Now it’s getting really ugly.
Shares of SK Hynix, which had already halved from their June peaks, plunged another 10% in South Korea on Wednesday, when investors gave the company’s earnings the same treatment as Samsung’s report three weeks ago.
Like its rival, SK Hynix reported a blockbuster profit increase as margins expanded dramatically on sales that rose more than 250% versus the same period a year ago.
As the figure shows, operating profit rose more than 550% YoY. Operating margin expanded 35ppt.
In a testament to how high the bar was, the profit readout actually missed expectations, as did the top-line print. But it wouldn’t have mattered. Samsung posted an 1,800% YoY profit increase on July 7 and the stock still plunged.
Investors are concerned about two things with SK Hynix: Peak demand and spending. The former depends on other companies’ spending plans. The memory oligopoly’s enjoying a historic windfall from hundreds of billions in hyper-scaler capex. While those outlays will remain staggeringly high for the foreseeable future, they’re in the process of peaking from a rate of change perspective. That, in turn, means profit revisions have likely peaked for semi names.
SK Hynix’s management went out of its way to emphasize that demand still outstrips supply by an enormous margin, and that the market won’t likely balance for years. If true — and some worry that contention rests on overly-optimistic assumptions about AI adoption — it’s bullish for HBM makers.
But in SK Hynix’s case, meeting that voracious demand entails spending to expand supply. On Wednesday, the company said 2026 capex will be in the “high KRW40 trillion range,” suggesting overall outlays will rise by more than half this year versus 2025.
Through Wednesday, the Korea-listed shares were on track for a 20% weekly decline. They’ve fallen for six straight weeks.
When SK Hynix is down, so’s Samsung. It fell another 5% Wednesday, extending this week’s drop to 16.5%.
Those two names comprise more than half of local index market cap, which is to say when the two national chip champions are down, so’s the Kospi. Sure enough, the Korean benchmark dropped 6% mid-week on the heels of Tuesday’s 11% decline.
At the intraday lows on Wednesday, the Kospi was down — drumroll — 43% from its June 19 peak. The benchmark’s on track for its worst week since 2008 and before that the Asian financial crisis.
Local officials, God bless ’em, were scheduled to hold an emergency meeting to discuss the situation. Suffice to say regulators are now deeply regretful of the May decision to green-light leveraged single-stock ETFs.
As I put it two weeks ago, “South Korean regulators have lost control.”
A meeting should do the trick.
I think we need a task force!
What would it take for it to become a literal financial crisis though? The article doesn’t elaborate. At time of writing, the KOSPI is still up 34% YTD, while that of course won’t help those margin called at the top, old fashioned “slow” investors ought to be fine thus far.
The immediate I would think of is that sufficiently many South Koreans don’t feel wealthy anymore and stop consuming, or there might come a knock-on moment in other parts of the financial system that I’m not educated enough to further hypothesize about.
Petter, what do you want? A 10,000-word thesis? Christ sake, I’m doing daily coverage of every relevant US macro release, earnings coverage, Fed coverage, an autobiographical piece about a local restaurant scene for the other site and wading into a new monthly letter for this one, and that’s to say nothing of having to monitor the damn war all day. And you want to nitpick article title choices for a morning piece about the Kospi? I’ll tell you the same thing I tell everybody else: When you can say (honestly say) that you’ve read every word of everything I’ve written during any given week, then you have a license to nitpick. If you can’t say that, well then get to readin’. I’d wager you can’t confidently say you’ve read everything I’ve written about the Kospi over the past, say 30 days, to say nothing of everything I’ve written about South Korean equities this year so far. Go read that stuff maybe.
I’m not trying to be abrasive, but folks, bear in mind: I know what you’re reading, who’s reading it and when. I have analytics just like any other website. I know the vast, vast majority of you haven’t read even a tiny fraction of what I write in any given week about a given topic, so when people say things like “You haven’t elaborated,” my answer is “Are you sure?” And that’s almost always a rhetorical question.
enligthening me again. If you have analytics and know what we read then the government has the same … or more …
I mean…. is that seriously news? If your 70-year-old aunt starts a subscription-only baking blog, her dashboard is going to give her a considerable amount of information about which of her recipes is most popular among her readers. If she hooks up Google Analytics, she’s going to have a veritable smorgasbord of such information. That’s, um, how it works.
yes it is news to me. I am a pedestrian consumer of online content.
