|
| Brand Name: | Soeteck |
| Model Number: | 20ft Cold Plate (Rack) |
| MOQ: | 1 Set |
| Price: | Negotiable |
| Delivery Time: | 30-45 working days |
| Payment Terms: | D/A |
AI racks have outgrown air. Once a cabinet passes roughly 30 kW, removing the heat with air alone demands airflow volumes that a sealed steel box simply cannot deliver—there is no raised floor to pressurise, no spare cold aisle, and every additional fan quietly eats into the very power budget you are trying to protect. At the edge, where there is no plant room and no chilled-water loop to connect to, that ceiling arrives sooner than most roadmaps expect.
The 20ft Cold Plate Container Data Center moves the cooling liquid to the chip instead. Five 42U racks sit inside a single 20ft High Cube shell at 50 kW per rack, giving 250 kW of total IT capacity in one ISO footprint. Each rack carries its own rack-mounted CDU, and coolant is distributed through a distributed manifold — so every cabinet gets its own supply and return path rather than sharing one long facility loop, and a single rack can be serviced without draining the whole system.
Because the CDUs travel with the racks, the container needs no separate plant room and no site-built coolant distribution. The complete system is assembled, piped and leak-tested in the factory, then shipped as a standard 20ft High Cube container measuring 6058 x 2438 x 2896 mm — by truck, rail or vessel — requiring nothing more than a level foundation on arrival. For AI training clusters, HPC and high-density colocation that have to go where the power and the land are, that turns a construction project into a delivery.
| Model | Cooling Architecture | Rack / Tank Count | Max Power Density | Total Capacity | Container Type | Dimensions (L*W*H) |
|---|---|---|---|---|---|---|
| 20ft Immersion › | Single-Phase Immersion (Integrated FDU) |
4 Tanks | 100 kW / Tank | 400 kW | 20ft High Cube | 6058 x 2438 x 2896 mm |
| 40ft Immersion › | Single-Phase Immersion (Integrated FDU) |
12 Tanks | 100 kW / Tank | 1.2 MW | 40ft High Cube | 12192 x 2438 x 2896 mm |
| 40ft Cold Plate (In-Row) › | In-Row CDU (Liquid-to-Liquid / Air Assist) |
12 Racks (Space for CDUs) |
60 kW / Rack | 720 kW | 40ft High Cube | 12192 x 2438 x 2896 mm |
| 40ft Cold Plate (Rack) › | Rack-Mounted CDU (Distributed Manifold) |
14 Racks (Maximized Space) |
50 kW / Rack | 700 kW | 40ft High Cube | 12192 x 2438 x 2896 mm |
| 20ft Cold Plate (Rack) › | Rack-Mounted CDU (Distributed Manifold) |
5 Racks | 50 kW / Rack | 250 kW | 20ft High Cube | 6058 x 2438 x 2896 mm |
Coolant is brought to the chip through cold plates, and every one of the five 42U racks carries its own rack-mounted CDU instead of drawing from a shared facility loop. A distributed manifold gives each cabinet its own supply and return path at 50 kW per rack, so a single rack can be isolated for service without draining the system, and the compressor-based cooling that normally dominates a container's power budget is removed entirely. Heat leaves as liquid rather than as air, which is what makes a sustained 250 kW load realistic inside a sealed steel shell.
Five 42U cabinets sit inside a single 20ft High Cube shell — 6058 x 2438 x 2896 mm of ISO footprint carrying 250 kW in total. Because the coolant loop holds junction temperatures steady across every GPU node, high-TDP AI processors run without the thermal throttling that quietly erodes training throughput, and with no server fans in the hot path there is no fan-induced vibration working on the boards. The result is predictable performance from the first rack to the fifth, in a package that ships as ordinary container freight.
Focus on the core cooling demands of high-density data center deployments, where our full range of premium products leverages advanced technological advantages to reliably safeguard continuous, efficient, and stable operation.
5-30 kW per Cabinet
Airtight cold aisle enclosure, 90%+ cold air utilization rate, PUE ≤ 1.2, fast deployment and flexible expansion for medium-density IT loads.
10-80 kW Total Power
Prefabricated all-in-one design, anti-vibration & dust-proof, plug-and-play, ideal for edge computing and temporary computing demand scenarios.
30-100 kW per Cabinet
Hybrid air-liquid cooling, direct heat dissipation from GPU/CPU, AI-ready design, high power density with low energy consumption.
50-200 kW per Cabinet
Immersion cooling with dielectric fluid, 100x higher heat dissipation efficiency than air cooling, PUE ≤ 1.1, perfect for ultra-high density AI clusters.
Empowering next-generation infrastructure with superior thermal management for high-wattage computing demands.
