Everyone says AI server board orders are weakening—that's why Wus Printed Circuit (002463) fell that hard. Kingboard Holdings (2368.TW) will announce October revenue before November 10; if year-on-year growth is still positive, does that count as refuting this claim?

Everyone says AI server board orders are weakening—that's why Wus Printed Circuit (002463) fell that hard. Kingboard Holdings (2368.TW) will announce October revenue before November 10; if year-on-year growth is still positive, does that count as refuting this claim?

It has not been overturned. Elian Electronics' October figures were already posted on the Market Observation Post System this morning: single-month 99.13 yi Taiwan dollars, up 76.2% from the same month last year and up 6.4% from September. September's +55.5% looked like the first signal of weakening, and one month later it was back to +76.2%. This back-and-forth shows that single-month revenue can neither overturn the weakening thesis nor prove there is no weakening.

What really should be answered is not whether this number can overturn anyone, but what this wave of decline is actually pricing in.

October revenue was out this morning

The deadline for monthly revenue of Taiwan-listed companies is the 10th of the following month, and these companies all reported before the deadline: Nan Ya PCB on November 3, Taiwan Union Technology on November 5, Compeq on November 6, Unimicron and Elian Electronics on November 7. None of the October revenue figures is still pending.

Revenue is in yi Taiwan dollars; year-on-year and quarter-on-quarter are percentages.

The four companies making high-end boards, substrates and base materials all had year-on-year growth above 67%; Compeq, which makes consumer HDI, had only 24.7%. This is not a whole-industry weakening; it is tiering: the more high-end, the faster the growth; the more generic the board, the earlier the deceleration. Only by putting Compeq's 24.7% and Elian Electronics' 76.2% in the same table can you see what "weakening" would look like if it really existed.

Elian Electronics' cumulative revenue for the first ten months was 832.92 yi Taiwan dollars, +69.9% year-on-year. A growth rate of this scale cannot be explained by any quarter in the PCB supply chain in which "orders are weakening."

Its three months of revenue in the third quarter summed to 297.19 yi Taiwan dollars, up 22.4% from the second quarter and up 69.6% from the third quarter of last year. Under Taiwan stock regulations, the company's third-quarter report must be disclosed by November 14 at the latest; the latest complete financial statements available now are still the half-year report. That is, through October its revenue curve was still accelerating upward, only the income statement has not yet been updated.

The swing in single-month year-on-year growth is much larger than the weakening

Over ten months, the lowest was 53.9% (February) and the highest was 87.3% (May), a range of 33 percentage points. September's 55.5% sat at the lower edge, and October's 76.2% returned to the center. Using a 20-point difference between two adjacent months to judge a trend reversal is equivalent to explaining noise with noise.

Monthly revenue also contains three variables unrelated to demand. Working days: February had five fewer days than other months. Exchange rates: the New Taiwan dollar strengthened against the US dollar this year, so the same shipments converted into Taiwan dollars would shrink or be inflated. Customers' pull-in rhythm: quarter-end push and quarter-start pullback. A move in any one of these three can create a year-on-year difference of more than ten points.

Wus's own third quarter shows not a trace

Wus disclosed its third-quarter report on October 28; breaking out the single quarter:

Revenue is in yi yuan; the last three rows are percentages.

Single-quarter revenue was +90.0% year-on-year and +27.6% quarter-on-quarter; net profit attributable to the parent was +160.8% year-on-year and +60.6% quarter-on-quarter; gross margin rose 2.74 percentage points quarter-on-quarter and 8.17 percentage points year-on-year. If order weakening were really happening, the first things to move should be capacity utilization and gross margin, not to wait two quarters before talking about it. This company's gross margin in the third quarter was moving upward.

Three forward-looking items on the balance sheet also did not turn. Contract liabilities were 1.03 yi, up 56.2% from the beginning of the year; inventory was 71.39 yi, up 68.1% from the beginning of the year, but cumulative revenue over the same period rose 71.9%, so inventory grew slower than revenue and is not a buildup; construction in progress was 33.05 yi, up 34.4% from the beginning of the year, still expanding capacity.

