On July 29, 2026, SK hynix reported revenue of KRW 79.3187 trillion and operating profit of KRW 60.5426 trillion for the April–June 2026 quarter. Both figures were record highs, with operating margin reaching 76%. Even so, the results fell short of market expectations. This is not an earnings report signaling a slowdown in HBM demand; rather, the overlap of HBM's annual pricing, a surge in commodity DRAM prices, and delays in shipping high-value-added products made sales volume and product mix harder to predict than before. The company has not disclosed the breakdown by factor, so the roughly 5% shortfall cannot be attributed to a single cause.
Record Profit, Yet a Shortfall of Around 5%
According to pre-earnings estimates compiled by Yonhap Infomax from 14 Korean securities firms, revenue was projected at KRW 84.1 trillion and operating profit at KRW 64.1 trillion. The gaps with actual results were KRW 4.7813 trillion and KRW 3.5574 trillion respectively, representing shortfalls of 5.7% and 5.5% versus forecasts.
| Metric | Q2 2026 (Apr–Jun) | Prior Forecast | Gap vs. Forecast | QoQ Change |
|---|---|---|---|---|
| Revenue | KRW 79.3187 trillion | KRW 84.1 trillion | 5.7% below | +51% |
| Operating Profit | KRW 60.5426 trillion | KRW 64.1 trillion | 5.5% below | +61% |
| Operating Margin | 76% | 75–77% | Within forecast range | +4 points |
What the table shows is that the margin itself did not collapse. Despite the revenue shortfall, the operating margin stayed within the forecast range, and year-over-year, revenue grew 257% while operating profit grew 557%. Operating margin also rose 35 points from 41% a year earlier. Calling this a "stinging failure" would be too strong given such robust profit growth.
That said, the roughly 5% forecast gap should not be dismissed either. The market had assumed SK hynix would capture even more of the historic price increases in DRAM and NAND. Since the actual margin came in within expectations, it is natural to conclude that the forecast gap arose mainly from a combination of sales volume, product mix, and average selling price.
Net profit reached KRW 93.9226 trillion, exceeding revenue. However, the company's preliminary earnings release does not break down non-operating gains and losses. In June, the consortium that invested in Kioxia sold its held shares, and there may also be valuation effects on related assets, so this net profit figure cannot be equated with the earning power of the memory business itself. The breakdown of non-operating profit will need to be confirmed in the detailed, pre-audit financial statements.
Commodity DRAM Overtook HBM in Per-Wafer Profitability
HBM is a high-value-added memory product that determines the performance of AI accelerators, but in the first half of 2026 commodity DRAM prices rose too quickly. According to TrendForce, contract prices for commodity DRAM rose roughly 93–98% quarter-over-quarter in Q1 2026 (Jan–Mar), with a forecast rise of 58–63% in Q2 (Apr–Jun). HBM, by contrast, is priced mainly on an annual basis under industry custom, and cannot immediately reflect quarterly market price increases. TrendForce analyzes that the company's HBM contract prices declined in 2026, which held back growth in the average selling price across DRAM overall.
The reversal also extended to per-wafer profitability. According to TrendForce estimates, HBM's revenue and profitability per wafer fell below that of 64GB DDR5 RDIMM in Q1 2026 (Jan–Mar). This does not mean HBM became technologically obsolete—rather, commodity DRAM prices overtook the HBM prices that had been set earlier.
Moreover, HBM consumes more manufacturing capacity to produce the same bit volume. TrendForce forecasts that HBM's share of the top three DRAM makers' wafer input will reach 22% by the end of 2026, while its share of bit supply will remain at just 9%. The more capacity is diverted to HBM, the scarcer commodity DRAM becomes, driving up prices for the remaining commodity supply. As a result, a paradox emerged: the companies most heavily weighted toward HBM had fewer opportunities to sell the now-surging commodity products.
SK hynix's concentration in HBM is high. According to a prospectus filed with the U.S. Securities and Exchange Commission, the company's global HBM revenue share in Q1 2026 (Jan–Mar) was 56.4% per IDC data. At the earnings call, the company also explained that shipments of some high-value-added products were pushed back to the second half of 2026, and that the resulting shift in product mix affected the company-wide average selling price. However, the company has not disclosed its HBM sales ratio or product-by-product pricing for the April–June quarter. HBM alone cannot account for the entire forecast gap; differences from shipment timing and product mix, including NAND, also remain.
