Wiwynn is spearheading the deployment of the NVIDIA Vera Rubin NVL72 platform, serving as a pilot build partner for the liquid-cooled, rack-scale architecture that combines 72 Rubin GPUs with 36 Vera CPUs. Simultaneously, the company is expanding its partnership with AMD to scale the Helios rack solution, which integrates Instinct MI450 GPUs and 6th-gen EPYC processors within the Open Rack Wide standard to support large-scale AI training environments.

To manage the thermal demands of these systems, Wiwynn is testing diamond-based microchannel cooling, which utilizes the superior thermal conductivity of diamond to accelerate heat transfer directly from the silicon. The firm also introduced a 6kW double-sided cold plate design that leverages vertical power delivery to reduce energy loss by 80%. These mechanical advancements are complemented by a new co-packaged optics (CPO) ecosystem, developed with partners like Ayar Labs and Corning, to address network bottlenecks through integrated fiber management and optical engine telemetry.