the dominant wastewater volume on the data-hall side of the train. Cooling systems account for most on-site operational water withdrawals in data centres (TNFD Feb 2026, citing Ceres 2025), so blowdown handling, not chemical complexity, is the design driver on the data-hall side.
| Anchor | Value | Source | Use in Mombasa envelope |
|---|---|---|---|
| Global semiconductor water use, 2012–2022 | Doubled | TNFD Feb 2026, citing Marcello 2024 | Growth-pressure context, not a sizing input |
| Single-fab UPW benchmark | ~14 billion litres/year | TNFD Feb 2026, citing WEF 2025 | Calibrates Mombasa 5,000 m³/day envelope downward |
| Municipal-to-UPW ratio | 1.4–1.6× | TNFD Feb 2026, citing IDE Technologies 2024 | Drives the CWWDA allocation request |
| Typical data centre annual draw | 25 million–770 million litres/year | TNFD Feb 2026, citing Ceres 2025 | Tier-III Mombasa hall sits toward upper end |
| Hyperscale data centre annual draw | May exceed 2 billion litres/year | TNFD Feb 2026, citing Hines Research 2025 | Not relevant unless Mombasa hosts hyperscale |
| U.S. national data-centre draw, 2023 | ~66 billion litres | TNFD Feb 2026, citing USGS 2015 | Macro reference only |
| Samsung reported daily water use | 344,000 tonnes/day (344 million litres/day) | TNFD Feb 2026, citing Samsung Water Stewardship | Global benchmark, calibrate envelope downward |
Stream Segregation: The First Design Decision
A "one wastewater plant" mental model is the wrong starting point for a Mombasa fab-or-data-hall campus. Each stream carries a different chemistry, a different NEMA effluent-class implication, and a different reuse potential. The first design decision is segregation; treatment selection follows. The same logic is implicit in the TNFD Feb 2026 case study, which treats fab-side pollution as a stream-by-stream issue when assessing dependency risk (TNFD Feb 2026).
On the fab side, four streams must be segregated at source: UPW reject (high-purity concentrate from the polishing loop, suitable for RO recovery), acid/alkaline cleaning rinses from wet-etch and post-ash cleans (high TDS, pH excursions), chemical-mechanical planarisation (CMP) slurry wastewater (high in suspended solids, colloidal silica, and metals such as copper or tungsten depending on the layer), and solvent-bearing rinses from photolithography. Co-mingling any of these with the others complicates both NEMA discharge compliance and any future water-reuse scheme, which is why segregation is the design step a permit reviewer will look for first. A regional comparison of this segregation logic against a different coastal regulator is in our West African regional guide.
Data-hall streams are chemically simpler: evaporative cooling tower blowdown (high in TDS, scale inhibitors, and biocides), humidification bleed-off (essentially pre-evaporated potable water with low TDS), and once-through cooling discharge where still permitted by CWWDA and any local sewer ordinance. Cooling systems account for most on-site operational water withdrawals in data centres (TNFD Feb 2026, citing Ceres 2025), which means the data-hall side of the wastewater train is dominated by blowdown volume, not chemical complexity.
| Stream | Source | Key characteristics | Reuse potential | Compliance risk if co-mingled |
|---|---|---|---|---|
| UPW reject | Polishing-loop concentrate | High purity, low TDS concentrate | RO recovery to UPW feed or cooling make-up | Loses recovery economics |
| Acid/alkaline rinses | Wet-etch, post-ash cleans | High TDS, pH excursions | Neutralise, then polish | NEMA pH excursions, sludge handling |
| CMP slurry wastewater | Planarisation step | Suspended solids, colloidal silica, Cu or W | Clarify, recover metals where feasible | Metal-loading breach at discharge |
| Solvent rinses | Photolithography | Organic load, low TDS | Segregate, separate destruction | Organic-loading breach at discharge |
| Cooling-tower blowdown | Evaporative cooling | High TDS, scale inhibitors, biocides | Limited; side-stream filtration | Biocide and TDS breach at discharge |
| Humidification bleed-off | Humidification system | Pre-evaporated potable, low TDS | Cooling make-up | Minor; mostly hydraulic load |
| Once-through cooling | Heat exchangers | Thermal load, low chemical load | Heat-recovery or discharge | Thermal-plume risk at tidal outfall |
Default 2026 Treatment Train for a Mombasa Site
The defensible default train for a 2026 Mombasa process-wastewater system runs segregated equalisation → DAF or lamella clarifier on the fab-side suspended-solids load → UF as RO pretreatment → RO on UPW reject for reuse → PLC-controlled chemical dosing for pH, biocide, and scale control → discharge or reuse to follow NEMA and WRA conditions. This order is consistent with the TNFD Feb 2026 case study's stream-by-stream treatment of fab-side pollution, and the 1.4–1.6 municipal-to-UPW ratio (TNFD Feb 2026, citing IDE Technologies 2024) makes UPW-reject RO recovery a structural part of any 2026 design, not an optional add-on.
