LITEON Highlights NVIDIA 800 VDC Power Architecture and Thermal Management at OCP Global Summit

LITEON Technology used the OCP Global Summit 2026 in San Jose to spotlight hardware building blocks it says fit the energy and heat demands of today’s AI data centers. In its presentation, the company tied its next-generation AI factory infrastructure portfolio to NVIDIA’s 800-volt direct-current, or 800 VDC, power architecture, alongside a thermal management portfolio intended to improve how systems handle high-density workloads.

The announcement came as compute providers and infrastructure vendors continue to converge on architectures that reduce power conversion losses and simplify power distribution at scale. In broad terms, 800 VDC systems aim to move electrical power across data-center racks with higher voltage, which can reduce current for a given power level and potentially lower resistive losses across cabed runs and connectors.

At the same time, thermal management has become a central constraint for AI servers, where power draw and heat flux can rise sharply with model size and utilization. LITEON’s focus on thermal management was presented as a complementary track to the 800 VDC approach, reflecting the industry’s push to coordinate power delivery and cooling designs rather than treating them as separate engineering problems.

According to the event-linked report, LITEON’s participation in the OCP summit centers on an “AI factory infrastructure portfolio.” The company framed its booth and showcase around NVIDIA’s 800 VDC architecture and the company’s thermal management offerings, positioning the package as relevant to the next wave of rack-scale and facility-scale deployments.

OCP Global Summit is organized by the Open Compute Project, a consortium that works on open specifications for data-center designs. OCP summits tend to serve as venues where vendors describe how their hardware and reference designs map to emerging standards and interoperability goals, particularly for large-scale cloud and enterprise environments.

While the report describes the themes of LITEON’s showcase, it does not provide additional specifics on product names, performance claims, or measurable outcomes such as efficiency improvements, thermal targets, or compatibility with particular NVIDIA platforms. It also does not disclose whether the company is already sampling any of the highlighted components, or when they might move into production deployments.

NVIDIA is widely associated with platform-level accelerators and the data-center ecosystem that supports them, so third-party hardware vendors often use summits like OCP to announcement how they plan to fit into power, cooling, and system integration requirements. In that context, LITEON’s messaging appears aimed at reinforcing its role in the physical infrastructure layer around AI compute, where uptime, maintainability, and power delivery stability are as important as raw processing performance.

For data-center buyers watching the next steps, the key question is whether LITEON and other infrastructure partners will translate these architecture and portfolio themes into concrete, standardized system components that can be evaluated across vendors. The next indicates to watch are any additional disclosures about specific modules, interoperability details, and any testing results tied to 800 VDC power distribution and the associated thermal-management approach.