Sorry it’s been a while, but we’re back.

Our call: HOLD. New money: don’t open a position. Holders: sell; this turns to trim at a beta of 1.31 or below.

It is a Japanese glass-fibre company from Fukushima, founded in 1923 as a textile spinner. It melts glass, draws it into filaments finer than a hair, spins them into yarn and weaves the yarn into cloth. One grade of that cloth, T-glass, forms the core of the substrate under almost every large AI chip, because it stays flat when the chip heats up. Nittobo is the only company we have verified making the thinnest grades at volume from its own yarn.

Where Nittobo sits: the glass cloth in the substrate core

Since 2024 the substrate makers behind AI chips have been short of that cloth. Nittobo’s Electronic Materials margin rose from 18.0% to ◆ 48.2%, and the shares rose from JPY 1,284 on 1 Oct 2025 to a closing peak of JPY 6,390 on 7 May 2026, then fell back to ◆ JPY 3,335 on 1 Oct 2026. At that price you are paying for the shortage to last a decade. We expect it to ease from early 2027, so we hold and would buy at JPY 1,400 to 2,150.

Every customer signal is captured in one shared context layer, always current and compounding. Agents and workflows build pipeline, chase every buying signal, and move deals forward, an always-on revenue engine running alongside your team.

With Attio, you’ll get:

Teams like Parallel, Turbopuffer, and Wordsmith build on Attio. Are you one of them?

Get started now

Basis: eight kessan tanshin (Japanese earnings releases) and company notices, read against the filing text. 237 public facts, with 40 load-bearing claims re-checked against a second source: 26 confirmed, 5 disputed, 5 refuted, 4 outdated. UN Comtrade and MOPS series, and a five-future DCF with a 20,000-path Monte Carlo. Facts as of 28 to 29 September 2026; prices at the 1 October 2026 close. Values marked with a red diamond, like ◆ this, change over time and move our view; each has a row in the Watch variables appendix.

The question is whether Nittobo holds a lasting toll on AI hardware or a head start that its own new capacity and three or four rivals close by 2028. We think it is the head start; the price assumes the toll. The schematic near the top shows where the cloth sits.

For two years substrate makers have sent executives to Fukushima to ask for more cloth, and most have gone home without it. Nitto Boseki, which trades as Nittobo on the Tokyo exchange under code 3110, is worth JPY 607bn at ◆ JPY 3,335, its close on 1 October 2026.

T-glass cloth matters because it keeps a large AI package flat as the chip heats up.

Glass cloth is a fabric woven from glass yarn. Each yarn is a bundle of glass filaments thinner than a human hair, and the cloth looks like fine grey muslin. Laminate makers soak it in resin and press it between copper foils to make the board material that almost every circuit board and chip substrate starts from.

An AI accelerator such as an Nvidia GPU or a Google TPU does not sit straight on a motherboard. It sits on a package substrate, a small multi-layer board that fans thousands of fine connections out to the coarser board below. The substrate has a stiff core of glass-cloth laminate, with wiring layers built up on each side from Ajinomoto’s build-up film (ABF), which contains no cloth (Ajinomoto product literature). In a conventional ABF substrate, only the core has cloth.

The core’s job is to stop the package bending as it heats. Materials grow as they warm, and engineers measure the growth as the coefficient of thermal expansion, or CTE: the millionths of its length a material gains per degree. Silicon expands about 3 parts per million per degree. The resin in a substrate core expands far more, and the glass cloth holds it back, so a core built on ordinary cloth still grows several times faster than silicon.

Heat the two together in a solder oven, then again in a data centre, and the package bows and solder joints crack. Engineers call this warpage, and it gets worse as the package gets bigger (Exhibit 1).

AI packages are getting bigger. New AI chip substrates grew from about 80 by 80 millimetres to 100 by 100 in 2026 (Commercial Times, 9 Nov 2025), and Nvidia’s Rubin substrate is larger again, with more layers (TrendForce, 30 Apr 2026). The bigger the package, the more a small expansion mismatch costs in failed parts.

T-glass is Nittobo’s answer. Ordinary electronic E-glass expands 5.6 parts per million per degree; T-glass expands 2.8, close to silicon, and it is stiffer, at 86 GPa of tensile modulus against 75 (Nittobo property table).

