[Red Envelope] The Truth About the Price Hike of Electronic Cloth: Japan Shakes, Global Cloth Shortage, NVIDIA and Qualcomm's Lifelines, Stuck by a Loom!

The Electronic Fabric Price Surge Storm: “Material Bottleneck” Driven to the Wall by AI Computing Power [Taoguba]

Deep Review of AI Computing Material Chain | May 2026

1. Core Judgment: This is not a price increase, but a re-pricing of materials by computing power hegemony

Electronic fabric (electronic-grade glass fiber fabric) is breaking away from the traditional pricing framework of cyclical commodities, entering a super shortage cycle driven by violent demand for AI computing power. Since 2026, the industry has completed four consecutive price hikes, with monthly adjustments becoming the norm. The price of 7628 ordinary fabric soared from 3.5 yuan/m at the beginning of 2025 to 6.3-6.5 yuan/m in April 2026, a cumulative increase of over 80%; Honghe Technology’s average price in the first quarter was even violently surged from 4.51 yuan/m to 9.78 yuan/m, a year-on-year jump of 116.85%. High-end varieties like Low-Dk second-generation fabrics are quoted at 160 yuan/m, and Q-Fabric (quartz electronic fabric) reaches 200-400 yuan/m, 50 to 100 times the price of ordinary fabrics.
The core conclusion is extremely clear: this round of electronic fabric price hikes is not simply a mismatch of supply and demand, but a structural material hegemony contest triggered by the AI computing power expansion rate surpassing the physical limits of high-end material capacity. Downstream cloud giants are highly receptive to price increases, with a complete price increase cycle—every 50% rise in electronic fabric leads to a 30-40% increase in copper-clad laminates, 15-20% in PCB prices, and only a 5-10% pass-through to the entire AI server system. A tiny “fabric” is now moving the cost curve of the entire trillion-dollar computing industry chain.

2. Three hard data points that must be directly faced

First data point: single-machine usage skyrocketed from 3 meters to 50 meters

Traditional server PCBs have about 10 layers, with 3-5 meters of electronic fabric per unit. Nvidia’s GB200 jumped to 14-16 layers, with 12-15 meters per machine; GB300 reaches 16-38 layers, with 18-24 meters per machine; the PCB layers of Rubin Ultra architecture, which will be mass-produced in the second half of 2026, are expected to exceed 78 layers, with single-machine usage reaching 30-50 meters. This means AI server electronic fabric usage is 8 to 10 times that of traditional servers, and it’s not ordinary fabric but high-end Low-Dk and Q fabrics priced at dozens or even hundreds of times higher. The demand curve is not linear but exponential.

Second data point: gross profit margin of 55.65%

Honghe Technology’s Q1 2026 report revealed a shocking set of data: despite the extreme cost pressure from raw materials like electronic yarn soaring 206.55% to 53.83 yuan/kg, the company’s gross margin not only was not eroded but increased by 98.32% year-on-year to 55.65%, with a net profit margin of 31.72%. What does this mean? It indicates that high-end electronic fabrics have extremely strong downstream pass-through and premium pricing power—downstream customers are willing to accept price hikes to lock in already tight capacity. Scarcity directly translates into pricing power.

Third data point: 18 to 24 months

This is the delivery cycle of Japan’s Toyota high-end jet looms, with current orders scheduled for 2028-2030. These looms are core equipment for producing ultra-thin and extremely thin low-dielectric and low-expansion fabrics. The global annual capacity is only about 2,000 units, almost monopolized by Toyota Japan. Domestic looms have a yield rate of about 85%, significantly lower than Japan’s over 95%, and cannot form effective substitutes in the short term. Even if midstream manufacturers face a flood of orders, they cannot expand production within a year by purchasing new equipment. Supply elasticity is physically locked.

3. The “triple strangulation” on the supply side: why this shortage is so desperate

First strangulation: equipment rigidity deadlock

High-end electronic fabrics rely on Japanese Toyota JAT910 and a few other jet looms, with lead times of 18-24 months. This means that the capacity that can be released in 2026 was actually locked in by the end of 2024 or even earlier. No matter how hot downstream demand is or how eager midstream manufacturers are, without looms, fabrics cannot be produced. This is a physical ceiling that cannot be broken—no short-term elasticity exists.

