Custom H100 Liquid Cooling Plate 700W Copper Microchannel Cold Plate

This custom liquid cooling plate is engineered for NVIDIA H100 SXM5 and PCIe 80GB GPUs running AI training and high-performance computing workloads in data centers. Mounted directly on the GPU, the cold plate draws heat straight from the chip into the coolant loop, keeping a 700W card stable under sustained full load. It is a direct-to-chip microchannel cold plate built for AI servers, HPC clusters and hyperscale data centers transitioning from air cooling to liquid cooling.

Photos note: The images in this listing are reference examples of this product type, shown to illustrate the product. The final product is manufactured to the customer's drawings and specifications.

• Oxygen-free copper base for fast heat transfer from the GPU die
• Microchannel fin, 50–100 μm, delivering a large heat-exchange area in a compact plate
• Vacuum brazing plus CNC machining for a 100% leak-free seal
• Thermal resistance ≤0.5°C/W for stable temperatures at 700W per card
• Pressure resistance up to 1.8 MPa for data center coolant loops
• 100% helium leak testing and pressure testing on every unit
• Fits H100 SXM5 and PCIe 80GB GPUs; dimensions, port positions and thermal interface customized to your drawing

The standard version is made from oxygen-free copper, the best choice for heat transfer from a high-power GPU die. A 316L stainless steel version is available for deionized water (DI water) loops. Aluminum alloy, stainless steel and brass builds are available for custom projects made to your drawing. One plate cools one GPU card; multi-card and complete HGX baseboard sets are available on request.

Every order is built to your drawing and your cooling requirements. Free ODM design and thermal simulation support is included: send your GPU model, target power, coolant type and flow rate, and the plate dimensions, port positions and thermal interface are designed for your server. Minor customization, drawing-based customization, sample-based customization and full customization are all supported.

• Send Inquiry
• Quote and Design
• Prototyping
• Testing
• Mass Production
• Delivery

Sample production takes 5–7 days. Mass production takes about 7–15 days after the deposit is received.

Every unit passes 100% helium leak testing and pressure testing to 1.8 MPa, plus a thermal efficiency test, before shipment. A machinery test report is provided, and video outgoing-inspection is provided. This is the same quality control used for data center liquid cooling deployments, where a leak is not an option.

Single item in an export carton, ready for ocean or air freight from Shenzhen Port. Sample orders ship in 5–7 days; mass orders in 7–15 days after deposit.

Shenzhen Elite Technology Co., Ltd is a manufacturer based in Dongguan, China, supplying custom injection moulds, CNC-machined parts and thermal components to customers worldwide. Every order is built to the customer's drawings and specifications, with full customization and export service. Contact us with your GPU model, quantity, coolant type and drawings for a quotation.