One Boom, Two Semiconductors: What a Single Word Erased

Introduction: One Boom, or Two?

"The biggest semiconductor boom on record." That headline has papered the business pages all summer. The evidence is plain enough: in June, semiconductor export value was up 171% on a year earlier — 2.7 times in twelve months.

Having watched this industry for a long time, I'll grant that on the numbers alone, the headline isn't wrong. But take that one word — "boom" — in your hand and turn it over, and two industries split apart inside it, moving in opposite directions.

One side makes and sells the chips. The other sells the machines that stamp those chips out. Under the same "semiconductor boom," one side shot up and the other bent down.

The shape is the letter K. The aggregate number surges upward, but look inside and it spreads into an upper arm and a lower arm. I'll call that spread the K. And I'll call the two industries inside it the two semiconductors.

So this piece asks one question. Is what we're looking at one boom — or two booms a single word has lumped into one? It splits that word along the line of who sells what, and lets the K show. Then it looks at how the statistical machinery of summing and averaging hides the split.

The Boom Is Real — But One Word Covers Too Much

This boom is not a statistical illusion.

Take exports alone. In May 2026, semiconductors' share of all Korean exports passed one-third. In the previous sixteen years, that share never once cleared one-fifth — 20.6% was the ceiling. It has now sailed well past that wall.

Still, the figure +171% carries a whiff of sleight-of-hand. The year it's measured against, 2025, was dismal. When the floor is low, a rebound rate looks inflated — the base effect. So I re-measured in absolute terms. May 2026 total exports ran 26% above the best month of 2025. The boom is real. It is simply not as dramatic as "record-breaking" makes it sound.

One thing I'll set aside honestly. That export value rose is certain; whether it rose because volume grew or because prices climbed, we can't tell the two apart from this number alone. So I'll go only this far: export value surged.

Here is what we've established so far, at a glance.

Table 1: The "Record-Breaking" Impression vs. What the Numbers Show

The industry has an old rule of thumb: "if the finished product sells this well, the equipment that makes it booms right alongside." A busy factory floods its suppliers with orders for new machines — perfectly intuitive. That intuition is breaking right now.

Two Semiconductors — Same Quarter, One Keeps ₩70 on Every ₩100, the Other Runs a Loss

Setting Device and Equipment on One Yardstick

Now I'll split "semiconductor company" into two kinds.

The firms that make and sell chips I'll call the device makers. Samsung Electronics and SK hynix belong here. They sell a finished product: memory chips.

The firms that sell the machines that stamp those chips out I'll call the equipment makers. Places like Wonik IPS, Jusung Engineering, and Hanmi Semiconductor. They don't sell chips themselves; they sell the gear that makes chips.

Set the two kinds against a single yardstick. The yardstick is operating margin — the share of every ₩100 in sales that a company keeps as profit from its core business, what people usually call the margin. The moment is the same for both: the first quarter of 2026.

Start with the device side. SK hynix kept ₩71.5 on every ₩100 of sales. Up sharply even from a year earlier (42.2%). Keeping over ₩70 on every ₩100 is a number you rarely see in any kind of manufacturing.

The equipment side is the mirror image. Jusung Engineering, a company that a year ago kept ₩28 on every ₩100, ran a loss this quarter instead (operating margin −12.8%). Its revenue was cut in half, too. Hanmi Semiconductor is still profitable, but its margin was cut to a third — from the 47% range to the 16% range — and its revenue shrank to about a third of a year earlier.

One caveat: the three equipment firms did not all move the same way. Wonik IPS actually swung from loss to profit. So to say "equipment was wiped out" would be a lie. The accurate statement is that the direction split — and the weight, clearly, is on the downside.

The device upper arm and the equipment lower arm. That gap, opened in the same quarter under the same "semiconductor boom," is the K. The chart below shows the split at a glance: the device bar (SK hynix) rises above the zero line, while the bar for one equipment maker (Jusung) drops below it.

Why Samsung Is Left Out

A sharp-eyed reader will ask why Samsung Electronics was left out — Samsung is a device maker too. There's a reason. Samsung's reported operating margin comes in at 42.8%, but that number is not the semiconductor business carved out on its own. It is the whole company (consolidated) — smartphones and displays folded in. Read it as "Samsung's chip margin" and you read it wrong. On a pure, chips-only basis, SK hynix (71.5%) is the more honest figure. Samsung shows the direction; the weight sits with SK.

Now set the two kinds side by side in one table. Since the chart above owns the margin, the table carries only the other axes.

Table 2: Two Semiconductors — One Boom, Two Report Cards

Laid out this way, it's clear: device and equipment are opposites in what they sell, in the direction of revenue, and in the market where they list. And yet we bundle the two under a single word — "semiconductor."

The Crux — The Boom Looks Uniform Because Summing Makes It Look That Way

The very name "the K" is a borrowed lens. It comes from the post-pandemic recovery debate — the one that said the economy looked like a V in aggregate but split into a K by class once you looked underneath. That view — the sum surges while the composition splits — I've carried straight over to semiconductors.

Two Erasures That Work by Different Rules

So why doesn't the headline show this K? Why does it collapse into the single phrase "record-breaking boom"?

The answer is in how the statistics are built.

There are two erasures, and they work by different rules. Take them one at a time.

First, the export number never contained equipment to begin with. What customs logs as "semiconductor exports" is chips. The manufacturing equipment that makes those chips is entered in an entirely different column of the customs ledger. So the +171% export figure didn't "erase" equipment. It simply never held it.

What is counted as "semiconductor exports" (HS codes 8541 and 8542) includes chips such as memory and system semiconductors, and — read broadly — LEDs and solar cells as well. Semiconductor manufacturing equipment, by contrast, is classified under entirely different codes and does not register in this export figure.

The second is where summing actually erases. Add up the revenue of the five companies (two device makers plus three equipment makers) and the device side is 99.86% of the total. Out of every ₩1,000 in sales, device makers are ₩998.6; all three equipment makers together come to a bare ₩1.4.

What does that mean? The instant you sum the five and call it "semiconductor industry revenue," that number becomes, in effect, the report card of two device companies. Equipment is plainly inside the sample, yet in the addition it vanishes arithmetically. Averaging does the same. Equipment's losses are buried under device's enormous profits, leaving not a trace.

One objection: "isn't 99.86% inflated because it folds in Samsung's whole-company revenue?" It is — the Samsung I dropped from the margin, I kept in the revenue sum. But drop Samsung entirely and put SK hynix alone against the equipment makers, and equipment's share is around 0.5%. Even without Samsung, equipment is under 1% of total revenue. The conclusion holds.

Here is what this means. The "record-breaking boom" looks uniform — even, consistent — not because the industry is uniform, but because summing makes it look uniform. Remove that one component, the addition, split it into two kinds, and the uniformity collapses on the spot.

The Two Samples Don't Carry Equal Weight

Still, I'll be honest that device and equipment don't carry equal weight. The two device companies (Samsung and SK) are effectively all of Korean memory. Their representativeness is high. The three equipment companies, by contrast, are only a partial sample of a far larger equipment industry. So read the K's device upper arm as solid, but read the equipment lower arm carefully — only as far as "these three companies did this."

Why Only the Device Side Put On Weight — The Money Is Crowding into the Back End

Same boom — so why did only one side put on weight? From here on, we're no longer dealing with settled numbers but with the mechanism that explains them. Take this not as a verdict but as the most plausible account.

Making a semiconductor breaks into roughly two stages. The front end (front-end process) etches circuits onto a thin disc — a wafer. The back end (back-end process) cuts, stacks, packages, and tests the chips so made.

The thing driving today's AI frenzy is the AI accelerator — a chip built solely for AI computation.

The high-performance memory that pairs with that accelerator and governs its performance is HBM. It's built by stacking thin chips in many layers. A textbook back-end job.

So the chain runs like this: AI accelerator demand surges → HBM orders explode → and that lands straight in the revenue and profit of the device makers (SK hynix, Samsung) that build the HBM. That is the first engine behind the device side's exploding margin.

And equipment? When a fab orders equipment to lay down a new line, it takes time for that money to register as the equipment maker's revenue. It's booked only after the order is placed, the machine built, installed, and signed off on inspection. Usually two to four quarters later — half a year to a year.

