# When the Door Is Shut, You Don't Even Know What's Missing Inside: Notes on a Semiconductor Materials Maker

Source: Realpha Blog (blog.getrealpha.com)
Original article and charts: https://blog.getrealpha.com/en/blog/caibaogou-2026-10-28-572-bbu-ft/

> Listening notes on Statementdog podcast episode 572 with Nytex chairman Chen Hsun-sen and president Chen Yu-han: unpacking the claim that semiconductors are a walled courtyard with the gate closed, and the conditions under which that moat stops working. Educational discussion of industry structure, not investment advice; no stock recommendations or price targets.

Published: 2026-10-28
Locale: en
Tags: Statementdog, semiconductor materials, BBU, composites, supply chain


![A deep factory corridor leading to a half-open heavy door, dim amber plastics shop outside, cold white cleanroom light spilling out from inside](/covers/caibaogou-2026-10-28-572-bbu-ft-cover.png)

> He who looks at the sky from the bottom of a well and calls it small — the sky is not small.
>
> —— Han Yu, *On the Origin of the Way* (Tang dynasty, c. 805; my own translation)

In the October 28, 2026 episode of the Taiwanese investing podcast *Statementdog* — episode 572, the industry-expert segment — chairman Chen Hsun-sen and president Chen Yu-han of the composites maker Nytex talked about how a 38-year-old plastics compounder got into semiconductors. Chen Hsun-sen described the industry as a walled courtyard with the gate shut: from outside you cannot see what is in there. He said he spent NT$60 million rebuilding the plant and buying equipment before holding a single order. Chen Yu-han gave the numbers: high-end applications — semiconductor parts, IC trays, and BBU materials — were two to three percent of revenue in 2023 and about twenty-seven percent in the first half of 2026. Two conditions hold that shift up: a mid-stream customer willing to run your qualification, and an upstream resin that nobody has yet turned into a commodity. Knock either one out and the same gate swings shut.

![Two bars differ dramatically in height: the short bar on the left represents high-end applications contributing 2 to 3% of revenue in 2023, while the bar on the right is ten times taller and represents 27% in the first half of 2026.](/figures/nitt-high-end-revenue-share-en.svg)

## One Gate, and the Things Behind It You Never Knew You Could Make

The hosts asked Chen Hsun-sen whether he worries about rivals following him in — plenty of compounders have been around as long as he has. His answer made me stop the tape. Semiconductors, he said, are like a walled courtyard with the gate closed; standing outside, you see nothing. Once he was inside, he could see how many of the things in there he was already capable of making. Before that, he had never handled PEEK — thirty years in composites and he first heard the name three years ago. PEEK is a high-temperature engineering plastic with real structural strength, used for wafer carriers and, lately, robot joints.

That line puts the barrier to entry somewhere an outsider would not guess. Intuition says the hard part is technology or equipment. The order that came out of this conversation runs the other way.

![Three walls are arranged from outermost to innermost: an arrow from outside passes through the first wall, “Materials,” but stops at the second, “Willingness to Validate,” while test data lies beyond the third wall.](/figures/nitt-three-walls-en.svg)

### Layer one: you cannot simply buy the resin

Nytex takes upstream engineering resins, adds its own formulation and compounding, and ships a spec the customer asked for. The problem is the top layer: the firms holding volume and patents will not sell to you. Chen Hsun-sen was blunt about it — they look at who your end customer is before deciding, and without volume you get nothing. The large Western and Japanese suppliers compound the high-end grades themselves and keep the resin in-house, because selling it to you creates a competitor. So for years Nytex sat in the middle tier, glass-filled nylon and glass-filled PC, with no path upward.

### Layer two: the resin is useless unless mid-stream will qualify you

This is the layer Chen Yu-han added, and I think it matters more than the first. A carrier or a compound is a tiny slice of total process cost, while a bad material lands its cost on the wafer. From where the mid-stream sits, nobody wants to carry supplier-switch risk to save a little on material. He said it plainly: if Gudeng had had no genuine supply problem, friendship alone would never have bought them a qualification slot. The episode describes a competitor pressing upstream suppliers to cut Gudeng off (that is the guests' account; I cannot verify the third party's side), which made a local partner willing to rebuild its plant worth having.

Stack those two layers and the scarce good surfaces: test data. Chen Yu-han said Gudeng shared its measurement results, which told them which direction to pull the formulation — without that, you do not know which knob to turn. Chen Hsun-sen said the same about BBU work: they get the temperature curves from nail-penetration and thermal-runaway tests, see where control is lost and how the fire spreads afterward, and only then know what the next formulation has to fix. Spending money, as he put it, is how you find the problem.

That is what the gate really means. From outside you do not know the spec in use, and even handed a spec you cannot be sure it is the right one. Co-host Sky made the point that stuck with me: of ten projects floating around out there maybe three go anywhere, every one of them costs time, and picking those three makes you a genius. Once the gate is open, what you hold is a whole working sense of what is worth making and what is not; the spec sheet is one page of it.

