Wiwynn Brings Vera Rubin NVL72 and AMD Helios Racks to OCP Global Summit – Unite.AI

Wiwynn Corporation unveiled its OCP Global Summit 2026 lineup on October 12, 2026, showcasing the NVIDIA Vera Rubin NVL72 and AMD Helios rack platforms, advanced cooling technologies, and a demonstration of co-packaged optics at booth C74 at the San Jose Convention Center.

In an announcement in its newsroom, the company said the showcase will span computing, cooling, power and connectivity.

“As AI workloads push compute density to unprecedented levels, the industry needs partners who can deliver proven, full-stack engineering across the rack,” said William Lin, president and CEO of Wiwynn. Lin said the company has end-to-end capabilities, from custom design to large-volume rack integration, as well as continued investments in cooling, power and optical interconnect, and that Wiwynn is committed to building next-generation AI infrastructure at scale with cloud provider customers and ecosystem partners.

Next generation AI rack platforms

Wiwynn said it and Wistron are among the first NVIDIA Pilot Build partners for the Vera Rubin NVL72 platform, handling engineering validation and design. Wiwynn describes the Vera Rubin NVL72 as a liquid-cooled rack-scale system that integrates 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs connected via sixth-generation NVLink.

The AMD Helios rackscale solution, based on the Open Rack Wide standard, unifies AMD Instinct MI450 series GPUs, 6th generation AMD EPYC server CPUs, AMD Pensando networking technology, and AMD ROCm open software into a single architecture for large-scale AI training and inference. Wiwynn said it is working with AMD to apply its expertise in rack integration and manufacturing to bring the Helios architecture to customers at scale, with configurations tailored to cloud providers’ requirements.

Advanced thermal engineering

The cooling portion of the stand covers in-packaging cooling, advanced materials and lightweight system-level design. For on-pack cooling, the company is exploring diamond-based composite materials as an alternative to copper, bonding diamond directly to silicon with microchannel structures integrated into the material. Wiwynn said the approach takes advantage of diamond’s thermal conductivity to shorten the heat transfer path at the chip level.

A second design pairs an ultra-thin double-sided cold plate with a cavity PCBA layout, intended for high-power AI ASICs using Vertical Power Delivery. Wiwynn reports 6 kW of cooling capacity from the design, an increase in thermal performance of more than 15%, and a reduction in power delivery loss by up to 80%, and says it jointly addresses thermal, energy, and space constraints in next-generation AI modules.

Lightweight cold plate and internal manifold design reduces the amount of metal used and is paired with lightweight tubing; Wiwynn said the result is at least a 30% reduction in the weight of cooling components, making it easier to integrate, transport and install high-density AI racks as deployments scale up. With Corintis, the company is also demonstrating topology-optimized microchannel layouts matched to individual chip power maps, focusing on high heat flux regions while keeping thermal and hydraulic performance in balance, an approach that Wiwynn says could further improve cold plate performance for next-generation AI ASICs.

Wiwynn’s event page for the summit lists additional cooling items at the booth, including 3D printed microchannel cooling for Intel CPUs, double-sided cooling for Vertical Power Delivery, lightweight plastic cold plates and manifold designs, and two-stage cooling solutions, with live demonstrations at the booth.

Optics and Summit Co-Package Program

For optical interconnects, Wiwynn introduces a co-packaged optics (CPO) scale-up rack solution built with ecosystem partners Ayar Labs, Corning, GUC and Senko. The rack-level work covers fiber routing, patch panel integration and multi-vendor fiber management, and includes a live demonstration of optical engine telemetry. Wiwynn said the demonstration highlights an open ecosystem approach and its ability to integrate optical, mechanical, cooling and system-level work into a scalable rack-scale AI infrastructure platform. The event page states that CPO’s live demonstration with Ayar Labs is intended to show how rack-scale optical interconnects remove the bandwidth, power efficiency, and scalability constraints of traditional copper within large AI clusters.

Wiwynn said the technologies will be explored throughout the summit in Executive Track and Expo Hall Talk sessions, engineering workshops and the Future Technology Symposium. The 2026 OCP Global Summit will take place October 12-15, 2026 at the San Jose Convention Center. On October 12, Wiwynn Vice President Tony Wen hosts an Executive Track session titled Building the AI ​​Factory: Overcoming the Infrastructure Challenges of AI at Scale. On October 13, 2026, senior director Lentis Pai gives a talk in the Expo hall titled Powering AI Infrastructure: Innovations in Compute, Power, Connectivity and Cooling.

The October 14, 2026 schedule lists a Future Technology Symposium data center sustainability talk by Nick Liu and Jake Hsieh on diamond-based materials for direct liquid cooling of next-generation AI chips; a Cooling Environments session by Jeff CA Chen and Shawn Ma on rack-scale weight reduction and reliability validation for liquid-cooled server systems using lightweight materials; a server session by Steven Chien with Microsoft’s Zi Ping Wu on board-level electrothermal co-simulation of VRMs and power delivery networks; and FTI workshop sessions on short-range optical interconnects presented in collaboration with Ayar Labs staff: Willie Chen with Nandita Aggarwal on rack-ready high-power ELSFPs developed to support CPO deployment in AI scale-up and scale-out data centers, and Kinny Chen with Vishal Chandrasekar on rack-scale open optical infrastructure for AI scale-up with co-packaged optics.

On October 15, 2026, Jake Hsieh presents Cooling: Cold Plate sessions on two-phase liquid cooling, one with ZutaCore’s Shahar Belkin on reducing operational complexity through standardized two-phase cooling for AI infrastructure, and one with Wiwynn’s Yuzhou Lin on system-level challenges in moving from single-phase to two-phase cooling.