2026.09.23 | AC | PCB | Glass Cloth: Rubin Ultra & Q-Cloth

摘要

1. Rubin Ultra electronic fabric upgrades and Q-cloth demand surge;
2. Gen-2 LowDK supply-demand gap and expansion cadence;
3. Q-cloth supply dynamics and quartz yarn capacity bottleneck.

Below are the expert's viewpoints:

Could you provide a detailed overview of the specific selection schemes of high-end electronic fabric across different components in NVIDIA's Blackwell, Rubin, and Rubin ULTRA series servers?

In NVIDIA's Blackwell series, the Computer Tray and Switch Tray use M8-grade materials, corresponding to Gen-2 LowDK electronic fabric, while remaining components primarily use M6-grade materials, corresponding to Gen-1 LowDK electronic fabric. For the Rubin series, the Computer Tray, Switch Tray, and midplane will all adopt M8-grade Gen-2 LowDK materials, whereas the OAM board uses M7-grade materials, which likewise correspond to Gen-2 LowDK. As for the Rubin ULTRA series, the solution remains unfinalized; the orthogonal backplane currently leans toward an M9-grade hybrid lamination scheme, while other components like the CPX board are still undergoing material selection testing between M8 and M9 grades.

What is the current pricing structure across different tiers of high-end electronic fabric in the market, such as Gen-1 LowDK, Gen-2 LowDK, T-glass fabric, and Q-cloth?

Current market prices show that Gen-1 LowDK electronic fabric is priced between RMB 35 and 40 per meter. Gen-2 LowDK sees a substantial increase, fetching RMB 150 to 170 per meter. T-glass fabric is priced close to Gen-2 LowDK, or even slightly higher. Q-cloth, as the top-tier offering, is priced between RMB 250 and 300 per meter. Notably, Q-cloth produced via chemical synthesis by Japanese vendors commands more than RMB 300 per meter due to higher purity, which is significantly above domestic counterparts.

What priority principles guide the selection of electronic fabric materials for high-speed circuit boards? How does the industry define specialized materials such as M9 and PTFE?

The core selection rule is to prioritize the lowest-cost solution that meets performance requirements. A more expensive Q-cloth scheme is considered only when Gen-2 LowDK materials cannot satisfy extreme performance demands. By industry definition, M9 typically defaults to a scheme combining Q-cloth with carbon fiber. Meanwhile, although PTFE materials feature lower DK and DF values, they are excluded from the standard M-series material definition due to poor CTE performance.

What core manufacturing challenges does Q-cloth face?

The primary pain points in Q-cloth production lie in two areas. First, yields across the entire value chain remain low, particularly in PCB processing where yields fall below 50%, resulting in heavy processing losses. Second, production throughput is restricted: because quartz is brittle and hard, loom speeds must be reduced substantially, limiting single-unit monthly output on Toyota looms to just 5,000 to 6,000 meters, far below the roughly 20,000 meters achieved with standard electronic fabric. In terms of adoption timing, Q-cloth is still in small-batch supply. High-certainty volume demand comes predominantly from Rubin Ultra's orthogonal backplane, which reaches up to 78 layers. Potential future incremental use cases include 1.6T switches and compute/switch boards in the next-generation Feynman architecture, which are likely to adopt Q-cloth down the line.

What does the supply-demand landscape look like across the Q-cloth value chain? Who are the core suppliers, and how will future bargaining leverage and profit margins evolve?

On the supplier side, domestic vendor Feilihua was among the earliest movers. Among foreign suppliers, Japan's Shin-Etsu relies on chemical synthesis with monthly capacity of around 100,000 meters, though its expansion cadence is relatively slow; Asahi Kasei weaves fabric by sourcing quartz yarn externally from Dehong, leaving its capacity constrained by upstream yarn supply. Regarding bargaining power and margin trends, yarn makers hold stronger pricing leverage in the near term because the quartz yarn segment faces a longer capacity ramp cycle and remains the primary bottleneck. As capacity gradually comes online, leverage will normalize toward balance. Going forward, as more players enter the Q-cloth field, excess margins in weaving are expected to compress first, while yield gains will further lower industry costs overall. On the supply-demand outlook, multiple vendors originally in military and solar quartz have crossed into the quartz yarn arena. By 2028, quartz yarn and Q-cloth capacities are both projected to match market demand, bringing industry supply and demand into balance while stabilizing prices.

What is the current supply-demand situation for Gen-2 LowDK electronic fabric? What are the capacity roadmaps of leading industry players?

Currently, Gen-2 LowDK represents the largest supply-demand deficit across the entire industry. The supply gap is expected to peak in Q1 next year (2027), followed by marginal easing in the second half of next year. Looking at specific capacity plans, the participating company currently produces 300,000 to 400,000 meters per month, targets 600,000 meters per month by the end of this year (2026), plans to scale to 1 million meters per month in the first half of next year (2027), and aims for a final target of 2.5 million meters per month in 2028. Other leading players, including CPIC, Taishan Fiberglass, and Guangyuan, likewise plan to bring Gen-2 LowDK capacity online in 2028, with footprints ranging from 2 million to 5 million meters. Capacity from new entrants typically takes roughly one year to complete customer qualification.