WPG Holdings Q2 2026 — The Record Was Borrowed Twice: From September, and From the Banks

A beat that isn't in dispute

Start with what checks out, because all of it does. WPG guided second-quarter revenue to NT$345B to NT$365B in May and reported NT$456.21B on the reviewed IFRS statement, 25.0% over the top of its own range (about US$11.34B at that top). Every US$ figure on this page uses about NT$32.2 per US$, the company's own Q3 planning rate, and is approximate. Net income was NT$8.62B (about US$0.27B) against a guide of NT$6.147B to NT$6.824B. EPS was NT$5.12 (about US$0.16) against NT$3.65 to NT$4.06. Nobody should argue with that.

One housekeeping point before the argument, since two revenue numbers are in circulation. The monthly releases sum to NT$458.28B (about US$14.23B), which is where the +44.8% QoQ in the June release comes from; the statement says NT$456.21B. The NT$2.07B gap (about US$0.06B) is 0.45% and the release doesn't explain it. In the three quarters before, the two bases matched to the rounding. I use the reviewed statement for the quarter, and monthly figures only for Q3, where nothing else exists yet. That choice matters later, because it means the monthly bar for September is if anything a little higher than the arithmetic below shows.

Management already told you what June was

June revenue was NT$192.92B (about US$5.99B), a record and 1.45 times the average of April and May. July came in at NT$146.10B (about US$4.54B) and August at NT$142.24B (about US$4.42B), down 2.64% on July. Two months of Q3 add up to NT$288.33B (about US$8.95B). So September needs NT$171.67B (about US$5.33B) for the low end of the guide and NT$191.67B for the midpoint. The midpoint needs another June.

Here is what management said about June on the August 18 call, per a third-party memo (unverified, no official transcript): customers "pulled orders forward at the end of the second quarter in anticipation of shortages and price increases, so third-quarter revenue is expected to grow by a single-digit percentage quarter on quarter." Read that sentence for what it concedes. A pull-in is revenue borrowed from a later month. If the end of June borrowed from July and August, the flat NT$146.10B and NT$142.24B are what the repayment looks like, and a September that repeats June would need a second pull-in on top of the first. That is possible. It's not the base case.

The series tells you which quarter-end month to expect. September 2025 and December 2025 each ran 1.15 times the first two months of their quarters. March 2026 ran 1.62 times, but February had been shortened by the Lunar New Year. June ran 1.45 times, the quarter management calls a pull-in. Q3 needs 1.19 times for the low end and 1.33 for the midpoint. One month is a phase until you have the series, and the series says the two clean quarter-end months ran at 1.15. As arithmetic, not a forecast, that ratio puts September near NT$165.8B and Q3 near NT$454.1B, below the guide. The last time Q3 looked like this, WPG guided NT$245B to NT$265B and reported NT$244.47B.

The core business, and the line under it

The results release says net income broke its record "driven by core business momentum" (translated). The income statement agrees only in part. Gross margin was 4.43%, near the top of guide and 6 basis points under Q1. Operating expenses rose 43.8%, in step with revenue, so the operating margin of 2.62% came in below Q1's 2.68%. Operating income of NT$11.95B (about US$0.37B) sat inside its guided margin band. The beat over the top of the net income guide was NT$1.80B, and investment income alone rose NT$1.41B quarter on quarter, to NT$2.00B (about US$0.06B), or 17.7% of pre-tax income. The core business did beat, on volume. The size of the net income beat came from below the operating line.

The Q3 guide agrees with me. Its pre-tax range implies a non-operating loss of NT$1.36B to NT$1.65B against NT$0.66B in Q2, which is what the quarter looks like without a repeat of the investment gain and with net interest, already NT$2.86B (about US$0.09B) and up 32.5%, still climbing. The operating margin step to 2.78% to 2.91% leans on operating expenses falling 6.6% at the midpoint while revenue rises. That's a plan. It is not yet a print.

The warehouse is the position, and memory is the warehouse

The release credits "strong pull-in momentum in power management, servers, networking equipment, energy storage systems and high-density connectors" (translated). Memory isn't on the list. The deck's product split puts memory at 54% of Q2 revenue, from 45% in Q1 and 26% a year ago, while core components fell to 18% from 40%. The AI label is real but smaller than the headline: computing including servers was 45% of revenue by application, down from 49% in Q1. "Near half" was the May story. In June it moved the other way.

For a distributor in a price spike, the warehouse is the position. The 4.43% gross margin is a spread between what the memory cost and what it now clears at, and the inventory on the balance sheet is the asset that earned it. That cuts both ways. Customers who pulled orders forward because they expected shortages and price increases are customers who stop pulling when the price stops rising, and nothing in WPG's guide tells you when that is.

The banks funded the quarter

Operating cash flow was negative NT$59.19B in the quarter, against positive NT$23.19B a year earlier, and negative NT$98.90B (about US$3.07B) for the half. Net operating assets absorbed NT$75.10B (about US$2.33B) on the cash flow statement. New borrowing of NT$63.15B (about US$1.96B) covered 84.1% of it.

