The setup
The headline says best quarter ever. The filing says something narrower and more useful: Macronix sold fewer NOR bits in the third quarter than in the second and still posted the best quarter in its history, so every dollar of the record came from what it charged, and nothing in the quarter locks that charge in. Chairman Miin Wu said it in one line on the call: “現階段營運成長主要來自產品價格提升” — growth at this stage comes mainly from higher product prices. The deck agrees. The NOR density shipment index fell from 242 to 213, down 12.0%, while parent-company NOR revenue rose 21%. NAND, including eMMC, rose 39% and passed NOR as the largest product line, 47% of parent sales against 46%.
That makes this a different print from the one three months ago. In Q2 the story was a mature-node fab that caught the cycle: 48- and 96-layer 3D NAND and 55nm NOR, the nodes the leading fabs walked away from, repriced while Macronix had a nearly fully depreciated plant to sell from. In Q3 the plant is full, the company is moving existing NOR capacity to NAND and eMMC, and the new capacity it is paying for does not ship until the second half of 2027. Wu again: “目前產能已相當有限” — capacity is already quite limited.
So the quarter has three ingredients worth separating: price (most of it), mix (NAND and eMMC taking wafers from NOR), and a NT$750M inventory write-back inside gross profit. And it has one structural fact that the Q2 dive flagged and nothing since has changed: Macronix prices monthly and has no meaningful long-term-agreement book. Nanya, reporting the same quarter, said its price rise “primarily reflect[s] renewed long-term agreements.” Decode: Nanya's customers have committed to its price; Macronix's have only paid it. Macronix's rent is still the market's, renewed every month. That is why the monthly revenue file is the right instrument to grade this call, and why the stock closed at NT$103.00 on the day of the print, 44.5% below its high of NT$185.50, while earnings per share rose from NT$3.91 to NT$5.52.
The numbers
Net sales were NT$24,237M (~US$763.6M), up 27% from NT$19,125M and 195% from NT$8,214M a year earlier. Gross profit was NT$17,086M at a 70.5% gross margin, against 64.4% in Q2 and 13.5% in 3Q25. Operating expenses were NT$4,153M, operating income NT$12,932M at a 53.4% operating margin, and net income NT$10,951M (~US$345.0M), up 42% on the quarter. Basic EPS was NT$5.52 on 1,983M weighted shares. EBITDA was NT$14,377M, a 59% margin. Book value reached NT$36.95 per share.
| Net sales | 24,237 · +27% QoQ · +195% YoY |
| Cost of sales (net sales − gross profit) | 7,151 · +5.1% QoQ (2Q26: 6,805) |
| Gross profit · gross margin | 17,086 · 70.5% (2Q26: 64.4%; 3Q25: 13.5%) |
| Inventory write-down reversal in gross profit | ~750 (CFO on the call; 2Q26: ~900) |
| Operating expenses | 4,153 · +24% QoQ · 17.1% of sales |
| Operating income · operating margin | 12,932 · 53.4% (2Q26: 46.9%) |
| Non-operating income · income tax | 65 · (2,046) |
| Net income · basic EPS | 10,951 · NT$5.52 (2Q26: NT$3.91) |
| EBITDA · EBITDA margin | 14,377 · 59% |
| Operating cash flow · capex · free cash flow | 13,392 · (4,354) · 9,038 |
| Cash · interest-bearing debt · inventory | 29,564 · 20,287 · 11,537 |
| Product mix (parent): NAND · NOR · ROM | 47% (+39% QoQ) · 46% (+21%) · 5% (+9%) |
| 1Q–3Q 2026 net sales · EPS | 53,831 (+155% YoY) · NT$10.34 |
Source: Macronix 3Q26 investor-conference deck, (MOPS). Write-back figures are from the CFO's remarks as reported by the Taiwan press. Cost of sales is derived.
Two lines in that table matter more than the headline. Cost of sales rose 5.1% while net sales rose 26.7%. Gross profit rose NT$4,766M on NT$5,112M of extra revenue, an incremental gross margin of 93.2%. Operating income captured NT$3,964M of the same NT$5,112M, a 77.5% incremental operating margin. A business whose extra revenue keeps 93 cents in gross profit is not selling more product. It is charging more for the product it already makes.
Second, the write-back. CFO Paul Yeh said roughly NT$750M of gross profit came from reversing earlier inventory write-downs, tied mainly to legacy ROM and NAND stock. Strip it out and the gross margin is 67.4%, not 70.5%. Do the same to Q2, which carried about NT$900M, and Q2's margin was 59.7%. Decode: the reported margin rose 6.1 points, but the margin after removing write-backs rose 7.7 points. The write-back is shrinking while the business underneath it improves faster than the headline says. The library's rule that a refund is not a margin applies here in reverse: a recovery is inflating the level, but a smaller one is understating the trend. Wu said reversals continue next year, with limited contribution.
Operating expenses rose 24% to NT$4,153M, NT$801M more than Q2, while sales rose 27%, so opex fell to 17.1% of sales from 17.5%. Wu explained the step: “公司將獲利的15%與員工分享” — the company shares 15% of profit with employees — and the CFO said actual spending did not rise noticeably. Decode: the opex line is now a profit-sharing accrual that scales with the quarter, not a cost base. It will fall if profit falls, which cushions a downturn a little, and it will rise every quarter the price does, which caps how much of a further price increase reaches operating income.
The year-to-date total already re-ranks the company. Nine-month net sales of NT$53,831M exceed the NT$50,573M Macronix booked in all of 2021, its previous record year. Nine-month EPS is NT$10.34 against a loss of NT$1.62 a year earlier.
The monthly file reconciles exactly. July NT$7,724.73M, August NT$8,145.27M and September NT$8,367.28M sum to NT$24,237M, the reported quarter. Each month was a record, but the steps are shrinking: July rose 11.0% on June, .4% on July, .7% on August. Hold that sequence; it is the one the rest of this page keeps coming back to.
The rest of this dive is for paid subscribers.
The headline numbers are above. The 9 sections below carry the mechanism, the peer read, the valuation work, and the dated tests that decide it.
- Price, not bits: reading the ingredients
- The cost line the capex will write
- The balance sheet
- Connecting the dots
- Valuation
- The counter-case
- What we'd watch
- The trade
- Bottom line
Not ready to pay? Get the free Monday issue and one full dive a week by email.
Already a subscriber? Sign in.