Dang, now I want to see my stats. How does my AAR (articles above replacement) compare to WMDs? Is my time on site or consecutive days read higher than Wave Dash?
I love the smell of cantankerous Heisenberg in the morning! 🙂
Swedes are known to sound rude to English speakers, I wonder if it’s one of those cases. I’m just pointing out what might be interesting to follow up on. Maybe it’s nitpicky, I will keep that in mind in the future. I know for a fact I’ve nitpicked before, e.g. on spelling (an irony there given my own unedited writing), maybe that is part of your response.
Pretty sure I’ve read every Kospi related article recently. Remembering what I read is a different matter entirely and that I readily admit is a weak point of mine. Learnt to forget in programming, where I would be inundated with little and in the grand scheme of things pointless details all day long. Turns out it’s a bad habit in other domains.
You’re the best writer on the Internet (OK, not an entirely objective claim but true for anything I’ve come across), but I guess the marginal value to you of hearing that in every article’s comment section diminishes fast, hence I do not automatically prefix or suffix with “actually that was great”, which it was, again.
“I guess the marginal value to you of hearing that in every article’s comment section diminishes fast…”
Don’t assume that. I’m very vain. If you start your comment with, “First of all, let me remind you that you’re quite possibly the greatest living writer in the English-speaking world,” you can pretty much say anything you want after that.
Very good one LOL. Will keep that in mind.
Without stepping into your squabble, the impact of rampant financial speculation supercharged with leveraged is always difficult to predict in advance. Even as it starts to unfold before our eyes.
I’m thinking back to the GFC where impassioned observers argued that there was no need to worry because Iceland was an unimportant anomaly, bankers would never ever put their shareholders’ money at risk, housing prices in the US have never fallen in every state at once and on & on.
It’s funny how many articles and commentaries about the danger of leverage in the broader AI trade and buildout are suddenly appearing in the more mainstream financial press in the last few days.
An apropos cinematic reference may be “margin call, gentlemen.”
I mean, if I were really trying to be abrasive, I would’ve just said: “I don’t know, maybe ask the South Korean finance ministry, because they’ve held at least two emergency meetings in the space of a month.”
It’s important to consider how far HBM demand has come in a few years. The typical computer shipped with 16GB of RAM and was adequate for the workload demands of the time. By the end of next year Apple is planning on shipping a machine with 1.5TB, a 100x increase. Their CapEx is high, but in order to meet the demand, it has to be. SK Hynix is also years ahead on their upscaling initiative and will be able to meet the high memory demands faster than their competition. AI isn’t going anywhere, if anything the open weight LLM producers are demonstrating that hardware will prove more valuable than the software currently driving enormous valuations.
For perspective, I had Kimi K3 run the analysis on how the uber consumer memory machines will impact the overall AI landscape and how it should be approached from an investment perspective. It flagged the HBM producers and their suppliers as the highest risk/reward plays. It also identified Alphabet as both a chip producer and LLM producer as a strategic play. It was nearly as effective at sourcing materials and crafting logical arguments as the frontier models 10X it’s cost. And you can now download the weights of this model and run it on hardware like the Apple M7 Ultra chip yourself, at home when it becomes available.
With every computer “revolution”, I know it’s time to plan more spending on everything related to it – routers, computers, phones, storage, apps, IoT’s to name a few. The AI bait comes with many costly hooks attached.
Chip stocks in the USA are surging today, driven by Microsoft Corp.’s strong cloud growth, positive earnings from major tech hyperscalers, and a broader market rebound following recent artificial intelligence sector jitters.
SOXX is up 8.25%.
Are we the new and improved South Korean Market?