CIOE 2026 Recap | Luxshare-Tech Deepens System-Level Component Synergy to Build a Solid Foundation for AI Computing Power in the Era of Supernodes

As the key metric for data centers shifts from “number of racks” to “computing power per watt,” a “transition period” for computing infrastructure has arrived. In the era of supernodes, higher computing density and larger cluster scales are driving high-speed interconnects, power supply, and thermal management—once considered “supporting roles”—to become the critical foundations supporting the efficient operation of computing systems. Rooted in the fundamental environment of AI computing and underlying physical processes, Luxshare-Tech continues to deepen its expertise in the three key areas of interconnect, power supply, and thermal management, building system-level synergy among core components to lay a solid digital foundation for the future development of AI computing.

On September 11, the 27th China International Optoelectronics Expo (CIOE 2026) successfully concluded at the Shenzhen World Exhibition & Convention Center. Against the backdrop of ongoing challenges in upgrading AI computing infrastructure, Luxshare-Tech—as a globally leading provider of intelligent computing center infrastructure solutions—has focused on upgrading the foundational capabilities of AI supernodes and aligning with the evolution of the global advanced computing ecosystem. The company made a major debut at the exhibition with its all-new “Xuanyuan Series System-Level Infrastructure Solutions” and a matrix of core components for advanced computing systems, The company demonstrated solutions such as the 3.2T NPO optical engine and high-speed interconnects via orthogonal backplanes, showcasing its system-level collaborative capabilities for AI supernodes.

Luxshare-Tech Booth at CIOE 2026

Born for AI Super Computing, Xuanyuan Series System-Level Infrastructure Solutions Redefining the Foundational Logic of the Supernode Era

To meet computing deployment needs at various levels, Luxshare-Tech has continuously iterated its solutions. In 2025, the company launched the Xuanyuan V2 solution, which focuses on single-cabinet-level infrastructure construction. It integrates in-cabinet CDUs, power racks, and cable modules, making it fully compatible with high-performance inference and light-load model training scenarios. At CIOE 2026, Luxshare-Tech showcased its all-new Xuanyuan Series system-level infrastructure solutions in their entirety for the first time in the form of standard ORW cabinets. Thanks to their systematic and integrated design philosophy, these solutions have garnered significant attention from the industry.

The Xuanyuan Series system-level infrastructure solutions are specifically designed for AI super-computing scenarios, precisely addressing the current pain points in the development of AI supernodes. Currently, super-node computing clusters face normally three core challenges: interconnect, thermal management, and power supply. As computing density continues to rise, it is driving the widespread adoption of full-range direct liquid cooling. Meanwhile, operations involving ultra-high currents lead to issues such as high line losses and insufficient overall energy efficiency. Additionally, cabinet space and load-bearing capacity are becoming increasingly saturated, propelling industry infrastructure into a new phase of development characterized by high-density racks, intelligent power supply, and extreme liquid cooling.

Xuanyuan V3 focuses on AI-POD prefabricated module-level applications and is positioned as the core infrastructure for high-speed interconnection in supernodes. The comprehensive solution includes side-mounted CDUs, power cabinets, and computing cabinets, with solid-state transformers (SST) available as an option for certain scenarios. The solution deeply integrates two core technologies—HVDC high-voltage power supply and full-range liquid cooling—and combines them with high-end core interconnect components such as NPC, CPC, and XPO to build a synergistic system that unites high-speed interconnection, intelligent power supply, and efficient cooling. This forms a highly complementary partnership with the Xuanyuan V2, which focuses on single-cabinet scenarios, thereby covering a wider range of computing deployment scenarios. At the core design level, the Xuanyuan V3 achieves comprehensive technological upgrades:

Luxshare-Tech's Demo of Xuanyuan V3 Rack

System Level

The solution delves deeply into the entire chain of underlying processes—including materials, manufacturing, reliability, and testing—to provide customers with comprehensive technical guidance and assurance for early-stage system planning, solution implementation, and performance validation.

High-Speed Interconnect Level

Through an on-site demonstration of an orthogonal cabling architecture, the solution visually illustrates the application of core products such as the 448G Intrepid™ Nexus backplane connector, 224G OmniStack™ (NPC), and 448G KOOLIO™ (CPC). This architecture further expands the deployment options for the Xuanyuan series’ high-speed interconnects, complementing the rear- and mid-cable backplane designs of the Xuanyuan V2.

Thermal Management

Featuring blind-plug-in capabilities for front and rear nodes, built-in cold plates, and a modular design with integrated flow channels, this solution provides a high-quality foundational thermal management solution for high-density computing through comprehensive flow channel optimization.

Power Supply

It supports a dual-mode power supply architecture, encompassing two solutions: 54V DC cabinet-integrated power supply paired with a liquid-cooled busbar, and ±400V HVDC high-voltage direct current busbar. Combined with an orthogonal power delivery design, this expands the range of power supply options for computing power.

At this exhibition, Luxshare-Tech also conducted an on-site demonstration of the “Xuanyuan V3 Front-to-Front Interconnect Architecture.” One of the core changes in Xuanyuan V3 is the shift from traditional middle plates or cable cartridges to front-to-front cable/PCB interconnects. The Compute Tray and Switch Tray are arranged in opposite directions and connect directly via components such as the NPC, NEXUS, and Cable Box, forming a shorter, clearer high-speed signal path. This architecture reduces excessively long interconnects between nodes, helping to improve signal quality and cost efficiency while minimizing the impact of complex cable management on cabinet space.

The continuous evolution from Xuanyuan V2 to V3 represents not only Luxshare-Tech’s in-depth exploration of AI supernode infrastructure but also the company’s relentless refinement of core technologies across the entire component supply chain. Against a backdrop of increasing homogenization in product design and architecture within the industry, the company is deeply committed to developing integrated solutions that combine interconnect, thermal management, and power delivery, helping customers rapidly validate and implement computing power projects. The component industry has now shifted from competition based on individual product iterations to competition centered on systematic solutions. During the R&D process for the Xuanyuan platform, Luxshare-Tech simultaneously refined its supporting testing and validation system, laying a solid foundation for large-scale mass production and delivery.

Leveraging Component Capabilities to Build the Foundation of the Ecosystem

Key Technologies in Optical Interconnect, Copper Interconnect, Thermal Management, and Power Supply

Full-Stack Empowerment of the AI Computing Infrastructure

Relying on continuous technological innovation and ecosystem development, Luxshare-Tech has established a full-stack product portfolio covering all computing scenarios. These products are deeply compatible with global advanced AI computing ecosystems and diverse standard architectures. At this exhibition, through the “Luxshare-Tech Thought Leadership Forum”, the company shared its insights on development trends, technical achievements, and industrial implementation experiences in core technology areas such as “copper interconnects, optical interconnects, thermal management, and power supply,” collaborating with industry partners to explore the evolution of next-generation optoelectronic interconnect technologies and AI computing infrastructure.

Addressing the computing power demands of the ultra-node era, Luxshare-Tech further showcased its forward-looking strategic planning and achievements in system-level collaborative innovation at CIOE 2026. Moving forward, Luxshare-Tech will continue to work closely with partners in the advanced computing ecosystem to jointly build a highly efficient, high-density, and reliable AI infrastructure, laying a solid digital foundation for the global computing industry’s transformation.