Specifications
Product Type
Coolant Distribution Unit (CDU)
In-Row Capacity
1.3 MW (liquid-to-liquid)
In-Rack Capacity
100 – 240 kW (standalone, liquid-to-air)
In-Row Approach Temp
45 °C (max allowable 60 °C)
In-Rack Approach Temp
15 °C
In-Row Flow Rate
≥ 1200 L/min
Pump Class
Class 1, overall efficiency ratio near 100%
Filtration
Filters on both primary and secondary sides
Monitoring
Coolant level, flow, humidity, leak, pressure, temperature
Serviceability
Field-replaceable fans, pumps, piping and sensors
Container Deployment
40-ft container, 510 kW equipment density
Rack Support
4x NVIDIA GB200 or GB300 NVL72 racks
Deployment Time
Container build in ~2 weeks
Cooling Method
Direct-to-chip liquid cooling
Overview
Aivres builds coolant distribution units in two distinct shapes for two distinct facility situations. Where a building has no facility water, the standalone in-rack liquid-to-air CDU delivers 100 to 240 kW at a 15 °C approach temperature with an integrated sensor set covering coolant level, flow, humidity, leak detection, pressure and temperature, plus field-replaceable fans, pumps, piping and sensors.
Where facility water is available, the in-row liquid-to-liquid CDU scales to 1.3 MW with a 45 °C approach temperature and a 60 °C maximum allowable temperature, maintaining a flow rate of at least 1200 litres per minute. Class 1 pumps with an overall efficiency ratio near 100% hold pumping overhead low, and filters sit on both the primary and secondary sides for loop protection.
For rapid multi-rack deployment Aivres integrates the liquid-cooling cabinets, network, power distribution, CDU, air conditioning, cold sources, monitoring and fire protection into a 40-foot container carrying 510 kW of equipment — enough for four NVIDIA GB200 or GB300 NVL72 racks, buildable in roughly two weeks.
Key Benefits
1.3 MW in-row capacity with 45 °C approach handles full NVL72-class racks on the technology loop. Two architectures — liquid-to-air in-rack for facilities without water, liquid-to-liquid in-row where water exists. Near-unity pump efficiency from Class 1 pumps reduces parasitic load. Dual-side filtration protects both loops. Containerised deployment ships four NVL72 racks' worth of cooling capacity as one unit in about two weeks.
Applications
NVIDIA GB200 and GB300 NVL72 rack deployments, AI factory builds on greenfield sites, colocation retrofits without facility water (in-rack L2A path), and rapid-deployment containerised AI capacity.
Request a Quote — AIVRES IN-ROW CDU 1.3 MW — LIQUID-TO-LIQUID COOLANT DISTRIBUTION UNIT
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →