The setup
Phison sold 26.8% more in the September quarter and its gross margin fell to 59.7% from 65.3%, which is what we said on it would do if the NAND cost layer had caught price. The call held. The way it held matters more than the grade, because gross profit rose NT$7.05B (about US$0.22B) on NT$18.19B (about US$0.57B) of extra revenue, so the margin on the extra revenue was 38.8%. It was 71.4% in June and 86.0% in March.
Our next call: December-quarter gross margin at or below 57.8%, which is what October's revenue run-rate gives if the increment stays at 38.8%. Above 57.8%, price or mix is outrunning the shelf again and we are wrong. The report should come in early , going by the 4Q25 deck, which came on .
The record is real and nobody should argue with it. Net income of NT$30.55B (about US$0.96B) is 3.49 times what Phison earned in all of 2025, and the deck says nine-month EPS of NT$325.49 exceeds the sum of the last decade. The argument is about what happens next, and the CEO has now joined it, telling the call that holding 60% is not reasonable and that keeping above 50% for the next few quarters would be excellent (unverified, press reports of the call). From 59.7%, that is 9.7 points of room by the company's own account. The shares closed at NT$1,500.00 on , before the results came out at the 17:00 Taipei call, so none of this is in the tape yet.
The numbers
Phison 3Q26 · NT$ billions unless stated · T-IFRS
| Line | 3Q26 | Y/Y | ||
|---|---|---|---|---|
| Sales | ||||
| Revenue | 86.08 | |||
| AI ecosystem modules (about) | 49.06 | +88% Q/Q | ||
| All other lines (about) | 37.02 | −12.0% Q/Q | ||
| Profit | ||||
| Gross profit | 51.39 | |||
| R&D | 15.29 | |||
| Operating income | 32.90 | |||
| Net income | 30.55 | |||
| Balance sheet | ||||
| Inventory | 111.77 | |||
| Receivables | 60.77 | |||
| Cash + current financial assets | 27.71 | |||
| Bonds payable | 25.65 | |||
Source: Phison 3Q26 announcement and results deck, (T-IFRS). The Q3 statements are due by , so cash flow, payables and contract liabilities are not yet filed. Notes ↓
The rest of this dive is for paid subscribers.
The headline numbers are above. The 12 sections below carry the mechanism, the peer read, the valuation work, and the dated tests that decide it.
- Seven quarters: the level before the delta
- The grade, in full
- What the AI line paid
- The line that saved the operating margin
- Who paid for the shelf
- What is already locked for 2027
- Connecting the dots
- The December bar
- Valuation
- The strongest case against this read
- What we'd watch
- Bottom line
Not ready to pay? Get the free Monday issue and one full dive a week by email.
Already a subscriber? Sign in.