### Layer three: once the gate opens, the capability spills over

What Nytex practiced at Gudeng was dispersion uniformity in anti-static compounds. Wafer carriers are demanding about this, and loading more additive pushes dimensions and warpage off target, so you hunt for a balance point. Four years of that let them into IC trays in a little over a year, because as a single package goes from a few dollars to twenty thousand dollars, the demands on anti-static uniformity and dimensional stability rise with it — and they had the capability on the shelf. Chen Hsun-sen said that once inside semiconductors he found anti-static and low-outgassing requirements everywhere, which is a subject you can keep walking down for years.

The spillover also went somewhere harder to foresee. They pursued aerospace certification because of one small U.S. customer making aircraft parts, and the projects showing up now are drones: as drone requirements rose, some builders began demanding aerospace-certified material suppliers, and few Taiwanese compounders hold that paperwork. That line was never laid for drones.

![From a core capability honed over four years, three branches of differing length extend to the right, and the one leading to drones forks off sideways from aerospace certification, heading in a direction no one originally planned for.](/figures/nitt-capability-spillover-en.svg)

### The conditions under which it fails

The company has already lived through it once. Chen Hsun-sen said electronic connectors kept shrinking while demands kept rising, he could not obtain the high-end grades, and that market fell by sixty to seventy percent — three, four, five straight years with no way to refill the hole, and a decade to climb back. The blunter example is nylon 66: once China brought on its own capacity, the price fell from a peak above US$5,000 per ton to a bit over US$1,000 today. When the thing behind the gate becomes a commodity, the gate stops being a gate.

![A price line falls steadily from a high point toward the ground, while the door beneath it collapses from a tall gateway into a slab nearly level with the floor.](/figures/nitt-door-flattens-en.svg)

So I listened to this as a window with a width, and the width is set by three things: how long a rival needs to finish safety and material certification (Chen Yu-han estimates six to nine months at minimum, plus separate filings such as an RTI thermal rating), whether the mid-stream customer has any motive to replace you, and whether the upstream resin gets commoditized. The first two look sturdy near-term. The third is exogenous and nobody reads it well.

![Three bars of different lengths stand for the three things holding the time window open, and the window's right edge is set by the shortest bar, whose own edge is blurred and diffuse.](/figures/nitt-window-width-en.svg)

### Why "good relationships" and "strong technology" don't carry the explanation

The guests disposed of the relationship story themselves: Chen Yu-han was explicit that qualification happens only when the customer is under supply pressure. The technology story has the sequence backwards — Nytex spent seven or eight months building a cleanliness-grade plant and buying equipment, then roughly seven more months producing a first qualified grade, and the NT$60 million went in with no orders behind it. Chen Hsun-sen said he dared to spend it because the balance sheet was clean and the pandemic years had been profitable. Capability grew out the far end of the money and the waiting.

![On a single timeline, the block for the 60 million of spending sits at the far left during the no-order period, the revenue block does not begin until fifteen months later, and the span between them is blank.](/figures/nitt-spend-before-order-en.svg)

One more explanation I considered: maybe Taiwan simply has a giant foundry and the supply chain beneath it rises on its own. The version in the episode is more specific — Gudeng walked Nytex through the gate and introduced the upstream resin suppliers too. A supply chain is a set of individual lines someone pulls you along, nothing like rain falling evenly on everyone.

## A Few Other Things Worth Remembering

- The BBU material sits between battery cells doing two jobs: spreading heat so cell life runs roughly twenty percent longer, and forming a char layer during thermal runaway that keeps the fire inside a single cell.
- These materials have to pass UL 9540A. Chen Yu-han said one safety certification alone runs over NT$6 million, approaching NT$10 million with the other tests; in the first year they cut their own molds, built a small pack themselves, and sent it to a third-party lab thirty or forty times, not counting runs done with customers.
- Taiwan has only a handful of labs that do this, and a busy season means a two-to-three-week queue. A failure sends you back to work out whether it was your wiring or your formulation — so "one test is quick" and "one cycle is slow" are both true.
- IC trays went into volume production last October and revenue has climbed each quarter since. Square trays appeal to the foundry because edge scrap falls and the same area yields more units.
- One very plain decision during the turnaround: to spread fixed cost, Chen Hsun-sen told his sales team to take anything that did not lose money. Two years later gross margin went from 3% back to 15–20%.
- They funded high-temperature materials research twenty years ago, far too early, and the money went nowhere; that market only turned good recently. Too early and never done look identical on the books.

## Further Reading

The episode itself (*Statementdog* #572, October 28, 2026) is the source — these are listening notes, so take the details and tone from the show. If you want to dig further: UL 9540A is the published safety standard for thermal-runaway fire propagation in energy storage systems; nylon 66 price history is on the plastics pricing sites; and for Nytex's and Gudeng's revenue mix and certification progress, the companies' own disclosures are the safer read.

## The One Thing to Take Away

One idea: in the places you feel locked out of, whatever is missing behind the door is probably not what you guessed from outside. Nytex thought the shortage was resin, and found after getting in that the shortage was somebody's willingness to share test data — and once the door opened, what appeared was a pile of work it had never occurred to them they could do.

Here's something I've tried, if you want it: pick one thing you've weighed for a while and never moved on — switching fields, approaching a new client, joining a group — find a person already inside, buy them coffee, and ask one question: "What are you short of right now?" Don't ask how to get in; they can't answer that, because from inside it's all just common sense. The first time I asked it this way, what came back had nothing to do with the thing I'd spent six months preparing.