The desk judged the 11-06 decline as a valuation contraction across the entire AI hardware chain, not a performance issue. To overturn it, the key is the October revenue that TAIWAN UNION TECHNOLOGY (2383.TW) will announce on November 10: below what year-on-year level would invalidate this judgment?

The desk judged the 11-06 decline as a valuation contraction across the entire AI hardware chain, not a performance issue. To overturn it, the key is the October revenue that TAIWAN UNION TECHNOLOGY (2383.TW) will announce on November 10: below what year-on-year level would invalidate this judgment?

The premise of this is invalid: Taiwan Union Technology (2383.TW)'s October revenue was disclosed on the Market Observation Post System after market close on November 5, not November 10. 244.73 yi, up 203.08% year-on-year, setting a new high for the eighth consecutive month. What is due on November 10 is TSMC (2330.TW)'s October revenue.

So this is not a forecasting question, it is an arithmetic question. For the judgment that "what fell on November 6 was positioning and sentiment, not facts" to be invalidated, October revenue year-on-year would have to be below +151.3%. The actual figure is +203.08%, 51.8 percentage points higher. The judgment was not overturned; it was nailed down.

What must be waited for on November 10 is another company

The two names differ by a single character: TSMC, Taiwan Union Technology. The former does wafer foundry, the latter does copper-clad laminate.

Taiwan listed and OTC companies file last month's revenue before the 10th of each month. Taiwan Union Technology was five days early this time, disclosing after market close on November 5; the stock had already fallen 9.99% that day, closing at 4,100 yuan. At that point no one had seen the October revenue.

On November 6, with the numbers in hand, the stock closed at 3,820 yuan, down another 6.83%. The same day the TAIEX weighted index fell 4.51%.

What fell was the multiple, not EPS.

Using the same trailing-four-quarter EPS of 93.36 yuan (10.44 yuan in Q4 2025, plus 82.92 yuan for the first three quarters of this year), the August 17 high of 6,350 yuan corresponds to 68.0x, and the November 6 price of 3,820 yuan corresponds to 40.9x. The stock has pulled back 39.8% from the high, and EPS has not fallen by a single cent. Revenue went from 192.07 yi in July to 244.73 yi in October, up 27.4%. Price fell faster than revenue; that is the definition of valuation compression.

Single-quarter breakdown: gross margin speaks before revenue

Q3 revenue rose 28.8% quarter-on-quarter, EPS rose 46.9% quarter-on-quarter, a gap of 1.6x. The entire difference came from gross margin rising 3.85 percentage points quarter-on-quarter. Operating expenses rose from 32.69 yi to 42.64 yi, a growth rate of 30.4%, roughly in line with revenue—neither a drag nor a help.

Operating profit rose 59.53 yi quarter-on-quarter, revenue rose 136.10 yi quarter-on-quarter; for every extra 100 yuan of revenue, 43.7 yuan fell through to operating profit. That is the marginal economics of this business.

Year-on-year is even more direct: Q3 revenue rose 142.1%, gross margin was 7.57 percentage points higher than the same quarter last year, and EPS rose 261.7%. No line item in the Q3 report is moving in the wrong direction.

The shape of this business is right there in the table. Formulation and certification lock in customers, new capacity takes years from decision to mass production, and pricing power is its own. Revenue up 30%, profit up 45%, and the piece in between is operating leverage.

This one October number writes half of Q4

October revenue was 244.73 yi, up 13.66% quarter-on-quarter. The quarter-on-quarter figures from March to September this year were +17.8%, +15.9%, +12.2%, +13.5%, +8.3%, +4.9%, +6.9%. October's +13.7% is the steepest since April.

Even if November and December stay flat, Q4 revenue is 734.18 yi, up 20.6% quarter-on-quarter and up 194.5% year-on-year. Q3's 608.85 yi is the sum of July, August, and September, matching the financial report exactly; this monthly revenue series can be trusted for now.

Peers: three companies hit new highs together

ITEQ's gross margin was 22.02% in Q2 and 34.33% in Q3; Taiwan Union Technology's rose from 29.71% to 36.93%. All three raised margins by 3.9 to 12.3 percentage points in the same quarter. This is not one company's product mix; it is the entire copper-clad laminate chain raising prices.

The next layer down checks out too: Unimicron's October revenue was 189.86 yi, up 67.6% year-on-year; Nan Ya PCB 62.89 yi, up 69.7%; Kinsus 53.68 yi, up 49.1%; Gold Circuit Electronics 99.13 yi, up 76.2%. Not one player in substrates and PCB is lagging.

Looking left to A-shares, the same direction. Shengyi Technology's Q3 single-quarter gross margin was 41.63% (35.36% for the first three quarters, 30.69% for the first half on a restated basis), Wus Printed Circuit's single-quarter was 44.02%. On November 6, Shengyi fell 9.37%, Wus fell 9.62%, Zhongji Innolight fell 9.19%, and Yuanjie fell 8.43%.

Prices are falling, while gross margins in contemporaneous reports are rising. Both things being true at the same time is what valuation compression means.

Transmission chain: where the money stops at each layer

Cloud capex sets the total, which flows to AI racks and 800G, 1.6T switches, then to the layer count of high-speed boards, then to the grade of copper-clad laminate. The higher the grade, the fewer the companies that can make it.