Long-Term Contracts Reshape Upside and Investment Decisions
SK hynix has signed multi-year supply contracts with roughly 10 major customers. According to the earnings call, contract terms typically run about five years, but pricing methods vary by customer and product. Rather than relying solely on fixed pricing, the company says it is negotiating multiple approaches to accommodate market fluctuations.
In quarters where supply shortages exceed expectations, contracted volumes cannot be resold at the higher prevailing prices, meaning price increases for contracted supply can lag behind non-contract sales. TrendForce forecasts that multi-year contracts with U.S. cloud providers will limit price hikes for some customers, and that server DRAM price increases in Q3 2026 (Jul–Sep) will be concentrated among customers without contracts or in additional volumes outside contract terms. This illustrates how long-term contracts can dampen short-term upside, but it is unclear how many trillion won this contributed to the April–June forecast gap.
What the company directly explained at the earnings call was that shipments of some high-value-added products were delayed to the second half of 2026, and that the resulting shift in product mix affected the company-wide average selling price. No materials have been presented that separate out the price gap from long-term contracts versus the volume gap from shipment timing. It is therefore reasonable to interpret this shortfall as the combined result of contract pricing, shipment timing, and product mix including NAND.
On the other hand, multi-year contracts make it easier to forecast demand several years out. The memory industry has repeatedly cycled through periods of expanding capacity in response to perceived shortages, only to see demand cool and swing into oversupply and price collapse. Locking in visibility over supply volumes and contract terms in advance makes it easier to align massive capital investment with actual customer demand. That said, this is no guarantee against demand declines or price drops themselves.
SK hynix expects capital expenditure in 2026 to be in the high-KRW-40-trillion range and plans to open the first phase of its Yongin cleanroom in early 2027. As of the end of June, cash and cash equivalents stood at KRW 88 trillion, against liabilities of KRW 18.6 trillion, bringing net cash up to KRW 69.4 trillion. Whether the company can bring new facilities online as planned, leveraging this financial strength together with contract-based demand visibility, will determine the value of these long-term contracts.
HBM4 Mass Production, and the HBM5 Cooling Technology That Wasn't New News
On the product front, progress with HBM4 will shape future earnings. SK hynix announced that it began mass-production shipments of HBM4 in Q2 2026 (Apr–Jun) and will ramp up production in the second half. At the earnings call, the company explained that yield and quality are approaching the maturity level of HBM3, and that supply to major customers is progressing—though it disclosed no figures on shipment volumes, customer names, or revenue contribution. Whether the delay in high-value-added product shipments has been resolved will be determined by sales volume and company-wide average selling price in Q3 2026 (Jul–Sep).
The cooling technology for HBM5 also drew attention at the earnings call, but this was not a first disclosure. SK hynix had already announced "iHBM" on May 26. The technology places a silicon-based Integrated Cooling Element—which conducts heat without conducting electricity—near the D2D PHY connecting the HBM base die and the AI accelerator, creating a new heat dissipation pathway inside the package. The company states that this reduces thermal resistance by 30% compared to existing structures.
iHBM is said to build on the proven MR-MUF process and can be introduced without major changes to existing System-in-Package designs. However, the 30% figure is the company's own evaluation, and the mass-production timeline for HBM5, adopting customers, and yield rates have not been disclosed. Before a technology announcement translates into revenue, two remaining hurdles must be cleared: customer certification and mass-production profitability.
The figures to watch in the second half of 2026 are HBM4's revenue contribution, the company-wide average selling price, and whether the delay in high-value-added product shipments has been resolved. If these show improvement, it would reinforce the hypothesis of a product transition and mix-driven gap—but it would not serve as retroactive proof explaining the April–June shortfall. Only once factor-by-factor price and volume disclosures, HBM4 contract pricing for 2027, and the supply volume generated by the new Yongin facility can all be confirmed will it be possible to judge whether SK hynix's technological leadership in HBM has translated into returns commensurate with its investment.