The difference is chemistry: T-glass is 64 to 66% silica, 24 to 26% alumina and 9 to 11% magnesia, with none of the calcium oxide or boron that make E-glass easy to melt. E-glass carries 20 to 25% calcium oxide and only 52 to 56% silica (Nittobo).

The same chemistry makes T-glass hard to produce. Furnaces melt glass fibre above 1,300°C and draw it through platinum-alloy bushings that need precise temperature control (Nittobo); T-glass runs at 1,600 to 1,700°C in dedicated electric furnaces (Postation, 24 Sep 2026).

Molten glass this rich in silica is stiff and hard to draw, and the filaments have to come out uniform enough to weave the thinnest cloths without breaks. Nittobo’s thinnest catalogue cloth, style 1010, is 13 micrometres thick, about an eighth of a sheet of office paper (Nittobo product table). Nittobo weaves the yarn on looms and finishes the cloth before it ships to the laminate makers.

A second special glass sells into the same boom. NE-glass and its successor NER-glass have a lower dielectric constant than E-glass, so they slow and smear fast signals less (Nittobo). These go into the server and switch boards that carry 112 Gbps links rather than into chip packages.

The scramble reached the general press in January 2026, when Nikkei Asia reported Apple competing with Nvidia, Google and Amazon for Nittobo’s cloth (14 Jan 2026). Trade press had reported seven months earlier that senior executives from Nvidia, AMD and Microsoft were visiting Nittobo to secure T-glass (DigiTimes, 27 Jun 2025); Nittobo has not confirmed any visit.

Hiroki Kajikawa, a senior executive officer, told Toyo Keizai that Nittobo sells only to laminate makers and used to have no direct dialogue with the chip companies; that, he said, had changed dramatically (21 Aug 2026).

WorldQuant BRAIN is a global simulation and research platform for students, aspiring researchers and quantitative thinkers from around the world.

On BRAIN, that spirit of exploration continues through challenges, collaboration and opportunities to develop new quantitative research ideas.

Learn more about WorldQuant BRAIN®

Nittobo reports six segments, and Electronic Materials earns almost all the profit.

In FY3/26, the year to March 2026, the group sold JPY 118.2bn and earned JPY 20.8bn of operating profit, with 2,793 employees at 31 March 2026 (FY3/26 tanshin; company site).

Glass yarn and cloth for circuit boards and chip substrates, the subject of this note. Sales JPY 49.3bn, operating profit JPY 19.4bn, 93% of the group total.

The other five segments earned JPY 3.6bn between them before JPY 2.2bn of corporate costs and eliminations.

Source: FY3/26 kessan tanshin, segment note.

Sources: Nittobo company history and TDnet filings; Nikkei (19 Apr 2019) for Baotek; TDnet 29 Aug 2025 and 6 Nov 2025.

Fiscal years: Nittobo labels them by the calendar year they start in, so the company’s FY2027 is our FY3/28, the year to March 2028.

The Fukushima Enterprise Center makes yarn and cloth, and all of the T-glass cloth for leading-edge logic comes from one building there, the former Fukushima No.2 plant (Nikkan Kogyo Shimbun, 30 Sep 2025).

In Taiwan, NITTOBO ASIA Glass Fiber melts and draws yarn in Chiayi. Baotek, listed in Taipei as Chien Jung, weaves in Taoyuan from group yarn and plans T-glass cloth mass production in FY3/28, with Nan Ya collaborating on weaving (Nittobo, Q1 FY3/27 results briefing, 5 Aug 2026).

17,212 square metres on the Fukushima site, under construction since October 2025, due for completion in December 2026, first output in January to March 2027 (TDnet, 29 Aug 2025). T-glass mass manufacturing there ramps through FY3/28 (Nittobo, Q1 FY3/27 results briefing, 5 Aug 2026), so the relief arrives in steps.

Hisanobu Hayashi became president on 1 April 2026 after running the Electronic Materials division and, before that, NITTOBO ASIA; he joined in 2013 and holds 899 shares (TDnet, 24 Dec 2025).

The board said it promoted him because the business was changing faster than the mid-term plan had assumed. Nittobo hit the plan’s final-year operating-profit target of JPY 20bn two years early, and the board raised it to JPY 36bn in May 2026.

The scarce step is the yarn, and Nittobo’s own spending shows it (Exhibit 2). The JPY 15bn Fukushima building buys weaving and finishing.

The yarn is a separate, slower programme, in four steps:

A furnace runs continuously once lit, and adding one takes about two years.

