Nvidia Tests Glass-Free PCB Materials for Next-Generation Rubin Ultra AI Servers
Nvidia is reportedly exploring new printed circuit board material systems for its upcoming AI server hardware, with a focus on glass-free PCB designs that could improve performance, signal quality, and reliability in extremely high-speed computing environments.
The company is said to be evaluating two advanced resin-based options: polytetrafluoroethylene, commonly known as PTFE, and hydrocarbon resin, often referred to as HC. These materials are being considered for possible use in the Rubin Ultra NVL576 Switch Tray, a critical component expected to play a major role in Nvidia’s next wave of AI server infrastructure.
As AI servers become more powerful, the demands placed on internal networking, switching, and interconnect systems continue to rise. Traditional PCB materials can face limitations when handling very high frequencies and dense signal routing. By moving toward glass-free PCB material systems, Nvidia may be aiming to reduce signal loss, improve electrical consistency, and support the faster data transfer required by next-generation AI workloads.
PTFE is already known for its low dielectric loss and strong high-frequency performance, making it attractive for advanced networking and computing applications. Hydrocarbon resin systems, meanwhile, can offer a balance of performance, manufacturability, and cost efficiency. Nvidia’s evaluation of both options suggests the company is looking closely at how PCB material choices could affect the scalability and stability of future AI platforms.
The Rubin Ultra NVL576 Switch Tray is expected to be part of a high-density AI server architecture designed to connect large numbers of accelerators with fast, efficient communication. In systems built for artificial intelligence training and inference, even small improvements in signal integrity can have a meaningful impact on overall performance and power efficiency.
Glass-free PCB technology could also help address some of the mechanical and electrical challenges associated with conventional fiberglass-reinforced boards. Removing glass from the PCB structure may reduce variations in signal propagation, which becomes increasingly important as server platforms push toward higher bandwidths and tighter design tolerances.
While Nvidia has not publicly confirmed final material choices for the Rubin Ultra NVL576 Switch Tray, the testing of PTFE and HC resin systems shows how much attention is being placed on every layer of AI server design. The race to build faster AI infrastructure is no longer only about GPUs and accelerators; it also depends on advanced packaging, power delivery, cooling, interconnects, and PCB innovation.
If Nvidia adopts glass-free PCB materials in future Rubin Ultra systems, it could mark another step toward more advanced AI server platforms built for massive-scale data center workloads. As demand for AI computing continues to grow, material-level improvements like these may become increasingly important in helping data centers achieve better speed, efficiency, and reliability.
