The Open Data Center Conference 2026 and the inaugural Computing Expo, known as ODX 2026, were held in Beijing from Sept. 2 to 4. During the event, Envicool showcased liquid cooling products for data centers and AI servers, spanning chip-level thermal management, node-level piping, rack-level temperature control and data center cooling systems.

Envicool describes its liquid cooling portfolio as covering the “full chain, full range of scenarios and full lifecycle.” The portfolio mainly includes cold plates, quick connectors, manifolds, CDUs, racks, coolants, piping, cooling sources and intelligent control systems.

Double-Sided GPU Cold Plates Target High-Power Chips

At the chip-level thermal management stage, Envicool showcased a high-power, double-sided GPU liquid cooling cold-plate module. The company said the module can meet the cooling needs of multiple chips within a 1U space.

As AI chip power consumption rises, servers must fit cold plates, piping and quick connectors into limited space while maintaining adequate flow, sealing performance and serviceability. Double-sided cold plates draw heat from both sides of a chip or module, making them one cooling design for high-power computing nodes.

Quick Connectors and Manifolds Support High-Density Deployment

For node-level liquid cooling systems, Envicool showcased an integrated-flow-channel manifold, stainless steel hoses and its MQD series of quick connectors.

According to Envicool, the MQD quick connectors use a steel-pin locking mechanism and a dual-ring sealing structure. They are about 40% smaller than the comparable UQD products cited by the company, while the MQDB series can compensate for installation deviations of plus or minus 1 millimeter. Envicool said the connectors had passed 5,000 insertion-and-removal cycles and have a designed service life of more than 10 years.

These figures come from Envicool’s product materials and internal testing. Actual service life and sealing performance may vary depending on system configuration, coolant, operating environment and maintenance conditions.

For rack-level liquid cooling, Envicool said its blind-mate manifold can support high-density racks with power levels of up to 300 kilowatts. The manifold also uses an anti-splash design intended to reduce the risk that accidental leaks will affect server equipment.

Prefabricated Piping Reduces On-Site Construction

Envicool also showcased liquid cooling piping manufactured and assembled in factories using a modular design. The piping is processed, welded, assembled and tested at the factory before being delivered to the data center for on-site installation.

Compared with cutting and welding on site, prefabricated piping can reduce processing work at data center construction sites and lower the risk of leaks caused by inconsistent installation precision. Actual installation efficiency and reliability still depend on project design, construction management and acceptance procedures.

The products on display also included Envicool’s SoluKing series of liquid cooling fluids. The company said the series is designed for use with a range of wetted materials, including copper, steel and aluminum, and includes propylene glycol-based and water-based coolants. Its corrosion resistance and third-party certifications should be assessed against relevant test reports and information publicly released by certification bodies.

Three-Tier Liquid Cooling Testing System

Envicool said it has established testing systems at the component, rack and data center levels. Testing covers pressure resistance, vibration, aging, thermal performance and cleanliness, among other areas.

In liquid cooling projects, cold plates, piping, quick connectors, coolants and cooling-source equipment from different vendors must operate together. Interface standards, material compatibility, installation cleanliness, leak detection and the allocation of after-sales responsibilities can directly affect system deployment and long-term operations.

Envicool said it hopes to reduce the integration work between products from different suppliers through a portfolio covering the main liquid cooling components and system-level testing.

High-Power Racks Drive Liquid Cooling Upgrades

During the conference, Envicool product manager Huang Sihao delivered a speech titled “Coordinated Full-Lifecycle Liquid Cooling Delivery for Full Racks: A Leap in Computing Density, with Reliability First.”

According to Huang, the power consumption of some high-performance computing chips has already exceeded 2,000 watts. The power requirements of some AI racks are rising from about 30 kilowatts to more than 100 kilowatts, with high-density configurations potentially reaching more than 300 kilowatts. As rack power and heat density rise, traditional air cooling systems face greater thermal management pressure.

These power figures reflect some high-performance products and high-density deployment scenarios. They do not represent the typical levels of all AI chips or data center racks.

Envicool divides the engineering challenges facing full-rack liquid cooling into three levels:

Chip-Level Thermal Management

Flow-channel design becomes more complex in multi-chip systems as local hot spots and heat flux increase, placing higher demands on the heat-transfer capacity of cold plates and flow distribution.

Node-Level Integration

Servers must fit cold plates, piping and quick connectors into limited space while meeting requirements for flow, size, maintenance and sealing.

Rack-Level Coordination

High-power racks require greater coolant flow and impose higher demands on manifold machining precision, material compatibility, leak detection and rack-level reliability testing.

Envicool said its liquid cooling solution incorporates cold plates, quick connectors, manifolds, CDUs, piping, coolants, cooling sources and control systems into a unified product and testing system to support the delivery and operation of high-density AI racks.

As per-chip and per-rack power consumption continues to rise, liquid cooling is being adopted by more high-density AI infrastructure projects. The choice between liquid cooling and air cooling, however, still depends on rack power, data center design, deployment costs and operating conditions.