SKHY — SK hynix
HBM leadership is printing absurd cash. The question is whether SK hynix can spend the next cycle without recreating the old one.
SK hynix is sitting right in the middle of the memory shortage with leading HBM share plus DRAM and NAND exposure. HBM is crowding conventional DRAM wafer capacity, enterprise SSD demand is tightening high-end NAND, and the business is printing absurd cash. Then there is the ADR, which can trade at a giant premium to the Korean ordinary. So I have two valuation problems here, not one.
SK hynix is basically sitting in the middle of the memory shortage holding a flamethrower.
Why I care about this one
TrendForce now expects 2027 HBM blended ASP to rise 121% year over year, driven by persistent supply tightness, a richer HBM4 mix and HBM4E ramping in the second half of 2027. Again, that is a blended ASP forecast, not “the exact same chip doubles in price.” (TrendForce — Sep. 29)
The more important detail might be the wafer crowd-out.
Samsung estimates HBM will consume nearly 30% of industry DRAM wafer capacity in 2027, versus about 20% now. HBM and conventional DRAM use the same wafer pool, so more HBM can keep standard DRAM tight rather than simply adding to total memory supply. (Reuters)
That is good for SK hynix in two directions.
It is the HBM leader, and the HBM build can tighten the commodity product it also sells.
Then there is NAND.
The consumer market is much less exciting, but high-end enterprise SSD is a different story. AI workloads and KV-cache demand have pushed cloud customers toward much larger SSD orders, and TrendForce is describing the enterprise segment as structurally short. For 4Q26 it expects enterprise SSD contract prices up 23%–28% q/q, versus 15%–20% for NAND overall. (4Q26 memory forecast)
That matters directly to SK hynix through Solidigm.
So the current setup is not one bottleneck.
It is HBM tightness, conventional DRAM crowd-out and enterprise SSD demand all pulling on the same broad memory supply chain.
SK hynix has approved roughly ₩54T for Yongin Y2 and Cheongju M17. But M17’s first cleanroom is targeted for December 2028, and Y2’s for June 2029. Equipment then gets installed progressively according to demand. (fab investment plan)
Construction spend is not the same thing as sellable bits.
That timing is one reason the shortage can remain stronger for longer than the headline capex numbers suggest.
It is also why I am moving the valuation checkpoint to FY2029.
If I want to test whether HBM really changed the cycle, I need to let the new supply actually arrive first.
Q2 2026 + Sep. 2026 shortage / ADR update
Primary sources: Q2 2026 business results ↗ · Sep. 29 HBM shortage / 2027 pricing — TrendForce ↗ · Sep. 29 HBM wafer-capacity estimate — Reuters ↗ · 2026 fab facility investment ↗ · SK hynix ADS prospectus — SEC ↗
HBM demand, conventional DRAM crowd-out and enterprise SSD tightness support earnings into 2027–2028, while M17 and Yongin Y2 do not provide substantial cleanroom capacity until late 2028–2029. My FY2029 Base uses about $30 ADR EPS at 9.5x with no terminal ADR premium, or $285. The difference between Korean underlying value and the U.S. wrapper remains a major valuation risk.
My position
What I actually bought and sold.
| Date | Type | Account | Dollars | Shares | Price | Return now* |
|---|---|---|---|---|---|---|
| Jul 13, 2026 | Buy | Taxable | $10,226 | 65.359 | $156 | +16.7% |
| Jul 20, 2026 | Buy | Taxable | $5,614 | 34.884 | $161 | +13.4% |
| Sep 9, 2026 | Fee | Taxable | $2.00 | — | — | — |
*Return now uses the latest available quote for buys and reinvestments. Sales and fees show “—”.
Valuation & scenario math
An ADR is not the same security as the Korean common, so I start with the operating business, translate the earnings into ADR economics, apply a fundamental multiple and then ask whether any terminal wrapper premium is justified.
| FY2029 | Bear | Base | Bull |
|---|---|---|---|
| Revenue | ₩410T | ₩560T | ₩680T |
| Operating margin | 48% | 65% | 72% |
| Net conversion after tax / other | ~78% | ~78% | ~78% |
| Shares | ~706M | ~700M | ~690M |
| USD/KRW | 1,450 | 1,350 | 1,250 |
| ADR EPS | ~$15 | ~$30 | ~$44 |
| Fundamental P/E | 6x | 9.5x | 10.5x |
| Terminal ADR premium | 0% | 0% | 15% |
| SKHY value | ~$90 | ~$285 | ~$531 |
The Base operating math is:
₩560T revenue × 65% operating margin = ₩364T operating profit
Then I assume about 78% converts to net income, divide by roughly 700M shares, translate one-tenth of that per-share earnings into dollars at 1,350 won per dollar, and get about $30 of ADR EPS.
The biggest valuation question is the wrapper. I use 9.5x on the underlying Base earnings and assume no terminal ADR premium, for roughly $285.
From the Oct. 6 close, that is still around 16% annualized. At 8.5x with no premium, it is still around 12%.
A persistent ADR premium belongs in Bull. Base already asks enough of the memory operating cycle without needing investors to keep paying extra for the U.S. wrapper.
