The hardest piece of glass cloth to buy in 2026 is the plainest one in the catalogue.
Style 7628 is the commodity E-glass weave, the default reinforcement under ordinary FR-4 copper-clad laminate. It is not a specialty product and nobody markets it as one. At the end of the third quarter of 2025 it was quoted around CNY 4.15 per metre. By the middle of August 2026 the average was around CNY 10 per metre, and by trade-press accounts some laminate suppliers had begun allocating material rather than supplying it on open order.
What makes that worth a buyer's attention is the mechanism rather than the number. Demand for 7628 did not have to rise for its supply to contract. The margin on AI-grade specialty electronic cloth is far higher than on ordinary electronic cloth, so leading producers converted traditional electronic-cloth lines to specialty production — and because the high-density thin cloth they converted to weaves slowly, each loom that moves yields far fewer metres, so the specialty demand absorbs more looms than its metreage suggests. Capacity left the ordinary weave twice over: once in the lines that moved, again in how many lines the move takes. A shortage built from the supply side behaves differently from one built by demand — which is why a buyer watching 7628 demand alone would not have seen it coming.
This article is about that supply picture. Whether a board needs the high-CTI grade in the first place is a separate question decided by the standard the board is certified against, and we work through that in what CTI 600 actually buys on a copper clad laminate.
Why the Plainest Weave Became the Scarce One
It helps to be precise about what is short, because the adjacent styles are not substitutes for it.
From BGF Industries' published technical data book, 7628 is a plain weave with a count of 44 × 32 per inch (warp × fill), ECG 75 1/0 yarn in both directions, 206 g/m² and 0.18 mm thick. One style number away, 7629 is a plain weave at 44 × 34, 211 g/m² and 0.19 mm — a different cloth by the numbers that decide resin content and dielectric thickness. The data book attaches its own caveat to every row: these are typical greige-fabric values (unfinished, before heat or chemical treatment), offered as a guide and not as a specification. Use them to understand what you are buying, not to write a purchase requirement.
ECG 75 is the same yarn that Chinese price reporting calls G75 electronic yarn, which is why the yarn price sits beside the cloth price in the table below: it is the input one step upstream, reported in the same articles. Over the same months the two series rose in the same direction — +61% for the yarn, +78% for the cloth — which is what the table shows and all it shows.
The conversion is documented rather than inferred. A Chinese financial daily reported in June 2026, citing Prismark and several brokerage research notes, that because the margin on AI-grade specialty electronic cloth is far higher than on ordinary electronic cloth, leading producers moved traditional electronic-cloth lines to specialty production; the same reporting describes high-end thin cloth as produced under a double squeeze of high density and low loom speed, with output per loom falling sharply on the converted looms, and says in its own words that the high-margin specialty cloth takes loom capacity first and ordinary cloth supply contracts passively.
Neither half of that says when it reverses. The report does not say whether converted lines come back when the margin gap narrows, and we do not speculate; what it does put a number on is the other route to relief — a new line takes 18 to 24 months to build — and the outlook section takes that up.
The Same Conversion Shows Up in Copper Foil
A single material's price chart is a weak basis for a procurement decision. The same structure in a second, independently reported material is a stronger one.
A metals data provider reported in July 2026 a notable supply gap in traditional electronic-circuit copper foil "due to conversion toward high-end categories" — the same explanation as the glass-cloth story, from a different source about a different material. Utilisation across the copper foil industry as a whole, battery foil included, rose from 88.56% at the start of the year to above 91% by the end of the second quarter. The processing fee for 18 µm HTE foil, the circuit-board grade, rose by CNY 3,000 to CNY 21,000 per tonne in late June.
So two of the three main laminate inputs are short for the same reason, documented separately. If you are modelling this as an AI-demand story that passes when AI server builds normalise, the second data point says the same thing about the mechanism: what moved was where the capacity points, not how much of it exists. Whether it points back when AI-server margins narrow is not something either source addresses.
