On July 7, Samsung Electronics reported preliminary Q2 operating profit of 89.4 trillion won ($58.4 billion), up more than 1,800% year-over-year and above the 87.3 trillion won analysts expected, on revenue of 171 trillion won. It was Samsung's third consecutive record quarter — and the single quarter beat both Nvidia's and Apple's, making it the highest quarterly operating profit any technology company has ever reported.
The number would have topped 100 trillion won if not for a one-time hit: a May wage agreement ties 10.5% of the semiconductor division's annual operating profit to employee bonuses, the result of a weeks-long labor protest that forced Samsung to scrap its old 1,000% base-salary bonus cap.
The market's response was to sell. Samsung's stock gave up as much as 10% intraday in Seoul and closed down 6.9%, wiping out more than $80 billion in market value in a single session. The selloff didn't stop there: SK Hynix fell 6%, the KOSPI index dropped nearly 5%, ASML fell 5.3% in Amsterdam, STMicroelectronics fell over 5%, and Infineon dropped more than 6%. Nasdaq 100 futures slipped about 1%. In premarket US trading, Micron, SanDisk, and Western Digital all sank roughly 7% on the same read-through.
This wasn't a miss — Samsung beat expectations. The real mechanism is that the beat was already priced in. Samsung's stock had run up roughly 150-165% year-to-date, and the market had spent months front-loading "record quarter" into the price. When the report merely confirmed what everyone already assumed rather than surprising to the upside, there was no incremental buyer left to reward it. eToro market analyst Zavier Wong put it bluntly: "It acts more like confirmation, and confirmation is what people sell into."
Two specific details added pressure. First, revenue came in slightly softer than hoped because DRAM price increases were more moderate than expected, putting a crack in the "structurally strong memory pricing" narrative. Second, Samsung simultaneously announced plans to spend roughly 400 trillion won building a new fab on a greenfield site in southwestern South Korea with none of the existing supply-chain infrastructure — read as another enormous capex commitment, feeding worry about whether AI-memory spending is starting to overbuild. JPMorgan's numbers are the sharpest version of that worry: AI memory already accounts for 52% of cloud-provider capex, and that share could exceed 70% by 2027. Any deceleration in AI infrastructure spending could reverse that highly levered profit structure fast.
History adds a sobering footnote: since early 2019, Samsung has beaten operating-profit expectations in 16 quarters, and in 10 of those 16, the stock fell after the release anyway. "Sell the beat" isn't a one-off for this stock — it's a recurring pattern.
Samsung and SK Hynix executives both warn "significant shortages" persist through at least 2027; some customers have already locked in supply through 2027, and Micron has sold out its entire 2026 capacity. Morgan Stanley raised its Micron price target from $520 to $1,050; Goldman Sachs revised its 2026 DRAM shortage forecast from 3.3% to 4.9%, calling it the worst in 15 years. Multi-year supply contracts are structurally dampening the industry's old boom-bust cyclicality.
All three major HBM producers are expanding capacity simultaneously, and when that capacity lands, oversupply risk arrives all at once. Micron has seen heavy recent insider selling — roughly $253.4 million in officer sales. Morgan Stanley also flagged hyperscalers shifting toward "more capex discipline," noting the historic run in semis since late March is losing momentum for the first time.
The second-order effects are already visible: capital is rotating out of AI-linked semiconductors and into names like insurers and banks, several of which hit record highs the same week — Bank of America, US Bancorp, BNY Mellon, plus Chubb, Travelers, and Cigna among insurers. Ordinary South Korean investors and pension funds, exposed through Samsung's and Hynix's outsized KOSPI weighting, felt the drop directly in account balances. Downstream, PC and phone makers that buy memory chips could eventually benefit if this round of capex anxiety actually forces suppliers to slow price growth over the next year or two.
The market's near-mechanical reaction to "sell the beat" obscures something real: Samsung's quarter is still the highest single-quarter operating profit any tech company has ever posted, and that fact wasn't disproven — only the idea that the news could push the stock higher was. Reading a positioning-driven selloff as proof that "the AI-memory story is breaking" overstates what one trading day actually tells you.
Worth being skeptical of, too, is JPMorgan's "52% to 70%" capex-share figure. It sounds like a demand-runaway signal, but a large part of that rising share is simply higher memory prices pushing up the dollar cost of the same purchases, not necessarily runaway growth in unit volumes. Treating a price-driven rise in spending share as proof that demand itself is unsustainable risks overstating the actual risk. And the fact that the whole semiconductor complex sold off together — ASML and Infineon, companies with no direct exposure to Korean DRAM pricing, dropped by similar magnitudes — looks more like ETF and basket-driven de-risking than analysts individually re-rating each company's fundamentals.
One takeaway worth keeping: whether an earnings report moves a stock has less to do with how good the numbers are and more to do with how much of that outcome was already bought into the price. Before chasing a stock into an earnings print after a 150%+ run, it's worth asking exactly how much of "the best possible outcome" is already sitting in that price.
This is informational analysis, not investment advice. Markets carry risk — form your own view.