The contract liabilities item should be discounted—for a company with 95 yi of single-quarter revenue, 1 yi of advance receipts is only a small fraction; it is more like a timing difference in customer payments, not an order indicator. It is less useful than looking at inventory and equipment under construction.

In the investor relations activity record of October 20, management wrote:

The company's overall capacity continues to operate at a high level, order demand is ample, and revenue and profit have hit record highs for several consecutive quarters.

This is the original text of the announcement. The same record also says that 1.6T switch supporting products have already shipped in volume, 224Gbps computing products are only "supplied in small batches," and 448Gbps is in preliminary research—progress is tiered, but the direction has not changed.

By structure, in the first half, data communication board revenue was 113.30 yi, accounting for 82.8% of revenue, with a gross margin of 43.3%; smart automotive board was 14.38 yi, with a gross margin of 19.0%. The two differ by 24 percentage points. As the AI business share rises, consolidated gross margin is naturally lifted; the third quarter's 44.02% came from this, not from cost savings.

Five peers, five statements, the same direction

All are single-quarter figures for the third quarter of 2026; the first two columns are percentages, and gross margin quarter-on-quarter is in percentage points.

Four of the five had quarter-on-quarter gross margin increases. The only one whose net profit attributable to the parent was halved quarter-on-quarter, Victory Giant Technology, looks like something went wrong, but turning to the income statement makes it clear: in the first half, fair value change gains were +5.95 yi, and in the third quarter alone they were -6.90 yi; in and out, more than 7 yi, which is book fluctuation of financial assets.

Excluding this, its third-quarter operating profit was around 12 yi, about the same as the second quarter, and gross margin still rose 2.1 points.

Among the five statements, not one has operating costs or gross margin pointing to order weakening.

From order intake to revenue recognition, two quarters lie in between

The chain is: cloud providers set annual capital expenditure → server ODMs give quarterly forecasts → PCB makers take orders (high-layer boards have long certification cycles, usually 12 to 16 weeks in advance) → production scheduling and material preparation (copper-clad laminate, electronic cloth, copper foil) → shipment, forming monthly revenue → entering the financial statements at quarter-end.

So October's monthly revenue corresponds to orders placed in July and August. If October orders truly weakened, the earliest it would show up is in December or January next year's revenue. Using October revenue to confirm October orders means there is a full quarter in between.

Financial statement line items also have a sequence. Weakening orders hit contract liabilities and backlog first, then inventory turnover days, and only last hit gross margin. None of these three line items at Wus has turned yet.

What's falling is the denominator

Unit: %.

Seven trading days, seven companies, and the declines all fall in the narrow range of 23.7% to 33.9%. Over the same period, the Philadelphia Semiconductor Index fell 12.5% in a single day on November 5, the largest single-day decline since March 2020; Nvidia fell 13.3% from October 28 to November 6, Celestica fell 31.1%, and Vertiv fell 24.7%.

The seven companies are in two markets, with different customer structures and different products; it is impossible for them to simultaneously show weakening orders during the same seven days. The only thing that could synchronize them is the denominator.

Goldman Sachs provided a reference point: the forward P/E of its "China AI stock basket" was compressed from a July peak of about 53x to about 25x, and the basket index fell about 55% from its early-July high (IPO Daily, November 7, 2026). Profits are still rising, but the multiple has been cut in half. That is what this decline is doing.

There is one more detail. On October 29, the day after Wus released its third-quarter report, five A-share PCB makers fell together: Wus -10.0%, Shennan -9.4%, Shenghong -11.1%, Shengyi Electronics -11.4%, and Jingwang -10.0%. That day Wus delivered a report with single-quarter profit +160.8%. On the day with the best report, it fell as much as those with the worst reports; clearly it is not the report that is setting prices.

