SK Hynix reported its Q2 2026 results before the Korean market opened Wednesday. Revenue came in at 79.32 trillion won, 5.5% below the 84 trillion won consensus. Operating profit was 60.54 trillion won, 5.4% below the 64 trillion won estimate. SKHY dropped 8% in premarket. Then it bounced. By the time Korean regular trading was underway, the stock had turned green. That reversal, from minus 8% to positive on a headline miss, is the most analytically important data point of the morning, and it is the inverse of the pattern that destroyed Micron's stock after its blowout June earnings.
The context matters enormously. KOSPI had already fallen 10.84% on Tuesday to 6,023.66. SKHY had closed at $130.17 in New York, down 8.98% on the day, before the earnings were even released. SK Hynix shares had declined 15.58% over the prior five sessions. The entire Korean market had pre-sold the earnings miss before it arrived. When the actual numbers confirmed the miss at 5%, not at 15% or 20%, the market had nothing left to sell and everything to cover.
What the Numbers Actually Show
Strip away the consensus miss and read the absolute numbers. Revenue of 79.32 trillion won is up 257% year over year and up 51% sequentially from Q1's 52.58 trillion won. Operating profit of 60.54 trillion won is up 557% year over year and up 61% sequentially. Net profit was 93.92 trillion won with a net margin of 118%, marking an all-time high quarterly performance. The company's cumulative first-half revenue crossed 100 trillion won for the first time in company history.
Management attributed the results to exactly the dynamic the bulls have been describing for months. High-performance products for AI servers led price increases. DRAM average selling price rose approximately 30% quarter over quarter. NAND ASP rose mid-50% quarter over quarter. These are not soft numbers dressed up with accounting adjustments. They are the pricing trajectory of a market in structural shortage. Big-tech customers are expanding infrastructure investment and memory procurement on surging AI service growth and a shortage of computing capacity, which management cited explicitly in the earnings release.
The company now has long-term supply agreements with 10 customers, up from what Micron disclosed in its June earnings. Those agreements, combined with the DRAM and NAND ASP increases, suggest the pricing power that has driven margins to 76% operating margin is contractually protected to a meaningful degree rather than purely spot-market dependent.
The Miss That Wasn't Really a Miss
The 5% revenue shortfall against consensus deserves scrutiny before being taken at face value as a negative signal. The consensus of 84 trillion won was constructed from analyst estimates built on the assumption that smartphone and PC memory volumes would contribute normally to the quarter. SK Hynix disclosed in the earnings release that smartphone and PC sales were temporarily adjusted because customers had difficulty securing memory volumes, meaning the miss came from the supply-constrained segments, not from weakness in AI server HBM.
The HBM business, which carries the highest margins and represents the strategic core of the bull case, was not the source of the miss. The source was the conventional DRAM and NAND business that hyperscaler customers were competing for so aggressively that it disrupted supply to other channels. A company that misses revenue because its highest-margin product customers were buying so much that conventional customers couldn't get enough is a very different miss than one that misses because demand fell short.
The tweet from TJ Research summarized the dynamic concisely: this is the inverse of Micron's sell-the-news reaction to a blowout June earnings. Micron beat massively and the stock sold off. SK Hynix missed modestly and the stock bounced. When bad news has been so thoroughly priced in that even a genuine miss produces a reversal, the selling phase has exhausted itself.
The KOSPI Reversal as Market Structure Signal
Tuesday's KOSPI session, which closed at 6,023 after falling 10.84%, was one of the most severe single-day declines in the index's history. Samsung Electronics fell roughly 13.4% on the same day. The combined market cap erasure across Korean equities represented a pricing-in event of extraordinary scale.
What makes Wednesday's initial bounce significant is that it is occurring into that backdrop. Korean margin accounts had already been partially deleveraged, with Goldman Sachs data showing margin balances falling from 38 trillion to approximately 33 trillion won over the prior weeks. The leveraged ETF mechanical selling that amplified every prior down session was operating on a smaller base. When the actual earnings miss came in at a level the market had over-discounted, the natural reaction is short covering and relief buying rather than continuation selling.
The bounce does not mean the bottom is definitively in. SK Hynix's earnings call commentary on HBM pricing sustainability, 2027 demand visibility, and the impact of CXMT's IPO expansion on conventional DRAM will determine whether today's move holds. But a stock that opens down 8% on a miss and closes green has told you something specific about market structure: the sellers who needed to exit have largely exited.
