Skip to content 
Here’s a puzzle for you.
A company reports its best quarter in history. Operating profit jumps nineteen times over last year. Revenue more than doubles. The number beats what every analyst on the street was expecting.
And the stock falls 10% in a single day.
That’s exactly what happened to Samsung Electronics this week. And if you’ve been following the AI trade even loosely, this one’s worth understanding properly, because it tells you a lot about where we are in this cycle.
Let’s get into it.
The numbers, first
Samsung’s preliminary Q2 print showed revenue of about 171 trillion won, up 129% year over year. Operating profit came in around 89.4 trillion won (roughly $58 billion), a nineteen-fold jump from the same quarter last year.
To put that in perspective, this single quarter’s profit was higher than Samsung’s entire 2025 profit. One quarter beat one full year.
The operating margin worked out to roughly 52%, compared to just 6% in Q2 last year. That’s not a typo. Samsung went from a 6% margin business to a 52% margin business in twelve months.
If you’d shown these numbers to an analyst in isolation, without telling them the stock reaction, they’d have assumed the shares rallied hard.
Instead, Samsung closed down 8%. SK Hynix fell 7%. Japan’s Kioxia dropped 11%. And the Kospi, Korea’s benchmark index which is heavily weighted toward these two chipmakers, fell as much as 8% before triggering a trading halt, eventually closing down 5%.
So what’s going on?
Why memory chips became the hottest trade on earth
To understand the fall, you first need to understand the rise.
DRAM and NAND are the memory chips that sit alongside every processor, including the AI accelerators powering data centers worldwide. Three companies, Samsung, SK Hynix, and Micron, control almost all of global production. That’s about as concentrated as an industry gets.
When AI labs and cloud providers started racing to build data centers, they didn’t just need more processors. They needed staggering amounts of memory to go with them, especially high-bandwidth memory (HBM), the specialized chips that let AI accelerators move data fast enough to keep up with the workload.
Here’s the catch. Building more memory chip capacity isn’t something you do overnight. Fabs take years to plan and build. So when demand suddenly went vertical, supply couldn’t keep pace. Classic economics kicked in. Prices went up, and they went up hard.
Citi Research noted that average selling prices for DRAM and NAND rose 44% and 53% respectively in Q2 alone, quarter over quarter. Nomura expects a further 25% and 24% rise in the current quarter. This is what happens when demand overwhelms a genuinely constrained supply chain.
Naturally, the stocks followed. Samsung’s shares had more than doubled this year before the fall, pushing its market cap past $1 trillion. SK Hynix, Samsung, and Micron had all crossed the trillion-dollar mark back in May. The Kospi itself was up more than 80% for the year, and Samsung and SK Hynix together made up close to 58-60% of the entire index. When two stocks become that dominant in a national market, everything else becomes a side character.
So why did great news trigger a sell-off?
This is the part that trips people up, so let’s slow down.
Markets don’t react to numbers in a vacuum. They react to numbers relative to what was already priced in. And going into this print, expectations were sky-high.
Samsung’s guidance of 89.4 trillion won did beat the consensus estimate of about 87.3 trillion won. But that’s only a 6% beat. After a rally where the stock had doubled, investors weren’t looking for “good.” They wanted “spectacular.” A 6% beat, after a run like that, reads less like a win and more like an invitation to take profits.
There’s also a subtler point buried in the results. Samsung’s operating profit was reduced by provisions for employee bonuses, after the company agreed in May to pay its chip division staff 10.5% of operating profit as a special bonus. Some analysts pointed out that without this provision, profit could have crossed 100 trillion won. So even the beat came with an asterisk attached, which gave the skeptics something to hang on to.
Then there’s a structural worry that’s been building for a while: what happens when all this new capacity actually arrives?
The oversupply question nobody can answer yet
Right now, the shortage is real. But the profits pouring in are being aggressively reinvested into new capacity, and eventually, that capacity comes online.
Industry data from TrendForce shows fab capacity is set to increase meaningfully starting in 2027. Samsung, SK Hynix, and Micron are all racing to expand at the same time, together planning to invest a combined $2 trillion to build out chip production in South Korea over the coming years.
Demand forecasts still look strong through the first half of next year, but they start tailing off toward the year end, according to TrendForce and BofA estimates. Since markets are forward-looking, some of today’s selling is really investors repricing a 2027 oversupply risk, today.