Monthly revenue flows onto the financial statements in three stages: deliveries record as operating revenue, price increases record as gross margin, and capacity utilization records as operating margin. All three stages lag by one to two quarters. Q3 was positive on all three at once: revenue up 28.8%, gross margin up 3.85 percentage points, operating margin up 3.76 percentage points.

What breaks first is never revenue, it is gross margin. Revenue drops 5%, gross margin may drop only 1 point, and operating profit will drop 9%—42.64 yi of quarterly operating expenses is sitting right there, and it cannot be spread thinner.

The bear case: revenue is the rearview mirror

The strongest rebuttal is this: the limit-down on November 5 happened before October revenue was published. The explanation the media gave that day was the "order transfer falling through" after the fire at Panasonic's Hukou plant, not earnings. Counting both days as valuation compression means using an order event to explain a multiple decline.

It is half right. On November 4, there was a fire at Panasonic's Hukou plant; the market's first reaction was that Taiwanese makers would pick up transferred orders, and Taiwan Union Technology rose 3.64% that day. On November 5, word circulated intraday that order transfers were below expectations, and Taiwan Union Technology and Taiwan Paiho both hit limit-down. This is order expectations moving — an earnings variable, not a multiple variable.

But it cannot overturn "November 6 was not earnings." If what was being sold was already-reported earnings, the stock should have risen on the day October revenue hit a record; in reality, with the number in hand, it fell 6.83%, underperforming the broad market by 2.3 percentage points. On the best earnings day, the stock looked the worst — this can only be the multiple moving.

There is a more technical point: the year-on-year base is rising. Last November was 81.29 yi yuan and December 87.05 yi yuan, both above October's 80.75 yi yuan. Even if revenue stays flat, November year-on-year would fall from 203.08% to 201.1%, and December to 181.1%. Using "year-on-year no lower than some number" as a test is itself dragged along by the base. So the line below is derived backward from profit, not set from the feel of the year-on-year figure.

Falsification condition: how the +151% line is calculated

Third-quarter operating profit was 186.87 yi yuan, operating expenses 42.64 yi yuan, gross margin 37.70%. To hold operating profit at 186.87 yi yuan in the fourth quarter, revenue needs to be (186.87+42.64)÷37.70%=608.8 yi yuan, exactly the third-quarter level.

608.85 yi yuan divided by 3 gives a monthly average of 202.95 yi yuan. Last October was 80.75 yi yuan; 202.95÷80.75−1=+151.3%.

Conclusion: if October revenue year-on-year is below +151.3% (single month below 202.95 yi yuan), fourth-quarter operating profit turns negative quarter-on-quarter, shifting from accelerating to decelerating, and the statement "the facts haven't moved" does not hold. Actual was 244.73 yi yuan, 41.78 yi yuan above this line.

This is not the dividing line between "good earnings" and "bad earnings." For October year-on-year to turn negative, the single month would have to fall below 80.75 yi yuan — two-thirds less than the current 244.73 yi yuan. That threshold is too far from the present to test the November problem. +151.3% is the dividing line between "still accelerating" and "starting to decelerate," and that is the test this judgment needs.

Three supplements; without any one of them this line does not count:

• A single month does not count; October and November must both be below 203 yi yuan for two consecutive months. One month is delivery rhythm; two months is orders.
• If fourth-quarter gross margin falls below 35% (third quarter was 37.70%), even if revenue holds, profit will not hold.
• TSMC's November 10 report. Same calculation: third-quarter revenue was 14942.43 yi yuan; for the fourth quarter not to decline quarter-on-quarter, the monthly average needs to be 4980.81 yi yuan; last October was 3674.73 yi yuan, so year-on-year cannot be below +35.5%. It is at the very upstream; if it falls behind, the good days for copper-clad laminate are water without a source.

Where this judgment is truly fragile

October revenue proves current-period earnings, not expected earnings. It cannot measure 2027 order share, nor can it measure who customers gave the orders to after the order transfer fell through. This judgment uses a delivery data point to fight an order expectation; logically it is asymmetric.

But as far as November 6 itself is concerned, every number that can be verified against filings is moving up: revenue, gross margin, operating margin, peers, downstream. No filing can support the words "earnings turning bad." The term valuation compression by nature describes only the price, not the company.

The next gate is TSMC's October revenue on November 10; the one after that is electronic cloth pricing in early December. Before these two numbers come out, this judgment can only be considered not yet overturned, not closed.

Sources: Taiwan Union Technology October 2026 consolidated revenue filing (Market Observation Post System, disclosure date 2026-11-05, monthly revenue filing has no announcement number); Taiwan Union Technology third-quarter 2026 consolidated financial report (balance sheet date 2026-09-30, published October 2026); Elite Material, Taiwan Paiho, Unimicron, Nan Ya PCB, Kinsus, Gold Circuit Electronics October 2026 consolidated revenue filings (Market Observation Post System, disclosure dates 2026-11-03 to 2026-11-07, monthly revenue filings have no announcement numbers); TSMC September 2026 consolidated revenue filing (Market Observation Post System, 2026-10-08, monthly revenue filing has no announcement number); Shengyi Technology third-quarter 2026 report (2026-10-30, announcement number 1225644504); Wus Printed Circuit third-quarter 2026 report (2026-10-28, announcement number 1225627804); Taiwan Stock Exchange weighted stock index close on November 6, 2026; Economic Daily News 2026-11-04 "Taiwan Panasonic Multilayer Materials Hukou Site Production Line Damaged"; Business Today 2026-11-05 "What Happened When Taiwan Union Technology Hit Limit-Down Again."