The Short Side Was Building Before the Numbers Arrived
One piece of context that reframes the entire setup is UBS's prime brokerage data from the past two trading days. According to UBS's TMT team, the semiconductor selloff heading into this week was not purely a continuation of the momentum unwind that has characterized July. Something structurally different was happening.
UBS's prime brokerage data showed that over Monday and Tuesday, hedge funds began establishing tactical short positions in US semiconductor stocks ahead of the hyperscaler earnings reports due Wednesday and Thursday. This is a meaningful distinction. The earlier momentum-driven selloff from late June through mid-July was almost entirely driven by longs getting cut, meaning investors who owned the stocks reducing or eliminating positions. That is a deleveraging dynamic. What happened Monday was a new active shorting dynamic.
On Monday, US semiconductor stocks saw net selling of negative 2.8 standard deviations, with more than 75% of that flow driven by short sales rather than long liquidation. A 2.8 standard deviation net selling event is roughly a 1-in-200 trading day occurrence. The fact that it was driven predominantly by new shorts rather than existing longs exiting tells you the positioning was tactical and event-driven: funds were making a specific bet that hyperscaler capex commentary or AI demand signals would disappoint, then planning to cover once that information was in the market.
SK Hynix's earnings release this morning, combined with the bounce from the opening lows, changes the calculus for those tactical shorts. If the thesis was that hyperscaler earnings would reveal a capex slowdown or demand plateau, SK Hynix's DRAM ASP up 30% quarter over quarter and NAND ASP up mid-50% are not consistent with that thesis. Shorts built before earnings on a capex disappointment thesis face a squeeze if Microsoft and Meta, reporting this week, confirm continued AI infrastructure acceleration. The magnitude of Monday's short positioning, at 2.8 standard deviations and 75% short-sale driven, means the covering of those positions would be a significant source of buying pressure in the same names that saw the most aggressive shorting.
What This Means for MU, DRAM ETF, and the Broader Complex
Micron is currently around $810, down roughly 35% from its all-time high of $1,255. The DRAM ETF is near $47. Both have been pricing a scenario where the AI memory earnings cycle was either peaking or about to be disrupted by Chinese supply. SK Hynix's Q2 result provides the most direct available evidence on both questions.
On the peak question: DRAM ASP up 30% quarter over quarter and NAND ASP up mid-50% quarter over quarter are not the numbers of a cycle that is rolling over. They are the numbers of a market where supply is still running behind demand by enough to sustain aggressive pricing. The 76% operating margin, delivered on a quarter that missed on the top line by 5%, means the profitability floor of the business is structurally higher than at any prior point in memory industry history.
On the China question: conventional DRAM disruption from CXMT is a 2028 to 2029 story. The miss in SK Hynix's Q2 revenue came from supply tightness causing customers to get less product than they wanted, not from Chinese supply pushing them toward alternatives. CXMT currently produces roughly 8% of global DRAM at nodes that cannot manufacture HBM. Its IPO proceeds will fund conventional DRAM expansion that takes 18 to 36 months to reach production. The 76% operating margin SK Hynix delivered today is built almost entirely on HBM, which CXMT cannot touch.
Morgan Stanley's chief equity strategist Mike Wilson added macro context on Tuesday that reinforces the setup. While acknowledging that geopolitical tensions, oil price volatility, and Fed policy uncertainty could keep markets choppy for another month, Wilson said he remains constructive on the S&P 500 for the full year. He identified 7,000 as a strong support level and reiterated his year-end target of 8,000, calling any move toward that support a buying opportunity. An S&P 500 with a defined floor at 7,000 and a credible year-end target at 8,000 implies roughly 14% upside from the support level. For memory stocks that have already corrected 30 to 35% from their highs while their earnings trajectories remain intact, that macro framework is a meaningful tailwind.
The next catalysts that will determine whether the bounce is the beginning of a recovery or a short-covering relief are Microsoft and Meta reporting this week, and the FOMC decision on Wednesday afternoon US time. If Microsoft confirms AI infrastructure spending acceleration and the Fed delivers a hold that signals September is not certain, the combination creates the conditions for a sustained recovery in the memory complex from deeply oversold levels.
Explore more exclusive insights at nextfin.ai.