On the third input, copper itself, the record is thinner than it looks. The LME copper cash price exceeded USD 13,000 per tonne for the first time in early January 2026 and through mid-October has stayed in the USD 14,000s according to trade reporting. We could not reach the exchange's own price pages — HTTP 403 — so every figure here is secondary and we assert no particular day's close. More importantly, we could not establish what share of a laminate's cost is copper foil or glass cloth. We therefore do not convert a copper move into a laminate move, and anyone who offers you that arithmetic should be asked for the cost breakdown before you accept the conclusion.
The Prices, and What a Percentage Without Its Baseline Hides
The June figures come from the Chinese financial daily cited above, the August figures from Chinese financial media, the October figures from a financial newswire citing a brokerage note of 9 October. A separate industry-research report, drawing on listed-company annual reports and the Chinese fibreglass and electronic-fibreglass industry associations, puts 7628 ex-works at about CNY 3.5/m in September 2025 and about CNY 6/m in May 2026 — around +70% on its own baseline.
Read those two sources together and the problem appears. The September 2025 baseline is CNY 3.5, 3.8 or 4.15 per metre depending on who you ask, the end points run from May to mid-August 2026, and the increase is quoted anywhere from about +70% to about +166% — all arithmetically correct, none comparable with another without both dates and both values. A percentage without its baseline — or its end date — should not be carried into a budget or a negotiation. The useful reply to one is not to argue the number but to ask for what makes it mean something: the baseline date and value, the end date, the grade and glass style, and ex-works or delivered.
Dollar-denominated 7628 series circulate in PCB trade blogs and do not reconcile with the yuan series — the exchange rates they imply differ from each other and from the market rate — so we use neither those series nor the forward projections attached to them.
Note also that the styles are not one market. Each of 2116 and 1080 ended June above CNY 9 per metre against CNY 7.40 for 7628, and in October each took CNY 2.5 against 7628's CNY 1.5. A laminate quotation that does not say which glass style it was built on has left out a variable that moved differently from the headline.
Laminate price notices. By trade-press accounts, a Japanese laminate maker issued a notice in early 2026 covering all copper-clad laminates and prepregs at about 30% over current prices, citing the rise in copper foil and glass cloth together with labour and freight. A market-research firm reported in May 2026 that two Japanese suppliers had raised prices across their lines by more than 30%. Chinese trade reporting describes a major Chinese laminate maker issuing multiple increase notices during 2026, but the counts and cumulative percentages we found come from secondary blogs only, so we do not quote them. On the first-half increase overall, one trade publication reports high-end materials up 10–15% against 2025 with some basic materials up 20–30%, while PCB blogs report 20% to over 40% by grade. Those are not the same claim, and we do not reconcile them into one figure.
The One Number a Korean Buyer Will Be Shown
Korea's imported copper-clad laminate averaged USD 20,728 per tonne in March 2026 against USD 11,880 a year earlier — up 74.5%, and the first time above USD 20,000 per tonne since the series began in 2000.
This is the figure most likely to arrive in a Korean buyer's inbox as evidence, and it needs its caveat every time. It is an HS-code-level aggregate across all grades, and the cause the reporting itself identifies is demand for AI semiconductor and AI server laminates, where procurement concentrates on high-layer-count, high-speed grades. The 74.5% is not the price history of high-CTI FR-4, or of any other individual grade. We found no price series specific to CTI 600-class laminate in any source, and we did not reproduce the customs figures from primary trade statistics ourselves.
Lead Times, and the Shift to Allocation
A trade publication citing a German PCB manufacturer puts standard FR-4 laminate at about four weeks, against "a few days" before, and high-performance grades at up to 140 days. Other reports give two to three weeks moving to six to eight weeks for standard FR-4, and 20 weeks or more for high-specification material; a Korean industry comment in the customs reporting above quotes at least six months for high-specification laminate. We treat that spread as reporting disagreement rather than pick a number from it, and where high-CTI automotive grades fall in the range we could not determine.
The structural change matters more than the weeks. By trade-press accounts — one in March, one in July 2026 — some laminate suppliers have introduced allocation quotas and purchasing restrictions, with last year's purchasing volume the commonest basis. In an allocation market the first question is no longer the price but whether you are on the list, which changes what a second source is for: a qualified alternative stops being a cost lever and becomes an availability one. In our reading it also moves qualification earlier: a grade you have not qualified is not an option on the day you need it.