Second strangulation: high-end squeezing low-end

Because high-end fabrics like Low-Dk and Low-CTE can achieve gross margins of 40-60%, while ordinary E-glass fabrics are only 20-30%, manufacturers with high-end looms naturally prioritize high-value-added specialty fabrics using their existing capacity, skilled labor, and kiln resources. This directly squeezes supply of ordinary fabrics used in consumer electronics and white appliances. The result is high-end shortages and simultaneous price hikes of ordinary fabrics due to production shifts, creating a peculiar cyclical pattern of price increases across all categories.

Third strangulation: Japan earthquake shatters global safety stock

On April 20, 2026, a 7.7 magnitude earthquake off the coast of Sanriku, northeastern Japan, severely damaged Nitto Boseki’s core factory in Shirakawa, Fukushima. This company nearly monopolizes over 90% of the global T-Glass (Low-CTE) high-end electronic fabric and over 70% of high-end Q fabric. The industry’s safety stock is only 2-4 weeks; Nitto Boseki’s shutdown lasts 20-25 days, with full recovery taking 3-6 months, directly causing a supply rupture in high-end supply chains. Global high-end electronic fabric lead times extend beyond 9 months, forcing giants like Apple, Nvidia, and Qualcomm to scramble for fabrics across borders, shifting large orders urgently to Chinese mainland manufacturers. This is not a simple supply-demand imbalance but a collapse of the material safety bottom line.

4. Scarcity premium: whoever masters the “thin” capability holds the pricing power

The scarcity of electronic fabric is fundamentally a competition over “thinness” technology barriers. The thinner the product, the more geometrically difficult the weaving, and the higher the gross margin: thin fabrics have about 26.1% gross margin, ultra-thin fabrics can exceed 53.9%. In this dimension, a clear echelon of scarcity has formed domestically—

Honghe Technology is the only global manufacturer to achieve mass production of 4μm ultra-thin electronic fabric, with about 50% global market share for ultra-thin/very-thin fabrics. In the sub-16μm ultra-thin electronic fabric field, only Nitto Boseki and Honghe Technology have stable supply capacity. After the Japan earthquake, Honghe became the only domestic and second globally mass producer of T-Glass, taking on urgent orders from giants like Apple, Nvidia, and AMD.

Philhua is the only domestic enterprise capable of full-chain independent control of “high-purity quartz sand—quartz fiber—quartz fabric,” with Q fabric gross margins exceeding 60%, certified for Nvidia’s Rubin architecture. Q fabric is the most extreme dielectric electronic fabric category, with only Japanese and Philhua capable of mass production globally.

CNMC Technology is the only domestic enterprise to achieve mass production of Low-Dk first/second/third-generation, Low-CTE, and Q fabrics across the full spectrum, with about 35% market share of second-generation low dielectric fabrics domestically. Its core scarcity lies in “one-stop substitution”—when customers switch from Nitto Boseki, only CNMC can meet all their product needs.

5. Domestic substitution: from passive chasing to proactive夺取 (seizing control) — a historic window

In 2024, domestic high-end electronic fabric self-sufficiency was only 35-40%; by 2025, it rapidly increased to 50-65%; and is expected to break through 80% in 2026. Behind these figures is not just capacity expansion but a natural result of technological breakthroughs—Honghe’s ultra-thin T fabric has been certified by giants like Nvidia, Google, AMD; CNMC’s second-generation low dielectric fabric has entered the supply chains of global top-tier PCB manufacturers like Taiyo Electronics and Shengyi Technology; Philhua has broken Japan’s Shin-Etsu Chemical and Nitto Boseki’s monopoly in M9-grade Q fabric.

In the context of Nitto Boseki’s capacity being hindered by the earthquake and new capacity release being extremely slow (expected only around 2027), global giants like Apple, Nvidia, AMD, and Qualcomm are desperately shifting orders to Chinese mainland manufacturers. This is a historic client-switching opportunity, and domestic high-end electronic fabric is poised to transition from “backup” to “main supplier” during this window.