There’s also a demand-side worry that’s less about chips and more about what’s happening one layer up. Data from Bloomberg shows LLM token spending, essentially how much AI companies are actually paying to run their models, has been pulling back lately. If enterprises get better at doing more with less, and start optimizing costs instead of throwing compute at every problem, the hardware trade eventually feels it too. JPMorgan flagged that investors are increasingly questioning whether AI memory can keep claiming an estimated 52% share of cloud providers’ capex this year, and over 70% next year. That’s a big number to sustain.
Adding to the unease, there were reports that Meta plans to sell excess computing capacity to external customers as part of a broader cloud push, a signal some read as early evidence that AI infrastructure spending might be outpacing actual usage.
The Korea-specific risk: leverage and concentration
Here’s where it gets particularly interesting for anyone watching Korean markets specifically.
Samsung and SK Hynix aren’t just large companies. They’re now so dominant that they effectively are the Korean stock market. When two stocks make up close to 60% of an index, that index stops behaving like a diversified basket and starts behaving like a leveraged bet on those two companies.
Layer onto that the fact that Korean retail investors have been trading with unusually high leverage. South Korea’s market has triggered circuit breakers multiple times in recent weeks, and the head of the country’s financial regulator has publicly warned about “excessive” leveraged positions among retail investors. There’s also been a proliferation of leveraged ETFs promising to amplify daily returns on these same stocks, which only adds fuel when things move quickly in either direction.
So when fundamentals even slightly wobble, the unwind isn’t gentle. It’s amplified by all the leverage sitting on top of the trade. By some measures, the Kospi has now fallen more than 20% from its June peak, technically putting Korea’s market into bear territory, even though it’s still up more than 70% for the year in local currency terms. That tells you how explosive the run-up was in the first place.
There’s one more wrinkle worth flagging. Analysts have pointed out that US competitors like Micron have moved toward longer-term purchasing agreements with customers, locking in demand visibility. It’s still unclear whether Samsung and SK Hynix have secured similar long-term contracts. Until Korean chipmakers offer more clarity on this front, some investors are choosing to stay cautious, even while the earnings numbers look extraordinary.
Is this the top, or just a breather?
Opinions are split, and reasonably so.
The bear case says the AI-memory trade has run too far, too fast, on speculative momentum rather than fundamentals, and a real correction is overdue as leverage unwinds. There’s even a chart doing the rounds comparing the semiconductor rally to the 2011 silver bubble, mapping the two price patterns on top of each other.
But the comparison has a hole in it. Silver’s rally was driven largely by a speculative narrative around a supply squeeze, with little underlying earnings power to justify it. Samsung’s rally, on the other hand, is backed by genuinely record-breaking, verifiable earnings. The capex, the demand, and the transformative nature of AI infrastructure are real, not narrative.
That’s the more constructive read: this looks less like a bubble popping and more like a very steep rally taking a breather after getting ahead of itself. As one analyst at CLSA put it, “I don’t think this story is over. It’s just a bit of a correction on the way up. Nothing goes up forever.”
What to watch next
If you’re tracking this space, a few things matter more than the daily stock move:
- 2027 supply data. This is the year fab capacity meaningfully increases. Watch whether demand forecasts hold up alongside it.
- Long-term contracts. Whether Samsung and SK Hynix follow Micron’s lead and lock in multi-year purchasing agreements could materially change how the market values them.
- AI capex trends at hyperscalers. If token spending and infrastructure investment keep growing, the memory story holds. If it slows, memory demand slows with it.
- SK Hynix’s US listing. The company is preparing a near-$30 billion ADR offering on Nasdaq, one of the largest-ever share issuances by an Asian company. How it’s received will say a lot about how global investors are currently pricing this trade.
The bigger picture here isn’t really about one bad day for Samsung. It’s a reminder that even genuinely great earnings can’t always outrun expectations that got too far ahead of themselves. And in a market as concentrated and leveraged as Korea’s chip sector has become, the swings, in both directions, tend to be bigger than the fundamentals alone would suggest.
Note: This piece discusses securities that do not trade on major US exchanges, including Samsung Electronics and SK Hynix, which are listed in South Korea. Investors should be aware of the additional risks, including currency and liquidity risk, associated with international shares.