Outlook: What We Will and Will Not Say
New standard electronic-cloth capacity is what resolves this, and the timing evidence is indirect.
A new line takes 18 to 24 months to build, per the Chinese financial daily's June report. An industry-research report describes Chinese fibreglass producers' electronic-glass and electronic-cloth projects as working toward commissioning in 2027, with at least one furnace already lit in 2026. A brokerage note carried in October 2026 expects tension to ease "as domestic loom technology advances and new capacity is gradually released" — without naming a date. And the looms are themselves a bottleneck: Chinese reporting in August 2026 said orders for electronic-cloth looms are booked out to 2030.
"After 2027" is therefore a defensible shape for the relief. We do not have a grounded view on when the squeeze ends, and the specific mid-to-late-2027 dates circulating in PCB blogs come from sources that cite nothing.
One correction, because it is widely repeated and load-bearing in most published accounts of this shortage. The glass-cloth expansion most often invoked as the reason relief cannot arrive before 2027 is a Japanese weaver's T-glass project. T-glass is a specialty cloth for AI and GPU substrates. It is not standard E-glass 7628, and the two sit in different markets with different customers. Whatever that project's schedule, it is not the supply that relieves the ordinary weave; that has to come from Chinese E-glass capacity, on the evidence above.
And we publish no price forecast. Every forward number we found was either an unsourced blog projection or a short-horizon brokerage estimate, and neither belongs in someone else's budget.
What a Buyer Can Do on Monday
None of the above is actionable as a market view. It is actionable as a set of questions, most of which can go out in one round.
- Ask whether the grade is allocated before you ask what it costs. Allocation makes list position the binding constraint. If the answer is yes, the next questions are what the allocation is based on — last year's volume is the basis most often reported — and what last year's volume with that supplier buys you this year.
- Refuse a percentage that arrives without its baseline and its end date. Ask for the baseline date, the baseline value, the end date, the grade and glass style, and ex-works or delivered. On 7628 alone the published baselines for the same month run from CNY 3.5 to CNY 4.15 per metre, and the published increases from about +70% to about +166% — figures that differ in their end month as well, so neither the baseline nor the end date can be left out.
- Ask for lead time by grade and glass style, not the figure the mill gives for everything. The public record spans about four weeks to 140 days by grade, and we could not place high-CTI automotive grades in that range. Your supplier can.
- Keep the customs average out of your own item's price history. When the +74.5% import figure is quoted at you, ask which HS code it covers and which grades sit inside it. It is a real number about a basket that is probably not yours.
- Ask for a cost breakdown before accepting an input-driven increase. A notice citing copper foil and glass cloth is credible on its face — one such notice says exactly that — but the step from an input price to a laminate price needs the share that input holds in the cost, and that is a number the mill has and you do not.
- Decide the inventory question deliberately rather than by default. Buffering against an allocation market is an option with a carrying cost, and inventory buffering and the cash conversion cycle puts that cost at 20–25% a year and sets out who ends up absorbing the risk.
- The rest of the round is the specification, and that is the other article's subject. Naming the glass style and resin content rather than the grade, asking the cut size and minimum order, and establishing which document fixed the grade before accepting a substitution — what CTI 600 actually buys covers each, and the Other Materials page sets out what we ask for ourselves before answering on feasibility.
Frequently Asked Questions
Why is standard 7628 glass cloth short when it is the most ordinary weave there is?
Because capacity left it. Chinese reporting documents that leading producers converted traditional electronic-cloth lines to higher-margin AI-grade specialty cloth, and that the high-density thin cloth those lines moved to weaves slowly, so each converted loom yields fewer metres and the specialty demand takes more looms. The same conversion-driven gap appears independently in traditional electronic-circuit copper foil. Demand for 7628 did not have to grow for its supply to contract — and new lines take 18 to 24 months to build.
How much has 7628 risen in 2026?