Customers say the same. “No extra capacity means no extra capacity. Pressuring Nittobo does no good,” a substrate-maker executive told Commercial Times; he expected relief only when new capacity opens in the second half of 2027 (16 Jan 2026).

Demand jumped from about May 2024, said Mikami, deputy head of the Electronic Materials division, and “honestly, our supply capacity is tight” (Denshi Device Sangyo Shimbun, 17 Oct 2025).

Which rivals, if any, make the thinnest low-CTE grades used in flagship GPU and ASIC substrate cores. Taiwan Glass has passed customer certification for a special-spec low-CTE cloth and started deliveries (Commercial Times, 11 May 2026). Fulltech’s low-CTE yarn has an overseas customer’s approval, but its cloth is still in certification (MoneyDJ, 2 Sep 2026). China’s Grace Fabric sells certified low-CTE cloth to laminate makers and had 5.0% of the market by 2025 revenue (Hong Kong listing filing, May 2026). Nan Ya weaves for Nittobo rather than against it. No rival was verified at volume in the thinnest grades on its own yarn, and no source names the grades each has qualified.

Both halves of the bottleneck wear down: a rival can buy a furnace, and qualifications come with time. Today’s margin is what Nittobo has to lose when they do.

Electronic Materials went from an 18% margin to a 48% margin in three years. Segment sales rose 65% between FY3/24 and FY3/26 while its operating profit rose 3.6 times, and the first quarter of FY3/27 printed a ◆ 48.2% margin on sales up ◆ 29.5% (Exhibit 3).

Electronic Materials: full years and the segment margin

Source: kessan tanshin 9 May 2025, 12 May 2026, 5 Aug 2026; share of group OP is segment OP over consolidated OP. Chokepoints analysis. Margins are on external sales, the tanshin basis; on sales including intersegment transfers, Nittobo put the Q1 FY3/27 margin at 38.9%, against 28.8% a year earlier (company materials, via Postation, 21 Sep 2026). Our thresholds use the external basis.

Three margin drivers:

The rises have stopped, by Nittobo’s own account: in May it had ◆ no further T-glass price revision planned and would not raise prices on supply and demand; in August it said it would look at price again once it could supply more, and that NER-glass would not rise this year (company Q&A, via Postation, 21 Sep 2026).

Japan’s trade data show the same shift. Exports of woven glass cloth to Taiwan, the main substrate hub, averaged USD 252 per kg in January to July 2026, up from USD 92 in 2022. Tonnage fell from 302 tonnes in 2022 to 166 in 2025 and 100 in the first seven months of 2026 (Exhibit 4).

The customs code mixes commodity and special cloth, so read the unit value as a mix signal rather than a T-glass price: a thin special grade weighs less per square metre, so a richer mix lifts it with no change in any price. In the same seven months Japan’s cloth fetched USD 32 per kg in Thailand and USD 289 in Korea.

Less glass is leaving Japan, at far more value per kilogram. It cannot show what any given cloth now costs.

Japan’s woven glass-cloth exports: less weight, far more value

Nittobo does not name its customers. Its direct buyers are laminate makers, and the chip companies sit two or three steps away; Apple stationed staff at one laminate maker, Mitsubishi Gas Chemical, in autumn 2025 to secure substrate materials (Commercial Times, 16 Jan 2026).

We found ◆ no disclosed long-term agreement, prepayment or customer-funded capacity. The nearest thing is Nittobo’s statement in May that it would ask customers to share the cost of any capacity beyond the current plan (company Q&A, via Postation). The one long-term supply deal on record runs the other way: Nittobo sells Nan Ya NER yarn that it sells to no one else, and Nan Ya weaves about 20% of Nittobo’s cloth by 2027 (Nan Ya release and Anue, 28 Nov 2025).

How much of Electronic Materials is special glass and how much is commodity E-glass. Nittobo discloses only the index. One outside yardstick: Frost & Sullivan, in Grace Fabric’s Hong Kong listing filing, sized the whole low-CTE cloth market at USD 128m in 2025. At Nittobo’s reported 50 to 90% share, that is JPY 10bn to 17bn of T-glass cloth, a fifth to a third of segment sales on our arithmetic. Our model assumes 70% special glass across NE, NER and T, so most of that 70% would be low-dielectric NE and NER cloth.

Three things distort the group numbers:

The 48% came from a sold-out plant in a shortage. The valuation question is what happens to it when the plant triples and customers have four suppliers to call.

