虎嗅

Global AI giants are competing fiercely for a piece of fabric

原文:全球AI巨头疯抢一块布

Hello! I'm your financial analyst friend. Today, we're going to talk about an industry that may sound very technical, but it's actually closely related to the phones, computers in our hands, and even the future capabilities of AI—electronic fabric.

Recently, news about the rising prices of electronic fabric has gone viral. Many people might think, "It's just a piece of fabric; how could it cost so much?"

Don't worry; this "fabric" is anything but ordinary. It's the backbone of printed circuit boards (PCBs), which are the "nervous systems" of all electronic devices, from the iPhones in your hands to the powerful AI servers in data centers.

To help you fully understand the logic behind these price increases and predict when they might stop, I've broken down this in-depth report from the Economic Observer into five easy-to-understand sections. We'll avoid using jargon and focus on plain language and the core facts.

---

1. Review of the Phenomenon: How Severe Are These Price Increases? Why Is Everyone So Excited?

First, let's clarify the situation: This isn't a minor fluctuation; it's an "epic" surge that's even accelerating.

  • Prices Have Doubled or More:
  • The most commonly used 7628 specification of electronic fabric has gone from 3.8 yuan per meter at the beginning of the year to 11.8 yuan per meter in September.
  • The thinner, higher-end 1080 and 2116 specifications have seen even more dramatic price increases, rising by 280%-290%.
  • For comparison, the previous price increase (2020-2021) only reached 8.78 yuan per meter, so the current prices are more than 30% higher than that peak.
  • Downstream Companies Are Struggling:
  • JianTao, a leading manufacturer of copper-clad laminates (the precursor to PCBs), has issued price increases seven times this year, with a total increase of over 100%.
  • Companies that produce PCBs, such as ShenNan Circuit and JingWang Electronics, are complaining about high costs and squeezed profits.
  • Stock Market Response:
  • Related stocks, like HongHe Technology, International Composite Materials, and China Jushi, have seen their prices rise by 3 to 7 times this year.
  • Their profits have also increased significantly: HongHe Technology's net profit in the first half of the year increased by 334%, and China Jushi's by 74%.

In short, electronic fabric has become a form of "hard currency"—companies with inventory can charge higher prices, and even the big players can't get enough of it.

---

2. The Core Logic: Why Is There a Shortage? It's Not About Lack of Machines, but a Three-Factor Stranglehold

Many think the shortage is due to a lack of weaving machines, but that's just the surface. Industry expert Su WanYi points out that it's a combination of three factors: demand, raw materials, and equipment.

1. Demand: AI Servers Are Devouring Fabric

  • Dramatic Increase in Usage:
  • Traditional server PCBs had 8-14 layers, but modern AI servers (like those from NVIDIA) require 20-78 layers, each of which needs electronic fabric as a backbone.
  • Conclusion: An AI server uses 3 to 8 times as much electronic fabric as a traditional server.
  • Key Point: This is a permanent increase in demand, not a cyclical one. As long as AI continues to develop, this demand will persist.

2. Raw Materials: A Shortage of Electronic Yarn

  • Slow Capacity Expansion:
  • Producing electronic yarn requires building new kilns, which takes more than two years from planning to full production.
  • Big players like China Jushi and JianTao prioritize using their own yarn for their own production, reducing the amount available for sale.
  • Small and Medium-Sized Manufacturers Are Struggling:
  • Some small and medium-sized manufacturers are already out of stock, only able to supply 60% of the demand.

3. Equipment: Japanese Monopoly

  • No Alternatives:
  • High-end electronic fabric, especially thin types, requires jet looms made by Toyota Industries of Japan.
  • Waiting Times:
  • Orders for Toyota looms now will be delivered in 2029.

Difference with Domestic Machines:

  • The issue isn't speed but stability. Toyota looms maintain tension within ±0.1 grams, while domestic machines can only achieve ±0.3 grams. Domestic machines perform well for thick fabric but have lower yields for ultra-thin fabric. Fabric manufacturers need reliability, so they're reluctant to switch to domestic machines.

In short, AI has driven demand skyrocketing, raw materials can't keep up due to slow capacity expansion, and equipment is monopolized by Japan, leading to these price increases.

---

3. The Truth About Profits: Who Is Making Money? The Cruel Game of Choosing High-End Products

During these price increases, different companies have faced very different fates. The core logic is simple: Those that produce high-end, special fabrics make big profits; those that only produce ordinary fabrics suffer.

  • Profit Differences:
  • Ordinary Fabric (7628): 11.8 yuan per meter, used in phones and appliances, with fierce competition and low profits.
  • Special Fabrics (low-dielectric, quartz fabric, etc.): Used in AI servers, requiring fast signal transmission and low loss.
  • Low-dielectric second-generation fabric: 160 yuan per meter.
  • Quartz fabric: 250-400 yuan per meter.
  • Price Differences: Within the same category, prices can vary by more than 20 times!
  • Companies' Choices:
  • HongHe Technology: Cut back on low-profit ordinary fabric and focus on special fabrics. As a result, their special fabric revenue increased from 16% to 35%, and their gross margin soared from 31% to 59.5%.
  • Contrast: Changhai Shares (mainly producing ordinary fiberglass yarn) saw its net profit decline by 10% in the first half of the year.
  • JianTao Laminates: Its electronic fabric business profit was 1 billion Hong Kong dollars in the first half of the year, a year-on-year increase of 280%.