That depends on the baseline, which is why a single percentage is not usable. One source chain has it at CNY 4.15/m at the end of Q3 2025, CNY 7.40/m in June 2026, around CNY 10.1/m on average by mid-August, and up another CNY 1.5/m by 9 October. A separate industry-research report starts from about CNY 3.5/m in September 2025 and reaches about CNY 6/m in May 2026. Published baselines for the same month differ — 3.5, 3.8 and 4.15 — and the increases quoted, from about +70% to about +166%, end in different months as well. Ask for the baseline date and value, and the end date, with any figure you are given.
Does the Japanese glass-cloth expansion fix the 7628 shortage?
No, and the claim that it explains the timing is a misattribution. The expansion usually cited is a Japanese weaver's T-glass project. T-glass is a specialty cloth for AI and GPU substrates, not standard E-glass 7628. Relief for the ordinary weave has to come from Chinese E-glass electronic-cloth capacity, where the projects described publicly point at commissioning in 2027.
When does the squeeze end?
We do not know, and we publish no date and no price forecast. "After 2027" is a defensible shape, on three pieces of evidence: a new line takes 18 to 24 months, Chinese electronic-cloth projects point at commissioning in 2027, and loom orders are reported booked out to 2030. The specific mid-to-late-2027 dates in circulation come from sources that cite nothing.
Is Korea's +74.5% laminate import figure the price history of my grade?
Almost certainly not. The figure — USD 20,728 per tonne in March 2026 against USD 11,880 a year earlier — is an HS-code-level aggregate across all copper-clad laminate grades, and the cause the reporting identifies is demand for AI semiconductor and AI server material, where buying concentrates on high-layer-count, high-speed grades. We found no price series specific to CTI 600-class laminate anywhere.
References (Public Sources)
- A Chinese financial daily, 27 June 2026, citing Prismark and Chinese brokerage research, for the June prices, the line-conversion mechanism and the 18-to-24-month build cycle; Chinese financial media, August 2026, for the mid-August average; Xinhua Finance, 16 October 2026, citing a brokerage note of 9 October, for the October moves and the easing expectation; an industry-research report citing listed-company annual reports and the Chinese fibreglass and electronic-fibreglass industry associations, for the CNY 3.5/m baseline and the 2027 commissioning programme; Chinese reporting of August 2026 on loom order books.
- Shanghai Metals Market, 7 July 2026, on the copper foil supply gap, industry utilisation and the 18 µm HTE processing fee. Benchmark Mineral Intelligence, 7 January 2026, and ADM Investor Services, August and October 2026, on LME copper; the exchange's own price pages were inaccessible (HTTP 403).
- Trade-press reports of early 2026 on a Japanese laminate maker's price-adjustment notice; a market-research firm's report of 6 May 2026 on Japanese suppliers' increases. Korea Customs Service figures for copper-clad laminate import unit prices as reported in the Korean press on 4 May 2026. Evertiq, 12 March 2026, citing ILFA GmbH, on lead times; Evertiq, 29 July 2026, on allocation programmes and their basis.
- BGF Industries technical data book (2022), for the 7628 and 7629 construction, with its own "not a specification" caveat. Trade and financial reporting during 2026 on the Japanese T-glass capacity expansion.
- PCB trade blogs consulted and not relied on for any figure here: the dollar-denominated 7628 price series and its forward projections, the 3-to-10-day inventory figure, the per-sheet Chinese FR-4 price, and the mid-to-late-2027 relief date.
What we could not establish: the share of total glass-cloth demand that 7628 represents, or the capacity split between standard and specialty cloth; the share of copper foil and glass cloth in a laminate's cost, which is why we convert no input move into a laminate move; a September 2025 baseline the sources agree on; any price series specific to CTI 600-class laminate; the lead time for high-CTI automotive grades; the Korean customs figures reproduced from primary trade statistics rather than press reporting; the LME's own published prices (HTTP 403); the increase counts attributed to one major Chinese laminate maker, which we found only in secondary blogs; and when the squeeze ends. Where each arises above, we have said so rather than fill the gap.
Nami Tech Solutions (NTS) works project by project, not from standing inventory. On an enquiry of this kind our work starts with the questions above — whether the grade is allocated, the lead time by glass style rather than in general, and what cut size and quantity a mill will actually take — put to mills in their own language against one written specification, so that what comes back is comparable between them. Where we have no basis for a view, as on the price path from here, we say so rather than supply one.
