A Comprehensive Explanation of the Three Generations of Electronic Cloth Technology: The Core Material for AI Computing Power

The AI ​​computing power race is intensifying, with core hardware such as GPUs, switches, and optical modules frequently becoming the focus of the market. However, few people pay attention to the fact that the upper limit of high-end computing hardware performance is determined by fiberglass electronic cloth. As a key upstream link in the PCB industry chain, the technological generation of electronic cloth directly determines high-speed signal transmission capabilities and is the core lifeline of AI servers and high-end chip packaging.

 

I. Core Positioning in the Industry Chain: The Dual Core Value of Electronic Cloth

 

To understand electronic cloth technology, it's essential to understand its complete industry chain: starting with quartz sand processing, through electronic yarn drawing and weaving to form electronic cloth, then composited with copper foil and resin to create copper-clad laminate (CCL), which is ultimately used in PCB printed circuit boards to power terminal computing hardware such as GPUs, switches, and optical modules.

 

Of the three core materials of copper-clad laminate-electronic cloth, copper foil, and resin-copper foil is responsible for conductivity, resin for adhesion, but the electronic cloth is the key to determining the upper limit of high-speed performance.

 

It undertakes two core functions: first, structural support, acting as the skeleton of the copper-clad laminate, withstanding high and low temperature changes and long-term mechanical stress, ensuring the stability of the hardware structure. Secondly, signal control hinges on two key indicators: dielectric constant (DK) and dielectric loss (DF). Simply put, a lower DK results in less signal transmission delay, and a lower DF results in less signal transmission loss. The industry's M8 and M9 material grades are essentially a selection process based on the DK and DF performance of electronic fabrics.

 

Second and Third Generation Electronic Fabric Technology: From Mature Applications to Future Breakthroughs

 

First Generation: E-Class First-Generation Fabric, Mature Production Capacity Squeezed Out

 

First-generation electronic fabrics are the most mature products in the industry, with DK values ​​maintained in the 5.8-6.5 range. Signal loss is relatively high, and they are mainly used in traditional servers, ordinary PCBs, and other conventional scenarios. Before 2025, most AI servers on the market still use first-generation fabric solutions. Domestic companies such as Taiwan Glass and Honghe Technology have achieved complete independent mass production, with mature technology and production capacity.

 

However, it is noteworthy that the current price increase of first-generation fabrics is not driven by a surge in demand, but rather by the industry's upgrade to high-end products. Many production lines have shifted to producing high-value-added second and third-generation products, leading to the squeezing out of ordinary first-generation fabric production capacity and creating a supply-demand imbalance. Second Generation: Low DK Second-Generation Fabric, a Core Necessity for AI Computing Power

 

Second-generation electronic fabric is currently the core battleground for AI servers. Through technological optimization, the DK value has been reduced to the 4.2-4.8 range, resulting in a significant decrease in DF (Distribution Deficiency) and a qualitative leap in signal transmission performance. This product is currently widely used in AI server motherboards, high-end switches, GPU servers, and high-end computing hardware such as NVIDIA GB200/GB300, Google TPU systems, and Amazon AI data centers, all of which rely heavily on second-generation fabric.

 

Currently, second-generation fabric is in structural shortage, not due to short-term supply and demand fluctuations, but rather a global supply shortage caused by technological barriers. Global technology leaders include Nittobo and Asahi Kasei, while domestic companies such as Sinoma Science & Technology, International Composite Materials, and Honghe Technology are rapidly catching up. In the next two years, second-generation fabric will remain the most scarce core material in the AI ​​computing power field.

 

Third Generation: Q-Cloth Quartz Fabric, the Ultimate Solution for the 1.6T Era

 

Third-generation Q-cloth quartz fabric is a top-tier material for next-generation computing power and a watershed moment in the industry's technology. Its core innovation lies in replacing glass fiber with quartz fiber, resulting in a significant performance boost. The coefficient of thermal expansion (DK) can be as low as 3.5-3.8, and the density of thermal expansion (DF) controlled below 0.01, perfectly suited for ultra-high-speed scenarios such as 1.6T optical modules, Rubin architecture, and high-end CPU packaging.

 

Currently, Q-fiber electronic cloth has not yet entered a period of full-scale development; it is in a transitional phase from verification completion and small-batch trial production to gradual mass production. The technological barriers are extremely high, and very few companies globally can reliably supply it. Overseas, Nittobo and New Moon Chemical are the main players, while domestically, only a few companies such as Feilihua and Quartz Shares have achieved mass production. This is the core battleground for future competition in electronic cloth technology.

 

III. Hidden Key Route: Low CTE Electronic Cloth, a Necessity in the Packaging Field

 

Besides the third-generation upgrade, Low CTE electronic cloth is an important sub-route that cannot be ignored. It does not belong to a completely new generation but is an optimization and upgrade based on second-generation cloth, with the core being the reduction of the coefficient of thermal expansion (CTE).

 

800G and above optical modules commonly use silicon photonics chips. Silicon chips have an extremely low coefficient of thermal expansion. If the PCB material's coefficient of thermal expansion is too high, the difference in thermal expansion and contraction can crack the encapsulation layer, leading to hardware failure. Therefore, Low CTE electronic fabric has become an essential material for GPUs, high-end AI chips, and silicon photonics packaging. HBM (Hybrid Machine Modeling) is also in the verification stage, with considerable future growth potential.

 

The core technological barrier of this route lies in the necessity of self-production of electronic yarn. Purchased electronic yarn cannot meet the precise performance requirements. Mastering independent production capacity of electronic yarn is the core moat for maintaining the Low CTE electronic fabric technology.

 

Electronic fabric may seem niche, but it is a critical bottleneck in the AI ​​computing power industry chain. The three generations of technology iteration clearly correspond to the upgrade rhythm of computing power hardware. From the production capacity mismatch of the first-generation fabric, to the structural shortage of the second-generation fabric, and then to the technological breakthroughs of the third-generation Q-fabric fabric, domestic companies are accelerating the breakthrough of overseas technological monopolies. As AI computing power continues to upgrade, the technological value and market demand of electronic fabric will be further released. This breakthrough battle in the field of materials will also become an important microcosm of the rise of domestic high-end manufacturing.