SK Hynix earnings miss triggers global AI chip stock selloff

SK Hynix's Q2 2026 numbers were, by any normal standard, spectacular: revenue of 79.32 trillion won, up 257% year over year, and operating profit of 60.54 trillion won, up 557% ^[1]. But both landed short of the roughly 84 trillion won in revenue and 64 trillion won in operating profit that analysts, per LSEG SmartEstimates, had priced in ^[2]. That gap between spectacular and expected - not the results themselves - is what sent the stock down as much as 19% intraday before closing off roughly 9.6% ^[7]. BNK Investment & Securities' Lee Min-hee flagged the deeper worry underneath the miss: 'There are concerns that tech firms will take a breather in infrastructure spending.' ^[3]Hyundai Motor Securities' Greg Roh added that SK Hynix 'needs to come up with a concrete shareholder return policy to turn around investor sentiment' ^[3]- suggesting the miss also exposed a payout gap investors were no longer willing to overlook, even with net profit hitting 93.9 trillion won on Kioxia stake gains ^[3].

Two structural choices inside the earnings compounded the miss. First, HBM4 shipments came in slower than expected, pushing revenue recognition later and causing HBM price gains to lag conventional memory ^[3]. Second, SK Hynix locked in roughly ten long-term supply agreements, typically five-year terms, that improve demand visibility but may have capped how much of the current DRAM and HBM price spike the company could capture in the near term ^[3]. That tension - revenue certainty versus missed spot-price upside - became the central debate among retail investors parsing the print: some read the long-term deals as a prudent hedge against cyclicality, others as evidence SK Hynix left money on the table during the hottest stretch of the AI memory cycle. DS Investment & Securities' Lee Su-rim added a related wrinkle, noting Samsung 'has greater pricing power and has raised prices more aggressively than SK Hynix' this cycle ^[3], meaning the two Korean rivals aren't equally exposed to the same criticism despite falling together.

Around the same time SK Hynix reported its results, China's CXMT (Changxin Memory Technologies) debuted on Shanghai's STAR Market, raising $8.6 billion and closing about 466% above its offer price - the largest semiconductor IPO in Asian history ^[6]. An equity analyst, Kim Seok-hwan, argued the market's real fear isn't CXMT's current output but its trajectory: 'The market's concern lies less in CXMT's current earnings and more in its potential for accelerated capacity expansion to rival Korean companies and technology development following its IPO' ^[4]. That framing turned an already high-stakes earnings season into a referendum on how much longer Korean firms can hold their memory-chip lead.

Because SK Hynix and Samsung together account for over half of the KOSPI's weighting, the earnings miss didn't stay a single-stock story ^[4]. The index fell 10.84% on Tuesday and another 5.99% on Wednesday - a combined two-day drop of roughly 12.6-18%, its worst two-day stretch on record - and triggered circuit breakers on consecutive sessions for the first time in the exchange's history ^[4][8]. Much of the prior AI-memory rally had been fueled by leveraged retail positioning, and that same leverage which amplified the run-up accelerated the crash as brokers force-liquidated losing positions ^[4]. Globally, Nvidia, SK Hynix, Samsung and Micron shed a combined roughly $700 billion in market value ^[5], with contagion reaching Kioxia, TSMC and ASML over the following week ^[10], and SK Hynix on pace for its worst monthly performance since October 2008 ^[9].

Not every voice in the aftermath read the selloff as a warning sign about the underlying business. Sarah Kunst of Cleo Capital, speaking on Bloomberg Television, framed the drop as ordinary profit-taking after an extraordinary run - SK Hynix and Samsung were both up roughly 500% over the prior year - rather than evidence of a fundamental crack in the AI memory story - a segment that also referenced SK Hynix's own leadership suggesting memory-chip shortages could persist well beyond 2030, meaning the demand backdrop hasn't actually changed. Joe Mazzola of Charles Schwab, on Schwab Network, offered a different reframe entirely: he tied the selloff less to SK Hynix's specific numbers and more to a broader mega-cap tech rotation already underway, with weakness bleeding in from Google, Amazon, Microsoft and Meta - a market-wide repricing of richly valued tech rather than a verdict on Korean memory makers specifically. Taken together, these framings suggest the bearish read on SK Hynix's miss - that it exposes real cracks in AI infrastructure demand - is contested even among those watching the same numbers: what looks like the market losing faith in the memory cycle can just as easily be read as an overheated trade cooling off inside a still-intact supercycle.