Feasibility judgment of alternatives: LCP (liquid crystal polymer) or MPI (modified polyimide) organic films can only be used in 5G RF antennas, flexible circuits, and other flexible scenarios, unable to meet the mechanical rigidity and dimensional stability requirements of high-layer motherboards. In the foreseeable 5-10 years, the core position of high-end electronic fabric cannot be replaced.

6. Core targets: four types of scarcity, four trading logics

From the perspective of scarcity, the four leading companies each have their irreplaceable core barriers, corresponding to four different trading logics—

China Baosteel (600176): The ultimate controller of scale scarcity
In 2025, electronic fabric sales hit 1.06B meters, a record high, with about 23% global market share; Huai’an Phase I started in March 2026, and after reaching full capacity, market share will rise to 28%. The company’s Q1 2026 gross margin was 39.6% (+9 percentage points YoY), net profit margin 24.0% (+8 percentage points YoY), confirming the “cost relatively fixed, most of the price increase translates into profit” strong elasticity logic. The 2026 forward PE is about 19 times, with the best valuation safety margin among core targets.
Key logic: volume-price growth + scale moat + controllable valuation

Honghe Technology (603256): The purest carrier of technical scarcity
Electronic fabric revenue accounts for up to 95.55%, making it the most sensitive pure target to electronic fabric price hikes in A-shares. The only global mass producer of 4μm ultra-thin fabric, with about 50% global market share. In the sub-16μm ultra-thin fabric field, only Nitto Boseki and Honghe Technology have stable supply capacity. After the Japan earthquake, Honghe became the only domestic and second globally mass producer of T-Glass, taking on urgent orders from giants like Apple, Nvidia, AMD.

Philhua (300395): The only domestic full-chain independent control of “high-purity quartz sand—quartz fiber—quartz fabric”
With Q fabric gross margins over 60%, certified for Nvidia’s Rubin architecture, and locked-in capacity with top clients in 2026. It is the only enterprise in China with full coverage of the Q fabric industry chain.

CMT (002080): The rare all-in-one replacement solution
The only domestic enterprise achieving mass production of Low-Dk first/second/third-generation, Low-CTE, and Q fabrics across the full spectrum, with about 35% market share of second-generation low dielectric fabrics. Its core scarcity is “one-stop substitution”—when customers switch from Nitto Boseki, only CNMC can meet all their product needs.

7. Endgame deduction: where is the price hike cycle heading

Short-term: The price increase logic is far from over
The 18-24 month delivery cycle of high-end looms locks the supply ceiling, while AI server demand for electronic fabric increases 5-10 times, restructuring the demand curve. As long as the core inequality “high-end demand growth > new capacity release” holds, monthly price adjustments will continue. Ordinary fabrics reflect industry inventory lows and capacity squeeze; high-end fabrics reflect AI material upgrades and domestic substitution windows—both lines run in parallel.

Mid-term: 2027 will be a key divergence point
Industry capacity under construction will be concentrated for release around 2027, putting downward pressure on ordinary electronic fabric prices. But high-end Low-Dk, Low-CTE, and Q fabrics, due to long certification cycles, high process difficulty, and high customer switching costs, are expected to maintain significantly stronger prosperity than ordinary fabrics. The stock selection criteria will shift from “whether there is electronic fabric” to three hard indicators: actual capacity and sales volume, high-end product certification and batch delivery ability, and whether price hikes can translate into profits rather than being dragged down by other businesses.

Risk warnings
AI capital expenditure underperforms expectations; industry capacity release in 2027 causes ordinary fabric prices to fall; high-end product certification delays; some high-elasticity targets face trading congestion risks.

8. One-sentence summary

The essence of the electronic fabric price hike is that the expansion speed of AI computing power outpaces the physical limits of high-end material capacity. In this super cycle driven by demand violence, supply rigidity, and accelerated domestic substitution, what is truly scarce is never just “fabric,” but the system barriers formed by equipment, processes, and customer certification that weave that fabric. Choosing the leading companies with scarce barriers is far more important than simply betting on price increases.