Alchemist Server is live. Browse our AI infrastructure line-up or talk to our team about a custom build.Talk to our team

Skip to main content

QuettaFlow Technologies

QuettaFlow VasRack

VasRack

Immersion-cooled racks at 80 to 126kW per cabinet with pPUE under 1.1, scaling from a single module to a multi-rack cluster.

Overview

80 to 126kW in one cabinet is roughly three to five times what a well-built air-cooled rack manages, and it is the number that decides whether a given room can host a cluster at all. Floor space stops being the constraint and power delivery becomes it.

pPUE below 1.1 means under 10 percent of the energy going into the room is spent on cooling rather than compute. Over a multi-year training programme that is a line on the operating budget, not a rounding error, and it is the argument that usually justifies immersion over a chilled water retrofit.

It takes the same Lead Wealth modular servers as the VasEdge and supports NVIDIA L20, RTX 40 and 50 series, A series and H series cards. Deployment runs from a single cabinet to a scalable cluster, so the first rack does not commit the design to a particular final size.

QuettaFlow states an operating range to 45C, which is five degrees wider than the VasEdge and is the specification that matters most for deployment in this region. A cabinet that tolerates a hot plant room does not need the plant room to be conditioned first.

Specifications

QuettaFlow VasRack Specifications
Compute
GPUNVIDIA L20, RTX 40 and 50 series, A series, H series
Power and Thermal
CoolantDielectric coolant
RedundancyActive plus standby pump, dual fans
Input Power400V
Cooling Capacity80 to 126kW per cabinet
pPUEUnder 1.1
Operating TemperatureMinus 10C to 45C
CoolingDielectric immersion
Physical
Form FactorSingle module or scalable cluster, Lead Wealth modular servers
Software
MonitoringFull telemetry over Modbus-TCP

Available Configurations

Manufacturer-supported options, shown for reference. Contact us to confirm a build.

Capacity
VasRack I34-28VasRack I102-86