A buyer at ◆ JPY 3,335 is paying for the shortage to last.

Nittobo trades at 30.4 times its own FY3/27 EPS guidance. The average sell-side target was JPY 4,532 on 1 October, down from JPY 4,690 on 29 September and JPY 5,451 at the end of June (minkabu, 1 Oct 2026).

To justify JPY 3,335 at a 9.2% discount rate, our model needs one of two things:

In our view the shortage is temporary and the yarn know-how is what lasts. Four facts put a date on the shortage:

Even so, the business stays good in our numbers. Our toll-road future gives Nittobo the thinnest grades, a 95%-full tripled plant and a margin easing from 47% to 42%, against 18% before AI. That future is worth JPY 2,189. The price sits 52% above it.

Hayashi said on 5 August that even the threefold expansion will not be enough (TrendForce, 6 Aug 2026), and management says it will invest beyond the plan if the opportunity is large (company Q&A, via Postation). A US broker still sees glass cloth about ◆ 30% short in 2028 (Postation, 15 Sep 2026). If Vlex, the next-generation low-CTE glass, becomes the standard from 2028, the lead renews. We give that world a 7% weight and value it at JPY 3,047, still 9% below spot.

Who Ibiden’s third and fourth suppliers are, which Japanese supplier is expanding, and whether the newcomers are qualified for ABF core cloth in leading-logic packages or only for BT and thicker grades. Ibiden names no one. In October 2025 it said it had qualified one alternative supplier but still worried about that supplier’s capacity (Ibiden Q&A, 31 Oct 2025). The expanding Japanese supplier is most plausibly Nittobo itself, whose Fukushima and Taoyuan lines ramp through Ibiden’s next fiscal year. Nippon Electric Glass fits the timing, but its new fibre is a low-loss board grade rather than low-CTE. Neither is confirmed. Taiwan Glass is the best-supported newcomer, but its Ibiden and Resonac certification exists only as market talk (Economic Daily News, 16 Apr 2026). Asahi Kasei entered in April 2026 with quartz cloth, a different material (Nikkei, 1 Apr 2026). If the Taiwanese incumbent is Nittobo’s own Baotek, the count is less independent than it reads.

We are wrong if Nittobo’s lead lasts. Two signs would show it, and either would change our call:

The second depends on Vlex. It expands about 30% less than T-glass, 2.0 parts per million per degree against 2.8 (Nikkei, via TrendForce, 4 Feb 2026; Exhibit 1), and Nittobo’s own deck adds lower signal loss.

Development is finished and samples are with laminate makers (Denshi Device Sangyo Shimbun, 17 Oct 2025). The roadmap puts the business at ◆ FY2028 onward, from April 2028 in Nittobo’s labelling (Nittobo, Q1 FY3/27 results briefing, 5 Aug 2026), while Nikkei reported practical use as early as calendar 2028. In February management said it would push Vlex to become the de facto standard against new entrants (company briefing, via Postation, 21 Sep 2026).

A lower discount rate softens the call for holders. Our line is sell while the price sits above every future we model, and trim once one future clears it. Beta, how far the shares swing with the market, sets the discount rate.

At 8.6%, our low case with a beta of 1.30, the weighted value is JPY 1,566. The supercycle future, at JPY 3,377, then clears the price by about 1%, so the holders’ line becomes trim. The switch comes at a beta of 1.31, within one standard error of our 1.42 estimate, so the line between sell and trim is thin. The weighted value reaches the price only at a WACC of 5.0%.

Whether Vlex needs a fresh qualification at each substrate maker or rides on existing T-glass approvals. No source says. The same logic cuts against Nittobo today: an approval a customer grants a rival during the shortage survives Nittobo’s expansion (Postation, 21 Sep 2026). If Vlex needs fresh approval, rivals get a chance to qualify alongside it in 2028.

What the price needs: the supercycle’s volumes with FY3/29 prices and margins held to about FY3/34, scarcity economics lasting about eight years longer than in any of the eight past component shortages we checked: MLCCs, silicon wafers, copper foil and DRAM in 2016 to 2019, glass cloth in 2016 to 2018 and 2020 to 2022, and ABF film and chips in 2020 to 2023. Held all the way to FY3/36, the same construction is worth JPY 3,633, 9% above spot.

The reverse DCF keeps our toll-road assumptions and asks what one number would have to change to justify the price.