Numbers like these are called lagging indicators. The invoice arrives long after the order was placed. So an equipment maker's results today are a shadow cast not by "today's orders" but by "past orders."

To sum up: the device boom registered in results immediately, while equipment's share, being a lagging indicator, simply hasn't reached the results yet — same boom or not. So the real line between the two isn't "semiconductor or not." Device and equipment are both semiconductor. The real line is "has that money been booked into results yet?"

The Market Draws the Same K — But It's a Rhyme, Not Proof

The stock market draws the same K.

So far this year the KOSPI is up 58%. The KOSDAQ, meanwhile, is down 16%. The large device companies list mostly on the KOSPI, the small equipment companies mostly on the KOSDAQ. The big-company market is smiling; the small-company market is in tears.

But you can't use this as evidence that "the market already knew of equipment's collapse and priced it in." The KOSDAQ holds every sort of small-cap, not just equipment. You can't pin the KOSDAQ's fall on equipment.

So I call this a rhyme. As the end-sounds of lines rhyme in a poem, the market's K rhymes with the earnings K, running in the same direction — nothing more. A rhyme firms up the mood of the piece, but a rhyme doesn't make the earlier line the cause of the later one. The weight of this piece rests entirely on one source: corporate earnings (DART filings). The market is only a chord laid on top.

A quick look at the US, too. The US semiconductor index (SOX) has fallen more than 20% from its June high. At a glance it looks like a crash, but measured from the start of the year it still sits 58% higher. Not a collapse — a correction. And this is a US index, so it doesn't point directly at Korean semiconductors.

"So Is the Equipment Industry Dying?" — Answering Four Objections

Now to meet head-on the strongest objection a reader might raise by this point.

The most dangerous objection is this: "If equipment revenue was cut in half and swung to a loss, isn't that a signal that semiconductor capital investment has peaked and is rolling over?" This drags the K's lower arm into a forecast of the future.

As we saw, equipment revenue is a lagging indicator. Today's number is the shadow of past orders, not a preview of orders to come. If anything, with the device side doing this well, it will soon lay down new lines with that money, and those orders may register as equipment revenue several quarters out. That is why today's weakness can't be used to read the future.

I'll add two more caveats, honestly. The equipment companies we looked at are just three domestic listed firms, and even their direction split (Wonik swung to profit), so these three can't deliver a verdict on "the whole equipment industry." Materials firms, the firms that do assembly and testing on contract, design-only firms — none are in the sample at all. And Hanmi Semiconductor's fall to a third in a year is a single quarter's number. Equipment revenue swings hard depending on which quarter one big contract lands in, so a one-quarter plunge is not the same as a trend decline.

The remaining three objections were each handled earlier, so here I'll fold them into a line. Even granting a base effect, the K's device upper arm is a margin (SK 71.5%), not a growth rate, so it survives with the floor stripped away (the boom-is-real section). KOSDAQ weakness only rhymes the earnings K; it is not evidence (the market section). And Samsung's 42.8% is a whole-company number, which is exactly why the pure benchmark was set on SK hynix from the start (the "why Samsung is left out" section). None of the three shakes the K.

So this piece never speaks about the future to begin with. My claim is exactly one: "at this moment, this boom is not uniform." That is not a prophecy but a photograph.

Conclusion: "Record-Breaking Boom" Isn't Wrong — It's a Word That Lumps

Looking back, "the biggest semiconductor boom on record" is not a lie. Exports and device profits really did explode.

The trouble is that the one word lumps too much into a single mass. Inside it are two semiconductors — device shooting up and equipment bending down. And what mashes the two into one is not the reality of the industry but a single piece of statistical machinery: the sum.

So when you meet a headline, there's a habit worth forming — to split it once and ask "whose boom is this?" The side that sells the finished product, or the side that sells the machines that make it? Because inside the same word, opposite report cards can appear.

Table 3: How to Split a Single Headline

This K is a photograph of this moment. It does not mean equipment's weakness is the herald of a downturn. The photograph six months from now may look different again. But in this one frame, taken today, there are plainly two semiconductors.