Why 2026 Is the Inflection Point for Sydney Data Center Water
ABC reported in August 2025 that AI workloads are projected to consume one-quarter of Sydney Water's supply within a decade, a structural shift that puts every Western Sydney and Alexandria campus in direct competition with residential and industrial users for the same pipeline. In July 2026, OpenAI dropped its Sydney blowdown-recycling plan, confirming that piped Class A recycled supply cannot be treated as a guaranteed input for hyperscale builds (Reuters, 2026-07-22; Data Center Dynamics, 2026-07). The WSAA December 2025 report on data centres and water now serves as the industry reference document, calling explicitly for higher cycles of concentration, blowdown recycling, and electrolysis-based cooling water cleaning. The 2026 designer's job is to convert a contested makeup supply into a closed-loop cooling system that does not depend on Sydney Water's pipeline schedule, and the three benchmarks any spec will be measured against are AirTrunk SYD1/SYD2/SYD3, the Goodman Western Sydney cluster, and the OpenAI site. AirTrunk's FY23 sustainability report disclosed 58% recycled water across its portfolio, but the Goodman Western Sydney recycled-water infrastructure is still largely unbuilt per WSAA 2025.
What Cooling Tower Blowdown Actually Looks Like in a Sydney System
Sydney data centre blowdown at 4-8 cycles of concentration carries TDS of 1,200-6,000 mg/L, typically 4-8× the makeup water TDS, with suspended solids of 10-50 mg/L even when basin filtration is operating correctly (Genesis Water Tech, 2026). The suspended fraction is dominated by biofilm fragments, corrosion products, and Sydney's characteristic airborne particulate from construction activity around Western Sydney greenfield sites. Scale-forming ions (calcium, magnesium, silica, and alkalinity) concentrate proportionally to COC, and they set the antiscalant selection and the recovery ceiling on any downstream RO. The dissolved phase also carries treatment chemical residues: oxidising and non-oxidising biocides, molybdate and azole corrosion inhibitors, and phosphonate or polyacrylate scale inhibitors, plus the by-products of their breakdown (Permachar, 2026-08). When reclaimed water is blended into makeup under Australian water-reuse conditions, TDS rises further, and the inhibitor package must be retuned with silica-specific dispersants and ammonia-tolerant corrosion control. Sanitary wastewater is a minor side stream in this design space; the train can be optimised for the single blowdown stream rather than a mixed municipal influent.
| Parameter | Typical range at COC 4-6 | Design implication |
|---|---|---|
| TDS | 1,200-6,000 mg/L | Sets RO osmotic pressure and recovery ceiling |
| Suspended solids | 10-50 mg/L | Requires UF pretreatment before RO |
| Silica (SiO₂) | 40-200 mg/L | Antiscalant must be silica-rated |
| Calcium hardness | 400-1,800 mg/L as CaCO₃ | Lime/scale risk; NF alternative if hardness-limited |
| Residual biocides | 0.5-10 mg/L active | Carbon stage or biocide-neutralising step before discharge |
| Phosphonate scale inhibitor | 3-15 mg/L | Limits RO recovery; can trigger P permit |
The Highest-Leverage Decision: Cycles of Concentration
COC is the single highest-leverage design variable because it determines both blowdown volume and blowdown chemistry simultaneously. At COC 4, 25-30% of makeup water exits as blowdown; pushing to COC 6 cuts that to roughly 12-15% (Genesis Water Tech, 2026). A 10-million-gallon-per-month facility losing 2.5-3 million gallons at COC 4 drops to 1.3-1.7 million gallons at COC 6, saving ~1 million gallons/month of makeup before any treatment-train investment. The WSAA December 2025 strategy report names higher COC and electrolysis-based cooling water cleaning as the two key levers for the Australian sector, signalling that planners and Sydney Water reviewers will increasingly look for COC as a design KPI. The trade-off is mechanical: higher COC drives higher blowdown TDS, harder RO feed, and more aggressive antiscalant demand, so the membrane system must be sized to peak COC, not annual average. The Open Engineering LCA found UF+RO raises reuse energy by more than 5× freshwater, but COC gains offset the blowdown volume and chemical footprint, leaving the system-level case for reuse intact for a system commissioned today and operated through 2050.
