LITEON Technology (2301) participates in the Open Compute Project (OCP) Global Summit 2026 in San Jose, California, from October 12 to 15, showcasing its next-generation AI factory infrastructure portfolio centered on 800 VDC architecture. The exhibition features power rack solutions supporting the NVIDIA Vera Rubin platform, integrated power and thermal management solutions designed for NVIDIA MGX and OCP ORv3 architectures, demonstrating its Grid-to-Chip technology portfolio.
LITEON showcases its 800 VDC Power Shelf integrated with Battery Backup Unit (BBU), demonstrating next-generation 800 VDC power architectures for AI factories. To address the growing thermal demands of high-density AI computing, LITEON debuts the 2.6MW Enterprise Cooling Distribution Unit (CDU) from its strategic investment partner, DCX Liquid Cooling Systems. The solution delivers up to 2.6MW of heat transfer capacity, achieves an approach temperature of 3°C, and incorporates an N+1 redundant pump design to maximize cooling efficiency while ensuring system reliability for high-density AI and HPC deployments. Also on display is the CDU's next-generation HMI and integrated software layer, designed to enhance operational efficiency and provide clear, at-a-glance visibility into key system parameters.
For the NVIDIA Vera Rubin NVL72 platform, LITEON is also showcasing an integrated infrastructure solution featuring a 95kW DC/DC Power Shelf and a 280kW Liquid-to-Liquid (L2L) In-Rack CDU. Designed based on NVIDIA MGX architecture, the solution combines 800 VDC distribution, energy backup, and thermal management technologies to reduce transmission losses and heat generation, improve energy efficiency and power density, simplify power infrastructure, and free additional rack space for high-density AI deployments.
"In the AI era, data center competitiveness is no longer defined solely by compute scale,” said John Chang, General Manager of LITEON Cloud Infrastructure Platform and Solution SBG. “Instead, it is measured by tokens per watt and how efficiently power can be converted into meaningful AI output. As AI platforms continue to scale in power density, data centers must simultaneously optimize energy efficiency, cooling performance, and deployment flexibility. High-efficiency power delivery and thermal management architecture have become the critical foundation for scaling future AI Factories."
LITEON is also showcasing an integrated AI power rack solution designed for the ORv3 architecture, integrating an 180kW Power Shelf and a 108kW Capacitor Backup Unit (CBU). Through centralized power delivery and supercapacitor backup capabilities, the solution addresses dynamic AI workloads while enhancing power stability and overall system reliability.
Optimized for liquid cooling compatibility, the ORv3 architecture enables higher-density deployments and greater infrastructure flexibility. LITEON shows key cooling modules including 120kW and 220kW In-Rack CDUs, which can be integrated with advanced cooling technologies such as Direct-to-Chip architectures. These solutions enable greater GPU density, enhanced thermal performance, and flexible infrastructure design for next-generation AI and HPC platforms. By integrating power, backup, and thermal management systems, LITEON helps customers build efficient, reliable, and scalable AI infrastructure.
As AI processors continue to increase in power consumption, efficiently delivering electricity from the rack level to the GPU has become a critical challenge for next-generation AI infrastructure. To address this challenge, LITEON is showcasing its comprehensive DC-DC power conversion portfolio, including 800V Power Bricks, IBC Modules, 50V Power Bricks, and Voltage Regulator Modules (VRM). These product lines bridge every stage of power delivery, from 800 VDC distribution and server-level power conversion to chip-level voltage regulation, demonstrating LITEON's complete Grid-to-Chip energy architecture for future high-density AI computing platforms.
Among these, the 800V DC/DC power brick can be deployed on the HVDC power distribution board (PDB) for downstream distribution and step-down conversion. Multiple bricks can be paralleled to match different power requirements, with active current sharing improving supply stability and reliability. The intermediate bus converter (IBC) then performs the next stage of voltage conversion. This raises overall power delivery efficiency, reduces distribution losses, optimises board area and system cost, and improves supply stability in multi-output architectures. Finally, the voltage regulator module (VRM) delivers a precise, stable operating voltage to the GPU, completing the “last mile” of power delivery from the rack to the chip.
As AI computing infrastructure continues evolving toward higher power density and greater energy efficiency, LITEON is expanding its capabilities across power delivery, thermal management, and system-level energy optimization. By extending its expertise from infrastructure power systems to comprehensive Grid-to-Chip energy management, LITEON is building a complete portfolio of AI infrastructure energy solutions.
LITEON continues to deepen collaboration with NVIDIA and the broader OCP ecosystem while leveraging its global manufacturing footprint and localized service capabilities. Through close ecosystem partnerships, LITEON is helping customers address the growing demand for high-performance, highly reliable, and highly scalable infrastructure in the AI Factory era, accelerating deployment and expansion of next-generation AI data centers.
LITEON Technology 2026 OCP Global Summit Exhibition Information:
Date: October 12–15, 2026
Venue: San Jose McEnery Convention Center, USA
Booth: C51, Hall 2