On November 6, the chairman proposed a buyback of 500 million to 1 billion yuan. The company's own attitude toward its share price is written in the announcement.

There is another layer that has not been priced. Kingboard's October revenue was posted on Saturday morning; Taiwan stocks would not trade until Monday, November 9, and A-shares opened the same day. The market has still not voted on this +76.2% reading. How it moves next Monday is more direct than any analysis.

The strongest opposing argument

If the October revenue of all four Taiwanese companies is above 67% year-on-year, then "orders weakening" is wrong as of October, and using it to explain the share price is even more wrong. I accept that.

What it overturns is "already weakened"; it cannot overturn "will weaken." The market is trading the latter: cloud providers' capital expenditure growth in 2027 falling from this year's high level, AI server boards going from supply shortage to supply-demand balance, and annual price cuts resuming. This judgment naturally cannot be seen in October and November shipment data, because it is about next year.

The second opposing point is sharper: Kingboard's strength in server boards does not mean Wus is strong in switches. Wus's own records state clearly that the strongest growth is in high-speed network switches and supporting routers, while computing-power products are still in small-batch production. Using Kingboard's server board orders to infer Wus's switch orders leaves a gap larger than imagined.

The third is cost. In this round, upstream electronic cloth, copper-clad laminate, and copper foil are all rising in price, and some in the industry have already voluntarily reduced low-gross-margin orders. Material price increases will hit gross margin first, not orders first. So the second number that should really be watched is fourth-quarter gross margin, not revenue.

What numbers would make me admit I am wrong

• Kingboard's November and December revenue both post negative quarter-on-quarter growth for two consecutive months, and year-on-year falls below +40%; then weakening orders turns from speculation into fact.
• Wus's single-quarter revenue in the fourth quarter of 2026 turns negative quarter-on-quarter, or single-quarter gross margin falls by more than 2 percentage points quarter-on-quarter, indicating that the breath it took in September did not carry through. These two numbers can be found in the annual report in April 2027.
• Wus's annual report shows inventory turnover days clearly lengthened and contract liabilities turning downward, indicating that the goods prepared did not turn into revenue.

Conversely, if Wus's single-quarter revenue in the fourth quarter is still growing quarter-on-quarter and gross margin holds at 44%, then weakening orders is disproven, and what is falling in this wave is valuation, not the business.

The next checkpoint is very close: before December 10, Kingboard's November revenue; next April, Wus's annual report.

Source: Wus Printed Circuit (Kunshan) Co., Ltd. "2026 Third Quarter Report," October 28, 2026, Announcement No. 2026-076; Wus Printed Circuit (Kunshan) Co., Ltd. "2026 October 20 Investor Relations Activity Record Form," October 20, 2026, No. 2026-1020-042; Wus Printed Circuit (Kunshan) Co., Ltd. "Indicative Announcement on the Chairman's Proposal to Repurchase Company Shares," November 7, 2026, Announcement No. 2026-078; Taiwan Stock Exchange Market Observation Post System "Listed Companies October 115 Operating Revenue Statistics Table," table release date November 7, 2026; Gold Circuit Electronics (2368) monthly revenue from January 2024 to October 2026 and consolidated income statements from 2024Q1 to 2026Q2, Market Observation Post System original data obtained via FinMind; Shennan Circuits "2026 Third Quarter Report," October 23, 2026; Victory Giant Technology "2026 Third Quarter Report," October 30, 2026; Shengyi Electronics "2026 Third Quarter Report," October 30, 2026; Kinwong "2026 Third Quarter Report," October 27, 2026; "Orders Growing but Materials Rising in Price: PCB Industry Chain Explores 'Seeking Efficiency from Processes'," Securities Times, November 7, 2026; "'Yi Zhong Tian' All Say Operations Are Normal — Are Optical Shares Still Worth Holding," November 7, 2026; Philadelphia Semiconductor Index and daily charts of Nvidia, Celestica, and Vertiv, Sina Finance US Stock Quotes, October 28 to November 6, 2026.