Source: Chokepoints valuation model v4.1 (29 Sep 2026), spot at the 1 Oct 2026 close; WACC 9.2%. The growth rate needed rises with the discount rate: 20.2% at 8.6%, 26.5% at 10.2%. No mix of our five futures reaches the price: spot sits above all of them.

The earnings multiples say the same. The price is 30.4 times FY3/27 guided EPS of JPY 109.87, 22.3 times our toll-road FY3/28 EPS of JPY 149.8 and 17.1 times our toll-road FY3/29 EPS of JPY 195.0.

The last multiple looks cheap until you see what FY3/29 is in our model: the year the new capacity ramps in, running 95% full at a 46% margin, and the peak of EM growth. After FY3/30 our toll-road sales grow about 2% a year on average and the margin drifts to 42%. Paying 17.1 times peak-year earnings only works if profits keep rising after the new capacity lands.

Our toll road sits on consensus operating profit to FY3/29, year by year: JPY 31.6bn, 40.7bn and 52.9bn in our model, against 31.6bn, 40.6bn and 52.9bn from the sell side (MarketScreener, 29 Sep 2026). FY3/27 consensus EPS rose from JPY 108.03 on 1 July to JPY 118.26 (minkabu, 29 Sep 2026), against our JPY 116.4. For FY3/28 the sell side expects sales of JPY 166.0bn (S&P Global, via StockAnalysis), above the plan’s JPY 155bn and our JPY 157.6bn.

Our disagreement with the sell side starts after FY3/29. Even on the toll road, the gap to the price is the fade after that year and the discount rate: on that path spot implies a 6.9% WACC, against our 9.2%.

Our method has leaned low before. Run on the information of 29 September 2025, it would have put FY3/27 operating profit about 30% below the JPY 30bn now guided, because it allowed no further price rise while the shortage was a year old.

The young-shortage rule we now apply, one more price step while a shortage is under about two years old, brings that backtest within 3%. This shortage is about 2.75 years old and Nittobo plans no further T-glass rise, so the rule adds nothing today.

Weighted across five futures, Nittobo is worth JPY 1,446 a share at our 9.2% WACC, and no weighting of them reaches the price.

Each future flexes a different named driver, so the weights estimate which driver wins. Each weight starts from a base rate, how often that outcome happened in a class of past cases coded on the outcome the future describes, and is then updated for 16 dated evidence items.

Glass-core substitution carries the largest weight, 39%: in 13 past packaging-material transitions, the new technology displaced the incumbent in flagship packages within about five years of its first pilot 39% of the time, and Intel’s glass-core pilot line dates from 2023.

The Monte Carlo varies capacity, price, utilisation, margin, the yen, the discount rate and terminal assumptions inside each future. One shared demand factor drives price, volume and capacity together, and the margin moves with price and volume.

Across 20,000 paths the median value is JPY 1,274 and the 90th percentile JPY 2,958. Between 3% and 7% of paths clear the price, depending on how wide and how correlated we make the shocks; that range measures how sensitive our model is to its own settings.

Each future’s central value sits below the price, but its paths spread around it: 42% of supercycle paths clear the price, 18% of toll-road paths and no more than 4% of any bear future’s.

Source: Chokepoints valuation model v4.1 (29 Sep 2026), spot at the 1 Oct 2026 close; WACC 9.2%. Weights are base rates updated for evidence, as set out above, rounded to whole percentages.

No future we can evidence reaches the price

The discount rate is our second largest judgement after price. We build it in four steps:

Our first version used the raw yield with a 5.5% premium, which counted part of Japan’s default risk twice, and got 10.2%. Each percentage point of WACC moves the weighted value by about JPY 173 to 225.

Exhibit 6 runs the weighted value across discount rates and terminal growth rates, the grid a fairness opinion shows. No cell reaches the price.

Exhibit 6. Probability-weighted value by WACC and terminal growth; every cell sits below JPY 3,335. Low case: beta 1.30 (8.6%); v1: raw JGB and a 5.5% premium (10.2%). ’None’ means spot sits above every future at that rate. Reversion row: each future’s EM margin runs from its FY3/28 level to 20% by FY3/36 (FY3/24 was 18.0%), with volumes, prices and weights unchanged. Source: Chokepoints valuation model v4.1 (29 Sep 2026), spot at the 1 Oct 2026 close.

Exhibit 7 moves one driver at a time to the 10th and 90th percentile of its range, set from its own history where it has one, with the margin tied to price and volume through the segment’s fixed costs.

