Can Korea's Chip Exports Predict Samsung and SK hynix Earnings?
Korea publishes its semiconductor export total on the first day of every month. For a quarter, that means the full three-month figure is out four weeks before Samsung and SK hynix report. This post lines up four quarters of export data against the two companies' revenue and asks a simple question: what does the trade print actually tell you in those four weeks, and what does it not?
KEY TAKEAWAYS
1. Quarterly chip exports went from $46.4bn in 3Q25 to $113.9bn in 2Q26, a 2.5x rise. Samsung DS plus SK hynix revenue went from KRW 57.5tn to 206.8tn, a 3.6x rise. Same direction, different multiple.
2. July and August 2026 alone delivered $87.7bn, 77% of 2Q's total. If September matches the July-August average, 3Q lands near $131.5bn, up 15% QoQ. September's number is published October 1, about four weeks before full results.
3. Exports give you direction and acceleration early. They do not give you the level: currency, overseas fabs, foundry and System LSI, and the timing gap between contract prices and invoice prices all move the multiple every quarter.
What comes out when
Korean trade data arrives in three layers. Korea Customs publishes a 1-20 day flash around the 21st of each month. The Ministry of Trade, Industry and Resources (MOTIR) and Customs publish the monthly total on the first of the following month. The science ministry's ICT trade report follows mid-month with a memory and system breakdown. So the last month of a quarter prints the day after the quarter ends: October 1 completes 3Q.
Companies are slower. Samsung's preliminary results land in the first week after quarter end (July 7, April 7). Full results and the SK hynix release come in the fourth week (July 29-30, April 30). That leaves a window of about four weeks between the completed trade quarter and audited numbers. The question is how much to trust the trade print inside that window.
The raw series first. MOTIR's monthly semiconductor exports rose from $15.0bn in June 2025 to $46.7bn in August 2026, a 3.1x increase in 14 months. June, July and August 2026 were all above $40bn. Chips were 47.5% of Korea's $98.3bn total exports in August. This is the kind of series that is hard to find outside Korea and easy to use once you know the calendar.
Exports against revenue
The direction matches, the multiple does not
Summing months gives quarterly chip exports of $46.4bn, $53.8bn, $78.4bn and $113.9bn for 3Q25 through 2Q26. Samsung DS revenue over the same quarters was KRW 33.1tn, 44.0tn, 81.7tn and 127.5tn; SK hynix was 24.4tn, 32.8tn, 52.6tn and 79.3tn. Combined: 57.5tn, 76.8tn, 134.3tn, 206.8tn. Indexed to 3Q25, exports reached 245 and revenue reached 360.
Sequential growth rates make the relationship clearer. Exports: +16%, +46%, +45%. Revenue: +34%, +75%, +54%. When export growth jumped from 16% to 46% into 1Q26, revenue growth jumped from 34% to 75%. When export growth stalled at 45% in 2Q26, revenue growth slowed from 75% to 54%. Direction and acceleration show up in the trade data first.
Four reasons the multiple drifts
Currency first. Exports are in dollars, revenue in won. Samsung disclosed about KRW 3.1tn of positive FX effect in 2Q26 from a stronger dollar. Scope second. DS includes foundry and System LSI; the export line includes other companies' chips. Overseas fabs third. Output from SK hynix Wuxi and Samsung Xi'an sold locally never appears in Korean exports but does appear in revenue. Price timing fourth: quarterly contract prices are set early in the quarter, while customs values reflect invoice prices at shipment, so the two diverge in a fast-rising market.
Reading 3Q26 from here
July plus August is $41.0bn plus $46.7bn, or $87.7bn, already 77% of 2Q's $113.9bn. If September equals the July-August average of $43.8bn, 3Q comes in around $131.5bn, up 15% QoQ. September's figure prints October 1, with a 1-20 day flash around September 21 to check direction first.
What that tells you: 3Q revenue rises versus 2Q, and the growth rate steps down sharply from the mid-40s of the prior two quarters. That agrees with TrendForce's 3Q26 conventional DRAM contract forecast of +13-18%. What it does not tell you is the level. The revenue-to-export multiple ran 0.124, 0.143, 0.171, 0.182 (KRW tn per $100m) over the four quarters, rising every time. Whether it keeps rising is a separate call about FX, overseas volume and LTA share.
What I actually watch
| Checkpoint | What it settles | Date |
|---|---|---|
| 1-20 day flash | Daily-average comparison with Jul-Aug narrows the 3Q landing zone; Chuseok (Sep 24-26) cuts working days, so use per-day figures | around Sep 21 |
| September exports | 3Q total: near $131.5bn is the +15% case; below $125bn is single-digit growth | Oct 1 |
| Samsung preliminary | The gap between export-implied and actual revenue growth is the size of FX, overseas-fab and LTA effects; log it as next quarter's multiple | around Oct 7 |
| ICT trade report | Memory vs system split and $/kg let you separate price from volume | mid-October |
Value chain read-through
| Question | Can trade data answer it | What to add |
|---|---|---|
| Is revenue rising | Yes, four weeks early, with a strong directional hit rate | Nothing |
| Is growth decelerating | Yes, the QoQ change in exports leads the QoQ change in revenue | TrendForce contract forecasts |
| What is the revenue number | No, the multiple moves each quarter | Average FX, overseas fab share, LTA commentary |
| What about operating profit | No, cost timing, inventory valuation and FX gains sit outside trade data | Preliminary results, consensus |
Risks to this view
- Four quarters of sample, all in an upcycle. The multiple may behave differently on the way down.
- SK hynix 3Q25 revenue is rounded from the company release; the 3Q25 export sum uses MOTIR's July ($14.71bn), August ($15.1bn) and September ($16.6bn) prints, which can differ slightly from Customs' final figures.
- The 0.124 to 0.182 multiple is not FX-adjusted. Deflating by average won-dollar rates would flatten the trend.
- The September run-rate assumes the July-August average; fewer working days around Chuseok make a lower print likely.
Closing
Trade data is not an earnings oracle. It is a directional signal that arrives four weeks early. Trust the direction and the acceleration, wait for the company for the level.
Sources: MOTIR monthly export releases (Jul 1 and Aug 1, 2025; Oct 1 and Nov 1, 2025; Jan 1, Aug 1 and Sep 1, 2026 and others); Korea Customs monthly final figures; Samsung Electronics 4Q25 IR presentation and 1Q26, 2Q26 results (Newsroom); SK hynix 4Q25, 1Q26 and 2Q26 results (Newsroom); TrendForce press release Jul 3, 2026.
Comments
Post a Comment