The unit operations map to the segregated streams as follows. A DAF system handles the CMP slurry and the suspended-solids load before downstream polishing. An ultrafiltration system sits ahead of the RO as pretreatment on the fab-side train, removing colloidal and particulate carry-over that would otherwise foul the RO membranes. An industrial RO system is then applied to the UPW-reject stream for reuse, and an automatic chemical dosing system handles pH neutralisation on the acid/alkaline rinses and biocide or scale-inhibitor control on the cooling-tower blowdown side.
Numeric effluent limits under NEMA and any WRA discharge condition are not in the supplied research and must be requested from NEMA at EIA scoping before final sizing. The defaulting risk is concrete: a design carried in on a "standard" specification will over-size on safety, or worse, miss a limit the regulator will reject at commissioning, which is more expensive than the cost of an early scoping letter. The equipment classes listed above map to steps 1–4 of the train; step 5, the discharge or reuse routing, follows the NEMA/WRA conditions once they are in writing.
Five Procurement Checks Before Tender
Before the wastewater plant is tendered, the project team should run five procurement checks. None of these are treatment-train decisions; they all sit upstream of treatment selection, which is why a datasheet-first approach gets re-scoped after the fact.
- Confirm the CWWDA allocation letter can support the 1.4–1.6× municipal-to-UPW ratio before any equalisation tank is sized. The allocation letter is a hard pre-condition, not a parallel workstream. A 5,000 m³/day UPW envelope implies 7,000–8,000 m³/day of municipal input, which CWWDA must confirm in writing before tender.
- Request NEMA effluent limits and any WRA discharge conditions at EIA scoping. The supplied research contains no numeric values, so the engineer must obtain them directly from NEMA and the WRA rather than carry generic numbers into the design. This is the single largest source of re-scope risk on a 2026 Kenya project.
- Size the budget request to the allocation and the limits, not to a generic vendor list price. The supplied research contains no Kenya-specific cost figure, so the engineer should request a budget estimate from the EPC sized to the CWWDA allocation and the NEMA/WRA conditions, not a packaged list price.
- Confirm the longest lead-time items at RFQ stage. RO skids and dosing skids are typically the long lead items on a 2026 Kenya project, so the engineer should request current fab-duty UF and RO delivery terms from the OEM at RFQ stage before tender. A 12–16 week delivery on a dosing skid is enough to push a Mombasa commissioning date on its own.
- Decide reuse vs. discharge early. The RO-on-UPW-reject step only pays back if a reuse end-use is sized into the campus (cooling make-up, scrubber make-up, toilet flushing). Without a reuse end-use, RO concentrate handling becomes the design driver and the 1.4–1.6× ratio advantage is lost.
Frequently Asked Questions
What is a defensible 2026 budget envelope for a process-wastewater train on a Mombasa fab or Tier-III data hall?
The supplied research contains no Kenya-specific cost figure, so a defensible 2026 budget envelope must be requested from the EPC sized to the CWWDA allocation letter and the NEMA effluent limits, not a vendor list price. The engineer should ask the EPC for a budget range that names the equipment classes (segregated equalisation, a DAF system on the fab side, an ultrafiltration system, an industrial RO system on UPW reject, and an automatic chemical dosing system) tied to the CWWDA allocation flow and the NEMA discharge conditions. A request that does not name those documents will return a generic figure, which is not what a permit-bound project needs.
Which equipment classes carry the longest lead time for a 2026 Mombasa project?
RO skids and dosing skids are typically the long lead items on a 2026 Kenya project, so the engineer should request current fab-duty UF and RO delivery terms from the OEM at RFQ stage before tender. The supplied research does not contain a Kenya-specific lead-time figure, so the check is to obtain a written delivery term from the OEM in weeks, not in months, and to put that term on the EPC's critical-path schedule.
Should the Mombasa design treat UPW reject as a discharge stream or a reuse stream?
Treat UPW reject as a reuse stream, not a discharge stream, in 2026. The 1.4–1.6 municipal-to-UPW ratio (TNFD Feb 2026, citing IDE Technologies 2024) makes RO recovery on the UPW-reject stream a structural part of the design rather than an option. A reuse end-use must be sized into the campus (cooling make-up, scrubber make-up, toilet flushing) so the RO concentrate handling does not become the design driver. If no reuse end-use exists, the design needs to be re-asked before tender, not after.
What is the minimum documentation the Mombasa engineer must obtain before sizing the train?
The minimum documentation is a CWWDA allocation letter sized to the 1.4–1.6× municipal-to-UPW ratio, NEMA effluent limits requested at EIA scoping, and any WRA discharge conditions. None of these numbers are in the supplied research, so the engineer must request them directly from NEMA, the WRA, and CWWDA before equalisation, UF, or RO sizing begins. A design carried in on a generic specification is the most common source of re-scope risk on a 2026 Mombasa process-wastewater project.
