Cap Peak Water Withdrawal. Unlock Permits & Protect Basins.
72-hour rolling MILP optimization that co-dispatches AI batch compute, cross-basin failover, cooling storage tanks, and grid-safe dry cooling — keeping facilities inside municipal limits before utilities spend $10M–$40M per MGD on peak supply capacity.
- Peak-day cut
- 45.4%
- 10-basin peak avoided
- 2.996 MGD
- Extra dry-cooling energy
- 1,792.9 MWh
- Homes protected
- 9,986
Precomputed PuLP CBC fleet compare for each registry site's 2026 peak window. Controls: flexible share 0.20, tank 40,000 gal, cap ratio 0.78, stage-2 drought (flex_event, tighten 0.30), reclaimed water 0.35, 100 MW, cycles of concentration 4, $85/MWh. Weather is the committed Open-Meteo archive fixture for that window. Capex uses Han et al. 2026, $10M–$40M per MGD of peak supply avoided.
Dubai, UAE
WRI Stress 83/100 · DEWA Desalination Cost & TSE Cooling Mandate
Peak Cut
—
—
5-Yr Net ROI
—
$11.50/kgal
CAPEX Avoided
—
—
Grid Peak Add
—
waiting for solve
Optimization Solver
PuLP CBC 72h Rolling MILP
Official US Grid Data
EIA-930 MISO · AZPS · PJM
Water-Stress Mapping
WRI Aqueduct 4.0 Tiers
AI Cluster Workload
Alibaba PAI 759,033 GPU Jobs
Regulatory Standard
EU EED (2024/1364) & ISO 30134-9
Every City Worldwide + 10 Calibrated Stress Hubs
Run Any City on Earth or Inspect Global Water Corridors
Search any global data center market — from Mumbai, Riyadh, and Tokyo to London, São Paulo, or Johannesburg — to plot it live on the basin map and run 2026 PuLP CBC MILP telemetry.
Basemap is OpenFreeMap (OpenStreetMap data, no API key). Pins use registry latitude and longitude. Color is the registry WRI score. The dashed arc is the model sister site for the selection. Scroll-zoom turns on after you click or focus the map.
Dubai, UAE
DEWA Desalination Cost & TSE Cooling Mandate
Expensive desalinated water ($11.50/kgal) & extreme wet-bulb penalties make sovereign failover + TSE reuse essential.
Sister Failover Route: Frankfurt Sovereign sister
23°C Northern European spread slashes MENA peak cooling OPEX while respecting data-residency batch windows.
- Shaved vs baseline
- 428,359 gal
- This site’s uncontrolled peak day minus its CBC peak
- Permit Shield (w/ TSE)
- 94/100
- EU EED Grade C
- 5-Yr Net Financial ROI
- $6.94M
- Households Protected
- 1,428/d
Precomputed CBC for 2026-07-20 to 2026-07-26. Stage-2 defaults: cap 0.78 tightened 30%, 35% reclaimed, 40,000 gal tank, 100 MW.
Physics + Compute Co-Optimization
Four Physical Levers Co-Dispatched Every 24 Hours
Unlike passive sustainability dashboards, ThirstCap solves a deterministic PuLP CBC mixed-integer linear program every rolling 72 hours — respecting grid peaks, wet-bulb humidity, and job SLAs simultaneously.
Real-Time Ingestion
72h Wet-Bulb Forecast
Open-Meteo T_db + RH% -> Stull T_wb
Official Grid Stress
EIA-930 MISO · AZPS · PJM Top-15% Guard
Alibaba PAI 759K Trace
Fixed (<15m), Deferrable, Pausable MW
Utility Water-DR Signal
Stage 0–3 Firm Cap + TSE Reclaimed Share
PuLP CBC
Rolling 72h MILP Core
• Firm Daily Cap Constraint
• Calibrated 0.19 L/kWh Tower
• Stull Wet-Bulb CoC Lift
• Zero Sample Fallback
Co-Optimized Controls (Click to Inspect)
Built For Every Water Stakeholder
Commercial Value Across Operators, Utilities & Finance
Unlock Stalled Zoning & Aquifer Permits Across High-Growth AI Corridors
The simulator shows what a firm daily cap does to a 100 MW load on real weather. The dispatch tab prints scheduler hints. It does not connect to Slurm or Kubernetes.
- Daily cap, shortfall, and permit-shield score come from the CBC plan response
- Dispatch windows are text hints for a batch scheduler, not live job control
- Reclaimed share is an input. The plan reports the potable and reclaimed split
Precomputed fleet figure
Permit shield in the default 2026 fleet
94 / 100
Peak-day cut 45.4% with the stage-2 default controls. Not a permit that has been filed.
Solver Basis
PuLP CBC MILP
CAPEX Benchmark
$10M–$40M/MGD
Enterprise Economics & Statutory Compliance
From Water Scarcity Risk to Boardroom Financial Return
CFO 5-Year Financial Model
100 MW Dubai and Phoenix, same default solve
- Dubai avoided supply CAPEX (Han et al. 2026)
- $4.28M–$17.13M
- Phoenix avoided supply CAPEX
- $3.69M–$14.76M
- 5-year net ROI, Dubai / Phoenix
- $6.94M / $4.32M
- Extra dry-cooling energy, Dubai / Phoenix
- 274.5 MWh / 286.8 MWh
Windows 2026-07-20 to 2026-07-26 (Dubai) and 2026-07-13 to 2026-07-19 (Phoenix). The model prices tank steel at $2.50 per gallon. Electricity in this solve is the $85/MWh default. These are scenario inputs, not utility invoices.
Permit Shield & EU EED (2024/1364)
Audit-Ready WUE & Community Impact Telemetry
Permit Shield
94/100 on all 10
From the fleet rows
Peak-day cut
45.4%
Cap held on every site
Daily homes spared
9,986
Sum of 10 campuses
Commercial Deployment Paths
What you can actually run
Run a basin
Live solverOpen the dashboard and solve one registry site or any geocoded city. Results come back from the FastAPI CBC service.
- Live 7-day forecast or the site's 2026 peak window
- Tank sweep and 10-city fleet for the default controls are precomputed
- Other control sets solve with a 4 second CBC budget per window
Compare 10 corridors
Precomputed CBCThe 10-city tab renders the committed 2026 summer fleet. Live mode on the free tier is labelled when it reuses that run.
- One row per registry site, with the window that was actually solved
- Capex uses Han et al. 2026 ($10M–$40M per MGD)
- WRI scores are the registry values, not a live Aqueduct query
Read the model
GitHubThis is a student research simulator. There is no pilot checkout, price, or booking form.
- PuLP CBC rolling-horizon MILP in backend/app/services/optimizer.py
- Open-Meteo fixtures, EIA-930 profiles, and the Alibaba GPU trace under backend/data
- Precompute script: backend/scripts/precompute_sweeps.py
Validate Your Basin's Peak Water Cap in Minutes
Run real-time 7-day forecasts or 2026 summer heatwave stress tests across all 10 flagship basins — or search literally any city on Earth live.