In short, the industry is undergoing a reshuffle. Big players and those capable of producing high-end fabrics are making substantial profits, while low-end capacity is being phased out.

---

4. Future Outlook: When Will the Trend Reverse? Don't Rush to Buy Low; 2028 Will Be the Real Test

If it's so profitable, why aren't new players entering the market? Why aren't prices falling?

  • Reasons for Entry Hesitation:
  • Past Lessons: After the 2021 price surge, many new players entered the market but faced price collapses in 2022, resulting in years of losses.
  • High Barriers:
  • State-Owned Enterprises: Their return on investment doesn't meet requirements, so they can't get approval.
  • Private Enterprises: The cost of core components (platinum-rhodium alloy masks) is too high, increasing the cost of building new kilns.
  • Existing Players: They're all focusing on expanding production of higher-profiting special fabrics, leaving no interest in ordinary fabrics.
  • Capacity Expansion Timeline:
  • 2026-2027: China Jushi and JianTao's expansion projects will start production, mainly by the end of 2027 to 2028.
  • Toyota Looms: Monthly production will increase to 200 units in the second quarter of 2027 and to 300 units by the end of 2027 or early 2028.
  • Domestic Looms: Production is expected to increase in the first half of 2027, but they still need to pass multiple tests for quality, stability, and certification before being widely adopted by big players.
  • Expert Opinion: 2028 Will Be the Real Test:
  • Su WanYi believes 2028 will be the true test of this trend.
  • Supply bottlenecks (raw materials, equipment) will remain until then.
  • Even with increased domestic production, competition will focus on factors like yarn composition, coating processes, and customer certification, not just equipment.

In short, supply will remain tight in the short term (2026-2027), and the price trend is unlikely to change. The true balance of supply and demand may not be achieved until 2028. Entering the market now is risky, but the industry's prospects are indeed very promising.

---

5. Potential Risks and Variables: PTFE as a Replacement? Overseas Capacity Withdrawal?

Finally, let's consider any potential "black swans" or new variables that could change the situation.

  • PTFE as a Replacement?
  • Concern: PTFE is another high-frequency material, but even if it's used in AI servers, it won't completely replace fiberglass electronic fabric.
  • Reality: NVIDIA's next-generation servers (Rubin Ultra) have tested PTFE, but mass production won't start until the second half of 2027, and it will be used in combination with fiberglass, not as a full replacement. So, this concern is exaggerated; demand for electronic fabric will still be strong in the short term.
  • Overseas Capacity Withdrawal and China's Share Expansion:
  • Giants like Owens Corning (US) and Electric Glass (Japan) are reducing or closing their ordinary electronic fabric production lines, shifting to high-end products or exiting the market.
  • Impact: Chinese manufacturers are receiving a large number of overseas orders, with China Jushi exporting over 100,000 tons of fiberglass in July and August, a record high.
  • Effect: China's share of the global ordinary electronic fabric market will increase, enhancing its influence.
  • Low-End Market Issues:
  • Some small manufacturers use inferior materials, leading to poor-quality products that can cause problems in PCBs.
  • Consequences: These issues can result in significant losses for small manufacturers, accelerating the consolidation of the industry towards leading players.
  • Price Predictions:
  • Industry Estimates: By the end of the year, the price of 7628 fabric from big players could exceed 15 yuan per meter, and protective fabrics (for specialized applications) could exceed 20 yuan per meter.
  • Current Market Signals: Secondary market prices (14-15 yuan) already exceed big player quotes (10 yuan), indicating extreme scarcity, suggesting prices may rise beyond expectations.

---

Summary for Everyone

1. For Investors: The electronic fabric industry is in a phase of high prosperity, high barriers, and high profits. The key points to watch are the release of capacity for special electronic fabrics and the progress of domestic loom production. However, stock prices already reflect some of these expectations, and 2028 will be the true test of supply and demand. Short-term fluctuations are likely.

2. For Those in the Electronics Industry: Prepare for rising costs. Downstream PCB and device manufacturers may pass on the costs through price hikes or design optimizations. Ensuring supply is more important than maintaining prices; those who can't supply will be at a disadvantage.

3. For Curious Onlookers: Remember, the seemingly insignificant fabric in your phone is undergoing an unprecedented revaluation due to the explosion of AI. It's no longer just an industrial product but a critical component of the computing era.

Final Reminder: The market is risky, and investments should be made with caution. This analysis is based on public news and logical reasoning and does not constitute direct investment advice.