| COC | Blowdown % of makeup | Approx. blowdown TDS (mg/L) | RO feed pressure band | Net water saving vs COC 4 |
|---|---|---|---|---|
| 3 | 33% | 900-4,500 | 150-220 psi | −8 to −15% |
| 4 | 25-30% | 1,200-6,000 | 180-280 psi | Baseline |
| 5 | 20% | 1,500-7,500 | 220-320 psi | +20-25% |
| 6 | 12-15% | 1,800-9,000 | 250-380 psi | +40-50% |
| 8 | 8-10% | 2,400-12,000 | 300-450 psi | +60-70% |
The 2026 Treatment Train: Side-Stream Filtration, UF, RO, Optional MVC
The defensible Sydney 2026 train is staged so each piece of equipment does one job and protects the next. Stage 1 is side-stream filtration, sized at 1-5% of total circulation flow, with self-cleaning spiral filters at 10-25 μm protecting the rest of the loop from suspended solids; CAPEX runs $50,000-200,000 for typical data centre flow rates (Genesis Water Tech, 2026). Stage 2 is UF pretreatment at 0.01-0.1 μm on PVDF hollow-fibre, operating at 10-30 psi with 90-95% recovery, which strips colloids, bacteria, and biopolymers before they foul the RO. A HydropureWater UF pretreatment system (0.03 μm PVDF, 2,000-40,000 L/h) is a fit-for-purpose option, tolerating up to 300 ppm turbidity in the feed. Stage 3 is RO at 95-99% dissolved-solids rejection, producing permeate at 10-50 mg/L TDS at 50-85% recovery on blowdown, with antiscalant injection essential to push recovery without membrane scaling; a 50,000 GPD unit runs $250,000-500,000 installed and $1.50-3.00/kgal OPEX. For hardness-limited rather than silica-limited blowdown, nanofiltration is the alternative branch: 75-150 psi, 70-85% recovery, permeate TDS 30-50% of feed. Stage 4 is the optional MVC polish, producing distillate below 10 mg/L TDS at 95-98% recovery of RO concentrate and 15-25 kWh/kgal energy, justified when discharge fees exceed $5-15/kgal or the operator targets COC above 6. The decision rule for 2026 is: if makeup is potable and discharge to Sydney Water is permitted, stop at RO permeate blending; if the site targets water-positive status or discharge fees sit above $5-15/kgal, add MVC. Across the train, a HydropureWater PLC-controlled antiscalant and biocide dosing skid handles pH adjustment, scale inhibitor injection, and biocide balancing. The permeate-spec role in the train is filled by a HydropureWater industrial RO system (recovery up to 95%), with permeate destined for cooling-tower makeup return.
| Stage | Technology | Operating parameter | Recovery / output | CAPEX band (AUD) | OPEX band |
|---|---|---|---|---|---|
| 1 | Self-cleaning side-stream filter, 10-25 μm | 1-5% of circulation | Continuous, no backwash downtime | $50,000-200,000 | Solids disposal only |
| 2 | UF (PVDF hollow-fibre, 0.01-0.1 μm) | 10-30 psi | 90-95% recovery | $80,000-300,000 | $0.30-0.80/kgal |
| 3 | RO (150-400 psi) | Antiscalant dosed | 50-85% recovery, permeate 10-50 mg/L TDS | $250,000-500,000 (50,000 GPD) | $1.50-3.00/kgal |
| 3b (alt) | NF (75-150 psi) | Hardness-selective | 70-85% recovery, permeate 30-50% feed TDS | $200,000-400,000 | $1.00-2.00/kgal |
| 4 (opt) | MVC polish on RO concentrate | 15-25 kWh/kgal | 95-98% recovery, distillate <10 mg/L TDS | $1-3M (10-30k GPD) | $4-8/kgal distillate |
| 5 (opt) | Crystallizer for ZLD | Waste heat preferred | 95-99% system recovery, solid cake | $3-8M total ZLD train | $5-15/kgal overall |
Three Sydney Deployment Scenarios Compared
Scenario A is makeup-from-mains plus RO blowdown recovery: lowest CAPEX, suited to inner-Sydney campuses with secure Sydney Water supply, treating blowdown for makeup return at RO permeate quality of 10-50 mg/L TDS. Scenario B is the AirTrunk-style hybrid, blending Class A recycled water where the mains exist with on-site blowdown RO, and the published reference is the AirTrunk SGP1 Singapore pilot, which achieved a 6% WUE improvement. Scenario C is zero liquid discharge, sending RO concentrate to MVC and crystallizer for 95-99% overall recovery at $3-8M CAPEX and $5-15/kgal OPEX, justified only where discharge is fully restricted or water-stewardship targets demand closed-loop operation. For Western Sydney greenfield sites, Scenario B currently carries permitting risk because the recycled-water infrastructure is still largely unbuilt; designers should treat on-site blowdown RO as the default and the recycled makeup as upside, not as the design basis. The OpenAI Sydney blowdown-recycling reversal in July 2026 is the strongest available evidence that piped Class A supply cannot be assumed on schedule. For a comparison of approaches outside Australia, see the Luanda data center blowdown treatment guide; for Latin American parallels, the Córdoba data center blowdown treatment guide and the Rosario data center blowdown treatment guide show how blowdown trains are being staged in other stressed basins.
