Upgraded version of liquid-cooled switch for cloud computing

Next-generation liquid-cooled switches deliver ultra-high throughput, energy efficiency, and thermal management for AI and cloud data centers.

Cisco Liquid-Cooled 51.2T Switch

Cisco's upgraded switch integrates direct-to-chip liquid cooling, removing up to 80% of system heat and enabling 51.2 Tbps total throughput with 64 ports at 800 Gbps each . Key features include:

• Targeted cooling for CPU, Network Processing Unit (MPU), and front-end optics, dissipating nearly 2,000 watts of heat .
• Energy efficiency: Up to 28% reduction in overall data center power consumption due to reduced cooling load .
• Scalability: Flexible configurations down to 400 Gbps or 200 Gbps per port.
• AI-ready design: Supports dense fabrics for large AI clusters, enabling high-performance workloads with minimal thermal compromise .

Alpha Networks 1.6T Liquid-Cooled Switch

Alpha Networks offers a 1.6T switch using the Broadcom Tomahawk 6 chip, achieving 102.4 Tbps per chip . Highlights include:

• Direct liquid cooling removing over 1.8 kW of heat per module.
• High-density AI/cloud deployment: Supports up to 100,000 GPUs and massive multimedia streaming.
• Compact design: Reduced rack footprint, lower acoustic emissions, and compliance with EIA/ORv3 water-cooled rack standards.
• Advanced management: BMC remote management and 1588/PTP precision synchronization.

Huawei CloudEngine XH9230-128DQ-LC

Huawei's CloudEngine XH9230-LC is a 400GE liquid-cooled switch with 128 high-density ports, optimized for AI cluster networks . Key innovations:

• 100% liquid cooling for both optical modules and main chips, with a heat dissipation ratio of 65%.
• Dual isolated liquid cooling pipes for improved thermal efficiency.
• High cabinet utilization: Eight switches per cabinet, doubling network density and reducing cooling noise by 8 dB.
• AI cluster integration: Works with Huawei core switches to support high-performance training and inference workloads.

Advantages of Upgraded Liquid-Cooled Switches

• Thermal management: Direct-to-chip and hybrid liquid cooling prevent throttling and maintain peak performance under high-density workloads.
• Energy efficiency: Reduces power consumption by 30–50% compared to air-cooled solutions, lowering operational costs.
• High throughput and scalability: Supports 400G, 800G, and beyond, enabling dense AI and cloud deployments.
• Compact and quiet: Smaller rack footprint and reduced acoustic emissions improve data center design flexibility. These upgraded liquid-cooled switches represent a major leap in cloud networking, addressing the thermal and energy challenges of modern AI and HPC workloads while enabling higher density, lower latency, and more sustainable operations. They are particularly suited for AI clusters, generative AI, and large-scale cloud computing environments.