SK hynix Inc. (KRX:000660) closed Friday at ₩1,647,000. Over the past twelve months the shares have traded as low as ₩265,500 and as high as ₩2,919,000. The stock is up more than six times from the low and down 44% from the high, and both of those happened inside one year.
The company's market capitalisation was ₩1,203.1 trillion at Friday's close. Against trailing four-quarter net income of ₩162.1tn, that is 7.4 times earnings. Against book equity of ₩262.7tn at June 30, it is 4.6 times book. My view is that neither multiple means what it usually means, because the denominators are moving faster than the price.
The eleven years of annual net income on file from DART start at ₩4.3tn, ₩3.0tn and ₩10.6tn. They then run ₩15.5tn, ₩2.0tn and ₩4.8tn. Next come ₩9.6tn, ₩2.2tn and negative ₩9.1tn. The final figures are ₩19.8tn and ₩42.9tn. This sequence runs from FY2015 to FY2025. The old cycle had a rhythm. Two or three years up, then a crash. The peak year, FY2018, earned ₩15.5tn. The trough, FY2023, lost ₩9.1tn.
Then came the last four quarters. They ran ₩12.6tn and ₩15.2tn. Then came ₩40.3tn and ₩93.9tn. The June quarter alone earned more than twice what FY2025, itself a record, earned in twelve months. Operating income in that quarter was ₩60.5tn on revenue of ₩79.3tn, a 76.3% operating margin. Gross margin was 83.2%. For reference, the FY2018 peak ran a 51.5% operating margin, and that was considered the top of what memory could do.
On FY2025 earnings the shares trade at 28 times. On trailing earnings, 7.4 times. On the June quarter annualised, about 3.2 times. Pick your denominator and you can make this stock look like anything.
The June quarter needs a footnote. Pretax income was ₩122.7tn against operating income of ₩60.5tn. The difference is ₩65.9tn of finance income. An earlier piece here flagged it and said to anchor on the operating line until it's explained. The filings I've read since don't explain it.
So here is the operating-only version. Take trailing operating income of ₩128.7tn and apply the June quarter's 23.5% effective tax rate. That gives roughly ₩98tn of after-tax operating earnings. On that basis the shares trade at about 12 times. Still cheap for a company growing at this rate. Not 7.4 times.
Whatever produced the ₩65.9tn, it also produced a chunk of the book value that the P/B ratio rests on. Which brings me to the number I think matters most.
Total equity was ₩120.7tn at the end of FY2025. It was ₩262.7tn at June 30. That is a 118% increase in two quarters. Retained earnings went from ₩106.6tn to ₩242.3tn over the same span.
Nothing in the history comes close. Equity took from FY2015 to FY2021 to go from ₩21.4tn to ₩62.2tn, roughly tripling in six years. FY2024 to FY2025 was a 63% rise, itself unprecedented. Then this.
A 4.6 times book multiple on a book that just doubled is a different thing from 4.6 times a stable book. If the second half adds even ₩80tn of retained earnings, that would require a sharp slowdown from the June quarter. In that case, book equity ends 2026 near ₩340tn. The multiple on Friday's price then drops to about 3.5. If the June quarter's pace holds, book equity could pass ₩400tn and the multiple falls below 3. The price hasn't moved to reflect a denominator that is compounding at this speed, or the market doesn't believe the denominator will hold.
That second reading deserves respect. In FY2023 the company lost ₩9.1tn and equity fell from ₩63.3tn to ₩53.5tn. Book value in memory goes down as well as up. What is different now is the scale. A loss year like FY2023 would take 3.5% off the current book, not 15%.
The board resolved on August 19 to buy back 24,070,000 shares for ₩40,004.3bn. The plan calls for exchange purchases between August 20 and November 19. The filing says the company will cancel every share it buys. The price used for the calculation was ₩1,662,000. At that price, the buyback is 3.3% of the 730.5 million shares outstanding. The daily order limit is 2,407,000 shares. The company had already cancelled 15.3 million treasury shares earlier in 2026.
The filing also shows the room. Distributable profit at the end of FY2025 was ₩91.6tn, and after dividends already declared, the remaining limit for buybacks was ₩89.4tn. The ₩40tn programme uses under half of that. Management has said the return target is more than 50% of cumulative free cash flow from 2025 to 2027. Korean press reports a dividend payout goal of 50% by 2030.
Measured against cash, the buyback is large but not stretched. First-half operating cash flow was ₩91.7tn and purchases of plant were ₩18.3tn. That leaves ₩73.4tn of free cash in six months. The ₩40tn buyback is 54% of it. Cash and equivalents were ₩26.8tn at June 30.
One more filing worth noting. On September 4 the company responded to an August report that its US subsidiary Solidigm was pursuing a ₩5tn pre-IPO round. It said Solidigm is reviewing options and nothing is decided, with a re-disclosure due by December 3. That is a possible source of finance income or a valuation mark in the second half, and I'd read it alongside the ₩65.9tn question.
Ten-year valuation data for Korean stocks is uneven across sources, so I'll stick to what the DART record supports. Return on equity in FY2025 was 44%. The June quarter annualised is 143%. The FY2018 peak, on the old numbers, was closer to 35%. The company has never earned anything like this on its capital, and the market has never had to price it.
The 52-week range is the market's own confession. From ₩265,500 to ₩2,919,000 is an eleven-fold move. That is not a range for a company whose value is understood. It is a range for a company whose earnings the market is still trying to believe.
The June quarter could be the peak. Memory has always peaked. If HBM pricing turns in 2027 as new capacity arrives from Samsung and from China, the trailing multiple will expand as earnings fall. Then 7.4 times becomes 15 times without the price moving.
The finance income could reverse. If ₩65.9tn came from marking an investment up, it can be marked down. The book value doubling would partly unwind.
Capital spending could climb. Purchases of plant were ₩27.5tn in FY2025 and ₩18.3tn in the first half of 2026. A fab race would eat the free cash the buyback rests on.
And the buyback is a three-month programme. It ends November 19. Whether another follows is a board decision, not a commitment.
The September-quarter results in late October answer the question this piece can't. If finance income is small and net income tracks operating income, the ₩65.9tn was a one-off. In that case, the trailing multiple is really about 12 times operating earnings. If finance income is large again, there is a recurring source the filings haven't named, and I'd want to know what it is before trusting any multiple.
The second number is retained earnings at September 30. Above ₩290tn means the book is still compounding at the June pace. The buyback completion report in late November will show the average price paid, which tells you what management thought the shares were worth.
kstock reads DART every morning and writes up what moved — the contract, the buyback, the number that does not add up. The daily post and a Saturday roundup, by email.