THE MODEL
Community Compute Hub 01.
The flagship model places approximately 500 NVIDIA RTX 5090 GPUs inside roughly 3,000–5,000 square feet of an existing building — closed-loop liquid cooled, grid-connected, solar-matched, and designed to leave the rest of the building productive.
500
GPUs
125
Four-GPU hosts
5,000 sq. ft. max
Compute footprint
$0.80
Net GPU-hour target
ONE BUILDING · THREE USES
The compute hub is part of the building — not the entire building.
3RD FLOOR
AI Compute
- 500 RTX 5090 GPUs
- 125 four-GPU hosts
- Closed-loop liquid cooling
- High-performance regional AI infrastructure
COOLING & ENERGY
Closed-loop. Dry-cooled. Solar-matched.
CLOSED-LOOP DIRECT-LIQUID COOLING
Coolant continuously circulates inside a sealed system.
NO ROUTINE PROCESS-WATER COOLING
Base design avoids evaporative cooling towers.
NO SURFACE-WATER THERMAL DISCHARGE
No river, lake or ocean is used as a waste-heat sink.
OPTIONAL HEAT REUSE
Recoverable heat can serve useful local thermal loads.
OPTIONAL HEAT REUSE
At approximately 70% utilization, the planning model estimates roughly 220 kW of potentially recoverable thermal output from one hub. Potential uses:
Domestic hot-water preheatingGreenhousesAquaculturePoolsCommercial buildingsApartmentsIndustrial hot-water processesCommunity facilities
All final installations are subject to engineering, utility, fire-code and building-code review.
Closed-loop direct-liquid cooling
Rooftop dry cooling
No evaporative cooling towers
No routine process-water cooling
No surface-water thermal discharge
Grid-connected reliability
Solar-matched annual electricity
Optional waste-heat recovery