HBM leadership + normalized-cycle value + total memory exposure
No static dollar band. Another dollar has to clear the underlying memory economics, the ADR premium versus the Korean ordinary, and the fact MU + SKHY are basically one memory bet in two tickers. A lower ADR can still be expensive if the wrapper premium stays stupid.
I model the underlying business first because SKHY is an ADR wrapper, not the Korean ordinary share. That means KRW operating economics → ADR EPS → a fundamental multiple → an explicit assumption for how much ADR premium survives. Otherwise I am mixing a great business with a potentially expensive wrapper and pretending they are the same thing.
| Case | What has to be true | Target range | Annualized return from today |
|---|---|---|---|
| Bear | FY2029 revenue ~₩410T, 48% operating margin, ~78% net conversion, ~706M shares, USD/KRW 1,450, ~US$15 ADR EPS, 6x fundamental P/E and no terminal ADR premium. | $75–$110 center ~$90 | -21.3% |
| Base | FY2029 revenue ~₩560T, 65% operating margin, ~78% net conversion, ~700M shares, USD/KRW 1,350, ~US$30 ADR EPS, 9.5x fundamental P/E and no terminal ADR premium. The business has to justify the value without assuming the U.S. wrapper stays irrationally expensive. | $255–$315 center ~$285 | +16.3% |
| Bull | FY2029 revenue ~₩680T, 72% operating margin, ~78% net conversion, ~690M shares, USD/KRW 1,250, ~US$44 ADR EPS, 10.5x fundamental P/E and a 15% terminal ADR premium. | $470–$590 center ~$531 | +43.5% |
Returns use the current quote and the time remaining to Sep 25, 2029. The scenario targets stay fixed until the underwriting changes.
Capital allocation
HBM leadership is earning the capex. Please do not celebrate by rebuilding the old memory cycle.
The balance sheet is extremely strong, HBM investment is justified by current demand, and the ₩40T buyback/cancellation plus >50% FCF return framework is unusually shareholder-friendly for memory. I still take points off for the sheer size of the ~₩54T fab commitment and the industry's incredible historical ability to turn scarcity into oversupply.
Stage-relative, not some objective rating-agency grade. The number is there to force the judgment into the open. See the rubric →
What management is actually doing with the cash
SK hynix is spending aggressively because the shortage economics are extraordinary, and it is also returning serious cash: a ₩40T buyback with full cancellation and a target to return more than 50% of cumulative 2025–2027 FCF. I like that. The fab timing is the part I keep watching. M17's first cleanroom is not due until Dec. 2028 and Y2's until June 2029, so ₩54T of announced investment does not equal immediate supply. Great for the current shortage. It also means FY2029 is where I want to judge whether the spending actually earned its keep.
What I’m doing with the next dollar
This is the part I care about.
The HBM shortage is real.
The wafer crowd-out is real.
The enterprise SSD shortage is real.
The timing of new fabs means a giant capex headline does not create giant supply tomorrow.
All of that makes me more confident in 2027–2028 earnings.
But the ADR premium means U.S. buyers are already paying extra for access to those economics.
The FY2029 endpoints are roughly $90 / $285 / $531. The live page calculates the return from today’s ADR quote.
The base return can still be good.
It is just nowhere near as ridiculous as the underlying Korean business looks if I ignore the wrapper.
What can break this
This is the other big change.
One SKHY ADS represents one-tenth of one SK hynix common share. That is not an estimate; it is the actual depositary ratio in the SEC prospectus. (SEC prospectus)
On Sep. 29, the Korean ordinary closed around ₩1,765,000. USD/KRW was about 1,359. That makes the local-share economic value per ADR roughly:
₩1,765,000 ÷ 10 ÷ 1,359 ≈ $130
SKHY itself traded around $187.
So the U.S. ADR was carrying roughly a 44% premium to the Korean-share equivalent.
That is absolutely not a rounding error.
MarketWatch was already flagging a 38% premium in July and tied it to limited arbitrage and conversion mechanics around the new listing. (MarketWatch)
The underlying company can look cheap while the U.S. wrapper is expensive.
If I ignore that, I am not really valuing what I actually own.
SK hynix approved a ₩40T share repurchase with full cancellation.
At the price used by the company when it announced the plan, that represented roughly 3.3% of issued shares, and management raised the 2025–2027 shareholder-return target to more than 50% of cumulative FCF. (buyback and cancellation)
That is exactly what I want to see when a cyclical business is printing ridiculous cash. There is plenty worth funding in the capacity roadmap, but I also want the balance sheet to stay strong and some of that cash to make its way back into the denominator.
The only part I do not want to underwrite is today’s shortage lasting forever.
The evidence I’m watching
HBM4/HBM4E leadership holds, the shortage lasts through 2027 without major demand destruction, enterprise SSD strength broadens through Solidigm, LTAs protect economics, M17/Y2 ramp into contracted demand and the ADR premium compresses without the Korean equity falling with it.
Competitors close the HBM technology gap, customers reduce memory content enough to wreck pricing, enterprise SSD tightness fades as NAND loosens, M17/Y2 add bits into a slowing market, or the ADR premium stays elevated and then collapses for reasons completely separate from the business.