The Five Variables Behind SK hynix's Operating Profit
SK hynix earned KRW 60.5 trillion in operating profit in 2Q26 at a 76% margin, up from 19.2 trillion and 58% three quarters earlier. Headlines report the number; they rarely say which lever moved it. This post splits the profit into five variables and sizes each one using what the company said on its July 29 call.
KEY TAKEAWAYS
1. On the 2Q26 base (revenue 79.3tn, operating profit 60.5tn), a 10% move in one variable changes profit by: price +7.9tn, won-dollar rate +7.6tn, bit shipments +7.0tn, unit cost +1.9tn. Illustrative, but the ranking is the point: at a 76% margin, cost is a small lever and price and currency are large ones.
2. In 2Q, DRAM ASP rose about 30% and NAND in the mid-50s, while bit shipments grew high single digits and mid-teens. Price did most of the work in the 51% revenue jump. The company attributed the softer-than-expected DRAM ASP to high-value products slipping into 2H and to mix.
3. Cost is small today and growing. Quarterly D&A is 4.0tn (16tn annualized) against 2026 capex guidance in the high 40s, three times as much. Once M15X and the first Yongin fab (cleanroom early 2027) start depreciating, the same price buys a lower margin.
What the 2Q26 call put on the table
The company's own numbers first. Revenue KRW 79,319bn (+51% QoQ), operating profit 60,543bn (+61%), margin 76% (72% in 1Q). Depreciation and amortization 4.0tn; EBITDA 64.6tn at an 81% margin. DRAM bits up high single digits with ASP up about 30%; NAND bits up mid-teens with ASP up mid-50s. Below the operating line, a 1.1tn FX gain and a 63.3tn gain on investment disposals and revaluation (the Kioxia stake among them) took net income to 93.9tn.
Guidance: 3Q DRAM bits up about 10% QoQ, server-led; NAND up low single digits. 2026 capex in the high 40s of trillions of won. LTAs concluded with roughly ten customers including key accounts, typically five years, with deposit provisions. HBM4 in volume shipment since 2Q, ramping in 2H. Cash 88tn and net cash 69.4tn at quarter end.
One thing worth flagging: operating profit came in 6.5% below FnGuide's 64.7tn consensus. Asked why DRAM ASP rose less than expected, management said some high-value product shipments had slipped into the second half and that portfolio composition weighed on the blended ASP. Of the five variables, mix decided the 2Q surprise.
The five variables, sized
1. Price: the biggest lever, and decelerating
At a 76% margin nearly every extra won of revenue is profit. A 10% price move on the 2Q base adds 7.9tn of revenue with almost no added cost, so profit moves by about the same 7.9tn. That is why price is the longest bar in the chart. The direction is deceleration: TrendForce has conventional DRAM contract prices going from +58-63% in 2Q26 to +13-18% in 3Q26, and the company itself said HBM prices are not set by commodity DRAM alone. 2027 HBM pricing is being negotiated now.
2. Bit shipments: the one variable the company controls
A 10% rise in bits adds 7.9tn of revenue but also lifts variable cost, so profit rises by roughly 7.0tn if half of cost is variable. DRAM bits grew high single digits in 2Q and the 3Q guide is about 10%, so shipment growth is accelerating even as price growth slows. If price steps down to the low teens and bits deliver 10%, revenue still grows double digits. Pulling M15X forward and opening the Yongin cleanroom in early 2027 is about making this lever bigger.
3. Mix: the variable that decided 2Q
The same bit sells at different prices as HBM, server DRAM, SOCAMM2 or eSSD. In 2026 the sign of mix flipped. TrendForce reported that in 1Q26 revenue per wafer on HBM was overtaken by DDR5 64 GB RDIMMs, and Korean coverage tied the SK hynix shortfall to its high HBM and long-term-contract share, which kept commodity price gains from flowing straight into ASP. That is why the chart shows HBM mix +10 pp at roughly zero: HBM did not get worse, commodity DRAM got that much better. HBM4 volume in 2H and the 2027 HBM negotiation can flip the sign back.
4. Cost: small now, larger from next year
Cost of revenue in 2Q was about 18.8tn (revenue less operating profit), so a 10% cost cut adds only 1.9tn. The direction matters more than the size. Quarterly D&A of 4tn annualizes to 16tn; 2026 capex in the high 40s is three times that. New fabs bring step-ups in depreciation. The 1c node transition pushes cost per bit down; new-fab depreciation pushes it up. The net of those two sets the 2027 margin.
5. Currency: inside and outside operating profit
Memory revenue is mostly in dollars; a large part of cost is in won. A 10% weaker won adds 7.9tn of revenue and only lifts the dollar-denominated slice of cost (assumed 20%), for a profit gain of about 7.6tn, second only to price. Samsung disclosed a roughly 3.1tn FX benefit in its own 2Q. On top of that, the 1.1tn FX gain sits below the operating line, so currency shows up twice: in operating profit and again in net income.
What I actually watch
| Checkpoint | Why | When |
|---|---|---|
| 3Q DRAM bit growth vs the ~10% guide | First quarter where volume has to offset slowing price; look for shipment growth above price growth | Late October |
| 2027 HBM pricing outcome | Analysts cited by TrendForce expect +50% or more; a signed number flips the mix sign | Through 4Q26 |
| Quarterly D&A trajectory | How fast 4.0tn climbs sets the ceiling on 2027 margins; read with M15X timing | Each quarter |
| Any disclosure of LTA share | Management calls it 'an appropriate level'; a number would reset price sensitivity | Earnings calls |
Variable summary
| Variable | 2Q26 | Direction from 3Q26 | Profit per 10% move (illustrative) |
|---|---|---|---|
| 1. Price | DRAM +30%, NAND mid-50s | Decelerating (commodity 3Q +13-18%) | +7.9tn |
| 2. Bits | DRAM high single digits, NAND mid-teens | DRAM ~+10% guide, NAND low single digits | +7.0tn |
| 3. Mix | Blended ASP hit by slipped high-value shipments | HBM4 ramp, 2027 HBM pricing | ~0 (parity phase) |
| 4. Cost | D&A 4.0tn per quarter | New-fab depreciation from 2027 | +1.9tn |
| 5. Currency | 1.1tn FX gain below the line | Depends on the won | +7.6tn |
Risks to this view
- The sensitivities assume all revenue in dollars, 20% of cost in dollars and 50% of cost variable. The company does not disclose its cost or currency mix.
- Shipment and price changes were given as ranges; chart 3 uses midpoints.
- The price lever works both ways: a 10% price decline removes about 7.9tn. A 76% margin is also a measure of exposure.
- The 64.7tn consensus is FnGuide's; other aggregators had 63.7tn.
Closing
SK hynix profit is most sensitive to price, then currency, then bits, then mix, then cost. Today the first two are large and the last two are growing. How much volume and mix cushion the quarter when price rolls over, and how fast depreciation climbs, are the questions for 2027.
Sources: SK hynix 2Q26 earnings call transcript (The Elec, Jul 29, 2026); SK hynix Newsroom 4Q25, 1Q26 and 2Q26 results; Hankook Ilbo and Hankyung Magazine (Jul 29, 2026); TrendForce press releases (Jun 2 and Jul 3, 2026) and news (Aug 25, 2026); Samsung Electronics 2Q26 results.
Comments
Post a Comment