| Scenario | Water recovery | CAPEX band (AUD) | OPEX band (AUD/kgal) | Permitting complexity (Sydney) | Best fit |
|---|---|---|---|---|---|
| A: Mains makeup + RO blowdown recovery | 60-85% | $0.8-2.0M | $1.50-3.00 | Low (existing trade waste) | Alexandria, inner-Sydney campuses with secure supply |
| B: AirTrunk-style hybrid (RO + optional Class A blend) | 70-90% | $1.5-3.5M | $2.00-4.00 | Medium (dual-source approval) | Western Sydney cluster where recycled mains confirmed |
| C: ZLD (RO + MVC + crystallizer) | 95-99% | $3-8M | $5-15 | High (discharge permit, waste haul) | Water-positive mandates, restricted-discharge sites |
Sydney 2026 Compliance and Permitting Checklist
Discharge to sewer must be confirmed with Sydney Water before any train is selected, and Sydney Water's June 2026 ASQ submission to NSW Parliament explicitly flags current capacity concerns that will factor into trade-waste acceptance. Permit triggers in blowdown include residual biocides, scale and corrosion inhibitors, heavy metals leached from piping, and in some refrigerant-leak cases PFAS compounds, and these may invoke EPA-style permit thresholds under NSW frameworks (Permachar, 2026-08). TDS discharge limits in some jurisdictions sit below 1,500 mg/L, which means a raw blowdown at 1,200-6,000 mg/L TDS cannot be discharged without RO or MVC polishing. Reuse water quality should be documented against the WSAA December 2025 framework for cooling-tower makeup reuse, with explicit cycles-of-concentration and chemical compatibility records. Reporting should follow the AirTrunk FY23 disclosure precedent (58% recycled water across the portfolio) and should cover WUE, PUE, freshwater displacement, and COC. Operationally, design N+1 redundancy on side-stream filtration and RO so a single component failure does not trigger an unpermitted discharge event.
Frequently Asked Questions
Why did OpenAI drop its Sydney recycled-water cooling plan in 2026?
OpenAI's Sydney campus abandoned its blowdown-recycling and Class A makeup scheme in July 2026 because the required recycled-water pipeline infrastructure was not delivered on the project schedule, forcing a return to higher freshwater makeup (Reuters, 2026-07-22; Data Center Dynamics, 2026-07). The decision confirmed that piped Class A supply cannot be assumed on time for Sydney hyperscale builds.
What CAPEX should a Sydney engineer budget for a 50,000 GPD blowdown RO system in 2026?
A 50,000 GPD RO unit treating blowdown at 50-85% recovery runs $250,000-500,000 installed in 2026, with OPEX of $1.50-3.00 per thousand gallons including energy, antiscalant, membrane replacement, and maintenance (Genesis Water Tech, 2026). Add $50,000-200,000 for upstream side-stream filtration and $80,000-300,000 for UF pretreatment to protect membrane life.
How much can blowdown RO improve a Sydney data centre's WUE?
The AirTrunk SGP1 Singapore pilot achieved a 6% WUE improvement from blowdown reuse, and AirTrunk's FY23 portfolio reached 58% recycled water utilisation including blowdown streams. For a Sydney campus moving from COC 4 to COC 6 plus RO blowdown recovery at 70-85% reuse, freshwater displacement typically lands in the 40-50% range against a COC-4 baseline.