Price moves the value most: a price index 18% lower from FY3/28, with unit costs unchanged, takes the weighted value to JPY 794, and 18% higher takes it to JPY 2,098. The discount rate comes second. No driver at either end reaches the price.

Price moves the value most; nothing at the 90th percentile reaches spot

JPY 1,062 is the weighted value if every future’s Electronic Materials margin reverts to a normal level, the last row of Exhibit 6. McKinsey’s method for cyclical companies bases the continuing value on a normal level of profit (Koller and Wessels, 21 Oct 2025), and our futures end FY3/36 at margins that weight to 32.1%.

In this check each future keeps its volumes and prices, and its margin runs in a straight line from the FY3/28 level to 20% by FY3/36, midway between the pre-AI 18.0% of FY3/24 and 22%. Across 18% to 22% the weighted value is JPY 1,005 to 1,119. Reverting only the two bear futures gives JPY 1,326 to 1,367.

Reversion is the main risk to the entry zone, and we keep the zone for now: FY3/24’s 18% came from a mix heavier in commodity E-glass, and our second-source and cycle-bust futures already end at 29% and 37%. The weak point is that FY3/24 is the only pre-AI year on the current segment basis. The 35% tripwire in section 11 tells us if we are wrong.

The share price has moved on how investors read Nittobo’s disclosures, and three times it fell on good results. Each line maps an event to the lever it moves.

Moves are daily closes, split-adjusted, from exchange data (Chokepoints analysis). Abnormal returns strip out the Nikkei 225 with a market model estimated over the 250 to 30 trading days before each event; a t-statistic above 2 in either direction is unlikely to be noise. Broker actions are from Kabutan and Monex’s IFIS rating log; IFIS does not name the 28 Sep broker. Average targets from kabuyoho, minkabu and JapanIR, 1 to 29 Sep 2026.

Up 160% in a year, down 48% from the May peak

The August fall is the one to understand. Q1 beat and the guide went up, yet the shares lost 17.9% the next day, 16.6 points more than the market model predicted.

The raise only just cleared the JPY 28bn QUICK consensus, the stock had risen 12.1% into the print, and AI-related names were weak that day (Fisco, 6 Aug 2026). Nikkei’s reading was that the high-function glass production plan had barely changed (6 Aug 2026). Investors read it as the end of the price rises, while the capacity plan, which sets the expiry date on the shortage, did not change.

The Taiwanese figures understate the cloth business because they are whole-company revenue. Glass fibre, mostly electronic cloth, rose from 23% of Taiwan Glass revenue in 2024 to 30% in the first nine months of 2025 (company spokesman, via Statementdog, 16 Jan 2026). The chairman expects about 40%, and analysts more than 50% by the end of 2026 (Economic Daily News, 12 Jun 2026; Anue, 11 Sep 2026).

Fulltech’s surge is mostly price and low-dielectric cloth. Special grades were about 55% of its first-half revenue, it raised E-glass cloth prices about 30% from 1 July, and its low-CTE cloth has not yet cleared certification (MoneyDJ, 2 Sep 2026; Money Weekly, 10 Aug 2026). For now, read Fulltech as a cloth-price signal rather than a T-glass rival.

Taiwan’s cloth makers are growing, on volume and on price

The relief is dated

35% is our tripwire on the entry zone. If the Electronic Materials margin stays below 35% for two consecutive quarters after Fukushima output starts in January to March 2027, we re-cut the zone toward the reversion case, JPY 1,005 to 1,119 at an 18% to 22% normal margin.

Bought in the entry zone, the supercycle future is our upside and the second-source and glass-core futures are the losses we accept. Until the price reflects the 2027 date, we do nothing.

Each of the 25 watch variables changes over time and moves the view; in the body its current value carries the red diamond. The full table, with each current value, source, cadence and threshold, is in the appendix of the web edition.

Not investment advice. This note is for information and education. It is not investment advice, a personal recommendation or an offer to buy or sell any security. Figures are from public filings and reports as dated in the text; estimates are ours and can be wrong. Do your own research.

All figures come from Nittobo’s kessan tanshin and TDnet notices, rival and customer disclosures, trade press and customs data, each dated where it is used, and the publisher’s own valuation model. Broker targets are cited as published averages; no broker or expert-network research is reproduced. © 2026 chokepoints.ai, Issue 008 All issues