SK海力士史上最赚钱的季度，为何仍“不及预期”？

Original | Odaily Planet Daily (@OdailyChina)

Author | Azuma (@azuma_eth)

On July 29, Beijing time, SK Hynix announced its second-quarter 2026 financial results.

The financial data shows that SK Hynix recorded revenue of 79.32 trillion KRW in the second quarter, up 257% year-over-year and 51% quarter-over-quarter; operating profit of 60.54 trillion KRW, up 557% year-over-year and 61% quarter-over-quarter, with the operating profit margin further rising to 76%, a historic high; including the one-time investment gain of 62.166 trillion KRW from the sale of part of its stake in Kioxia, the company's net profit reached 93.92 trillion KRW.

In any industry, this would be a report card shocking enough for the market.

However, the initial reaction from the capital market was quite the opposite. As revenue (actual 79.32 trillion KRW vs. market expectation of 84 trillion KRW) and operating profit (actual 60.54 trillion KRW vs. market expectation of 64 trillion KRW) both slightly missed market expectations, coupled with the fact that SK Hynix's stock price had already pulled back over 40% in the previous month, pessimistic sentiment intertwined. After the earnings release, SK Hynix's US ADR stock price fell about 9% in after-hours trading (it had already fallen nearly 9% in regular trading yesterday). However, as investors gradually digested the earnings details, the stock price quickly recovered all losses and even turned positive.

Meanwhile, after the opening of the South Korean stock market this morning, SK Hynix's stock price initially opened higher, rising 4%, but then gradually weakened. As of 10:00 AM, it had fallen over 9% again.

Why did a record-breaking earnings report first face heavy selling, then quickly recover its losses, and then sharply turn downward again? The answer may lie in the fact that what the market truly cares about is far more than how much money SK Hynix made in the second quarter; it is about how to re-price future growth potential – and the bulls and bears have clearly not yet reached a consensus on this point.

The Most Profitable Quarter Ever: Why Did It Still Miss Expectations?

Looking purely at the numbers, SK Hynix is almost still at its peak profitability stage.

In the second quarter, the company's gross margin reached 83%, and its operating profit margin reached 76%. This means that for every 100 KRW of products sold, about 76 KRW was converted into operating profit. This profitability level even surpasses the vast majority of global semiconductor manufacturers. At the same time, the company's cash and short-term financial assets continued to grow rapidly to 87.96 trillion KRW, further expanding its net cash position and providing ample ammunition for future capacity expansion.

But the problem is that the market had already set expectations even higher. Previously, the consensus expectation was for SK Hynix's second-quarter revenue to be around 84 trillion KRW and operating profit around 64 trillion KRW. The final actual figures were about 5% and 6% lower than these expectations, respectively.

For most companies, such a deviation is not significant. However, for SK Hynix, which has been labeled the "biggest beneficiary of AI" and whose valuation is built on high growth expectations, any data falling short of expectations gets magnified by the market.

A closer look at the earnings report reveals that this "miss" did not stem from a deterioration in market demand but rather from changes in the profit structure.

First, a counterintuitive point: the continuously increasing share of HBM products has actually weakened profit elasticity. In past quarters, a key driver of the rapid profit expansion in the entire memory industry was the sustained price increase of traditional DRAM and NAND spot markets. However, because SK Hynix's HBM revenue share is significantly higher than its peers, and HBM relies more on long-term supply agreement (LTA) pricing, it cannot fully capture the benefits of rapid spot price increases like ordinary DRAM can.

Additionally, SK Hynix disclosed that the average selling price of ordinary DRAM in the second quarter increased by about 30% quarter-over-quarter. While still growing, this was a significant deceleration from the first quarter. Similarly, the average selling price of NAND increased by 50%-55% quarter-over-quarter, also slower than the first quarter.

In other words, more AI products were sold, but the pace of price increases for traditional products slowed down; long-term contracts secure future revenue but also limit short-term profit elasticity. This is why a record profit still failed to meet the numbers the market had previously "fantasized" about.

Is the Memory Super Cycle Still On? What the Earnings Tell Us

If the operational data answers how much money SK Hynix made in the second quarter, then the information provided by management in the earnings report and subsequent conference call answers another question the market is more concerned about – is the AI memory super cycle starting to cool down?

For now, SK Hynix's answer remains leaning towards optimism.

First, regarding demand outlook, the company did not release any obvious cautious signals as the market had feared. SK Hynix forecasts that global DRAM market demand in 2026 will still grow by mid-20% year-over-year, and NAND market demand will grow by high-teen percent year-over-year. During the post-earnings conference call, management also stated that they have not yet observed any signs of an AI investment slowdown and expect AI infrastructure investment to maintain steady growth even after 2027.

Secondly, another key information worth noting is the further advancement of long-term supply agreements (LTAs). SK Hynix disclosed that it has completed LTA negotiations with approximately ten customers and is continuing discussions with other major industry clients. The new generation of long-term agreements will adopt pricing mechanisms capable of responding to price fluctuations and will utilize corresponding financial mechanisms to ensure contract performance, thereby enhancing the stability and predictability of future demand.

For the memory industry, the significance of this change is considerable. In the past, products like DRAM and NAND relied more on spot market pricing, and violent price fluctuations prevented the entire industry from shedding its "cyclical stock" label. However, with the increasing share of HBM products in the AI era, more and more large cloud companies are starting to lock in supply capacity years in advance. Supply-demand dynamics are gradually evolving from short-term bargaining into more long-term and stable cooperative relationships. While long-term agreements can compress profit elasticity during periods of rapid spot price increases, as seen this quarter, they exchange this for higher revenue certainty in the coming years.

Furthermore, the rollout pace of SK Hynix's next-generation products proceeded without any surprises. The earnings report shows that SK Hynix began shipping HBM4 products in the second quarter and plans to ramp up volume fully in the second half of the year; the next-generation HBM4E has already been sampled to major customers in the first half of the year; additionally, SO-DIMM2 products based on the 1c-nm process have also begun supply.

This means SK Hynix maintains its leading product cadence for the next-generation AI GPU platforms. Considering that HBM4 will be a crucial companion memory for NVIDIA's next-generation AI platform like Rubin, its smooth volume ramping also indicates that the company currently firmly holds its leading position in the high-end AI memory market.

Finally, in terms of "Capital Expenditure" (CapEx), which is the most accurate reflection of management's true judgment, SK Hynix not only maintained its expectation for 2026 CapEx to be in the upper end of the over 40 trillion KRW range, but also plans to advance the mass production schedule for the M15X fab, accelerate the construction of the Yongin Fab Phase 1, and continue to push forward medium- to long-term projects such as P&T7, M17, and the new semiconductor cluster in South Korea.

For a company that has experienced multiple memory cycles, such an aggressive expansion plan is itself a statement – Management remains confident that the AI memory demand in the coming years will be sufficient to absorb this new capacity.

The Focal Point of Bull-Bear Battles

Today, SK Hynix has become a core battleground for bulls and bears in the AI memory cycle.

For the bulls, record profits, continuously expanding HBM demand, and the AI infrastructure investment cycle still support the company's long-term growth thesis. For the bears, the earnings miss, valuation pressures, and concerns about the sustainability of AI capital expenditure are amplifying short-term adjustment pressures. Bulls are betting that the expansion of AI infrastructure will continue, while bears worry that the market has already overdrawn future growth.

Heavy is the head that wears the crown. SK Hynix enjoys the valuation of an industry leader but must also bear the pressures that come with it – when the market already believes in your story, excellent performance is no longer enough; only consistently exceeding even higher expectations can drive the valuation higher.