Data cutoff: October 28, 2026, 9:55 a.m. Asia/Shanghai (October 27, 9:55 p.m. ET). SK hynix announced its results on October 28 in Seoul; the latest completed U.S. regular trading session was October 27.
A memory supplier has become a critical test of the AI buildout
SK hynix's third-quarter results, announced in Seoul on October 28, are among the most consequential semiconductor disclosures of this earnings window. The company reported revenue of KRW 104.6155 trillion and operating profit of KRW 83.3338 trillion. Its operating margin reached approximately 80%, and quarterly revenue exceeded KRW 100 trillion for the first time. Those numbers represent a much bigger change than an ordinary earnings beat. They suggest that the most advanced memory products have become scarce enough to support exceptional profitability at a major global supplier.
The company recorded net income of KRW 39.7756 trillion. That is positive growth from the year-earlier period but a decline of about 58% from the unusually high second-quarter net income of KRW 93.9226 trillion. Net income and operating profit measure different things, and this divergence should not be flattened into a claim that the operating business suddenly weakened. The operating result grew 38% from the second quarter. Non-operating investment-related effects and the accounting composition of profit must be analyzed separately.
What the quarter actually showed
Revenue rose 32% quarter over quarter from KRW 79.3187 trillion, while operating profit rose 38% from KRW 60.5426 trillion. Versus the third quarter of 2025, revenue was up 328% and operating profit was up 632%, according to the company’s preliminary K-IFRS release. The operating-margin comparison is equally dramatic: 80% versus approximately 47% a year earlier. These are consolidated figures and should not be mistaken for a pure HBM product margin.
SK hynix attributed the performance to stronger prices for key memory categories and greater sales of high-value products. Server DDR5 demand helped DRAM results, while HBM4 shipments moved into a more substantial phase. In NAND, 321-layer TLC enterprise SSD products expanded. This matters because the current upcycle is not solely an HBM story. Enterprise DRAM, AI-related SSD demand and a richer product mix are also contributing to earnings.
Why HBM4 is the next catalyst, not just another product name
High-bandwidth memory is packaged close to powerful AI accelerators so that processors can access data faster and with greater bandwidth. Its production is more complex than conventional commodity DRAM, and capacity decisions involve advanced processing, stacking, packaging and qualification. Those constraints can support better pricing when demand rises quickly.
SK hynix expects HBM4 to surpass HBM3E in both shipment volume and revenue in the fourth quarter of 2026. It plans to ramp the M15X facility to maximum capacity by year-end to support this transition. Those are management expectations, not realized fourth-quarter results. The distinction is important: successful qualification and yield improvement must still convert into commercial output.
The U.S. market read-through reaches Nvidia and other accelerator designers, which depend on memory availability for high-end systems. Micron is a directly relevant U.S.-listed competitor, while Samsung is another major memory supplier. A tight HBM market can improve suppliers' earnings and constrain accelerator shipments simultaneously. It may be good for a memory producer without being unambiguously good for every downstream technology company.
Long-term contracts change the durability debate
One of the most important disclosures is not in the income statement. SK hynix said it has signed long-term agreements with 14 major customers, including cloud providers, covering more than 50% of its total production capacity for 2027 and 2028. These agreements give the company greater demand visibility, though the release does not say that all future pricing is fixed at today's elevated levels. It describes arrangements that can reflect changing market conditions within agreed price ranges.
This is a meaningful structural shift. Traditional memory producers have often suffered severe boom-and-bust cycles because factories commit capital years before demand becomes visible. Multi-year customer commitments can moderate that risk, but they cannot eliminate it. Customers may change product mix, supply conditions may improve and future contracts may have pricing mechanisms that share gains or losses.
Management believes memory demand could exceed supply through 2030 because leading-edge processes are technically difficult and new facilities take years to build. This is a company forecast rather than an independently established fact. Investors should test it against the pace of capacity expansion by SK hynix, Samsung, Micron and other suppliers.
The shareholder-return announcement is large, but the components matter
SK hynix also laid out an approximately KRW 100 trillion shareholder-return program for 2026. Its described components include KRW 30 trillion or more in cash dividends and KRW 60 trillion or more in repurchases and cancellations, including KRW 40 trillion announced in August. The "approximately 100 trillion" headline is a rounded program description rather than a license to add every board action twice.
For the third quarter, the board approved a KRW 14 trillion cash dividend and KRW 20 trillion of share purchases. Of the latter, about KRW 10 trillion is intended for cancellation; another roughly KRW 10 trillion is for employee compensation and is explicitly excluded from the shareholder-return total. The dividend includes a special component tied to realized gains on the Kioxia investment. Management also discontinued its review of a Solidigm IPO, saying current market conditions did not support the originally expected outcome.
Investors should distinguish recurring operating cash generation from returns of non-operating investment gains. Both can benefit shareholders, but only the first offers a durable basis for future capital allocation.
Risks: record margins invite new supply
An 80% operating margin is impressive, but it also creates incentives for competitors to increase capacity and for customers to accelerate second-source qualification. A major risk is that HBM and conventional DRAM supply eventually catch up with demand faster than the company's long-term shortage thesis anticipates. Yield improvements can also lift effective supply without equivalent new factory construction.
Demand concentration remains another risk. AI infrastructure spending is heavily influenced by a small number of cloud and accelerator platforms. Long-term agreements reduce uncertainty but do not make hyperscaler capital budgets infinite. Rising memory prices can also hurt PC and smartphone unit demand, so strength in data centers can coexist with weakness in consumer electronics.
SK hynix expects investment to increase significantly in 2027 as its Yongin project and other infrastructure expand. That may support revenue growth, but higher spending could reduce free cash flow during construction even if accounting profits remain strong. Currency movements, export restrictions and process-transition yields introduce additional uncertainty.
What to watch next
The clearest near-term test is the fourth-quarter HBM4 transition: whether the product actually overtakes HBM3E in revenue and shipments as management anticipates. Investors should also look for confirmation that M15X ramps on schedule and that 321-layer enterprise SSD production scales successfully.
For 2027, watch whether the 14 long-term customer agreements translate into steady shipments and resilient prices, rather than only headline contracted capacity. Compare SK hynix's margin trends with Micron and Samsung, and track whether customers continue signing multiyear commitments as new supply enters the market. A sustained margin advantage would be more persuasive than one exceptional quarterly number.
Finally, follow the cadence of shareholder returns versus investment requirements. The ability to distribute large amounts of cash while funding advanced memory capacity would strengthen the investment case, but reported net income should not be treated as interchangeable with free cash flow.
Conclusion
The SK hynix release is a leading indicator for the economics of the entire AI hardware stack. Revenue above KRW 100 trillion, an 80% operating margin, the HBM4 transition and production agreements covering more than half of 2027–28 capacity point to unusually strong pricing and demand visibility. The unanswered question is durability. If high-value memory remains scarce while new capacity ramps carefully, the company's earnings power may persist. If capacity catches up or hyperscaler spending slows, today's extraordinary margin could normalize rapidly. That tension makes this one of the most important results of the week for investors following AI semiconductors.