The balance sheet shows where it went. Receivables were NT$294.80B (about US$9.16B), inventories NT$215.02B (about US$6.68B) and payables NT$212.44B (about US$6.60B), so net operating assets reached NT$297.38B (about US$9.24B). Through 2025 that line sat between NT$171.93B and NT$192.64B. Then it went to NT$222.49B in March and NT$297.38B in June, NT$109.04B in six months. Short-term borrowings went from NT$114.55B to NT$204.91B (about US$6.36B) over the same stretch, 2.07 times equity of NT$98.94B (about US$3.07B), against 1.34 times in December.

The model I built for TD SYNNEX fits a Taiwan distributor just as well, the ODM pays three ways. The growth is paid for in working capital, in bank debt, and in equity, through convertibles. The equity leg is already running. Share capital rose to NT$17,285M from NT$16,791M in the quarter as the second and third convertibles converted, per the dividend notices (unverified — third-party reproduction). The fourth is on the way.

The composite hides the component

The deck's headline working-capital measure improved: a net operating cash cycle of 52 days, from 59 in Q1 and 67 a year earlier. That is a real improvement, and it's smaller than it looks. The deck's days are annualised on average balances, and in a quarter when balances jump, the average lags the ending balance. The composite hides the component. Recomputed on period-end balances against each quarter's own revenue and cost of revenue, the cycle is 59.3 days in Q2 (receivables 58.8, inventory 44.9, payables 44.3), against 63.4 in Q1 and 62.7 a year earlier. A few days better, not fifteen. The working capital didn't balloon because the cycle got worse. It ballooned because volume did, and volume at a 4.43% gross margin needs roughly a quarter's worth of cost of goods tied up at all times. The cycle is the efficiency. The balance is the bill.

A convertible sized for a different problem

On August 11 the board approved CB4: up to NT$15B (about US$0.47B), three years, a 0% coupon, proceeds to repay bank loans. The release explains the choice: the board "judged that a CB has a lower equity-dilution effect than global depositary receipts (GDR)" (translated). Two things in that sentence. Straight equity was on the table, which tells you how management sees the balance sheet. And the instrument chosen is a refinancing, not new money for growth. At its cap, CB4 is 20.0% of one quarter's increase in net operating assets and 7.3% of short-term borrowings. It swaps a slice of floating bank debt for zero-coupon paper that converts if the stock cooperates. It doesn't change the shape of the funding. The dividend of NT$6.38B (about US$0.20B), paid on September 30 (unverified — amount from a third-party reproduction), comes out of the same pool of borrowed cash in Q3.

The strongest case against me

Here is the other side at full strength, and it's not weak. WPG beat the top of its revenue guide by 15.1% in Q1 and 25.0% in Q2. It set the Q3 range on August 18 with July's NT$146.10B already booked and most of August in view, so the NT$460B floor is a number management chose while looking at the same flat months I am. The planning rate moved to NT$32.2 from NT$31.6, which adds to reported revenue on its own. Memory is 54% of the mix and memory prices are the thing everyone agrees is rising, into the first half of 2027 on management's own telling (unverified, third-party memo). And March showed that this business can put 1.62 times its first two months into a quarter-end. Believe all that, and September at NT$191.67B is a normal quarter-end for a company in a shortage, not a second pull-in.

I still come down on the other side, for one reason. The two quarter-end months that were not distorted by a holiday or a pull-in ran at 1.15, and the company itself told you June was a pull-in. A guide set in August can price the price. It can't un-borrow the orders.

What decides it

One number, on one date: September revenue, due by October 10, 2026. Below NT$171.67B, Q3 misses its own low end on the monthly basis, the first miss since Q3 2025, and I'm right with room to spare. Between NT$171.67B and NT$191.67B, Q3 lands in the bottom half of the guide and I'm right. At or above NT$191.67B, the June level held without a pull-in to explain it and I'm wrong. Remember the basis note from the top: the Q2 statement came in NT$2.07B under the monthly sum, so a borderline September is weaker than it looks.

The second check comes with Q3 results in November. Net operating assets against NT$297.38B and short-term borrowings against NT$204.91B. If revenue only grows single digits and working capital still rises, the pull-in has turned into inventory. What I can't determine from anything published is how much of the June receivables were memory bought at the old price and sold at the new one. That is the question the next statement answers and no analyst on the call appears to have asked.

The trade (analysis, not advice)

WPG at this point is a memory-price instrument wearing a distributor's balance sheet. The upside case is memory prices; the carrying cost is bank debt that grows with every quarter the price rises. That's an asymmetric setup into October 10: a September near the low end resets a guide the market has been reading from June, while a September at or above the midpoint mostly confirms what the Q3 guide already says. I wouldn't pay for the top of the guide before the number prints. Anyone who wants the memory cycle without carrying NT$297.38B of someone else's inventory and receivables can take it one layer up through picks-and-shovels capture, from the makers that set the price. WPG holds the spread, and the spread is what reverses first.

Bottom line

WPG printed its best quarter and borrowed twice to do it. It borrowed demand from the months after June, by management's own account, and it borrowed the working capital from its banks, NT$63.15B of it in one quarter, with a NT$15B convertible to refinance a slice. My view is that Q3 lands in the bottom half of the guide. The number that decides it is September revenue below or above NT$191.67B, published by October 10.