These results were driven by surging AI infrastructure demand, tight memory supply, and strong DRAM and NAND price increases . DRAM average selling prices rose about 30% quarter-over-quarter, while NAND prices climbed in the mid-50% range .

Why They Missed Estimates — and Why the Stock Fell 9.6%

Despite these record numbers, both revenue and operating profit fell short of consensus expectations:

Source:

The reason is not company weakness but sky-high investor expectations. The AI hardware boom has pushed analyst forecasts to extremely aggressive levels. SK Hynix's own performance was historic, but the market had already priced in even more . The stock fell 9.6% on the day — as much as 15% intraday — and the miss also dragged down U.S. memory peers Micron and SanDisk .

"The results fell short of brokerages' consensus estimates of 83.94 trillion won in revenue and 63.99 trillion won in operating profit, " reported The Korea Times .

HBM Focus and Multi-Year Supply Agreements

SK Hynix's strategy centers on high-bandwidth memory (HBM), the critical component for training large AI models, and locking in demand years ahead through long-term agreements:

• HBM4 commercial debut: SK Hynix began mass production shipments of HBM4 in Q2, making this the first quarter with commercial HBM4 revenue . Yields and quality are already approaching those of mature HBM3E .
• HBM4E sampling: In June 2026, the company shipped samples of 12-layer HBM4E chips to a major customer, with mass production targeted for 2027 .
• Long-term agreements (LTAs): SK Hynix has signed or is negotiating multi-year supply contracts with about 10 key customers. These are typically 5-year frameworks covering volume commitments, pricing mechanisms, and deposits . Management said the agreements provide a "demand floor" and reduce performance uncertainty .
• Demand outstrips supply: Requests from clients for HBM chips over the next three years already "greatly exceed" SK Hynix's production capabilities, the company stated during its Q1 2026 earnings call .
• No AI slowdown seen: Management stated explicitly, "We currently do not see any signs of a slowdown in AI investment," expecting AI infrastructure buildout to remain robust beyond 2027 .

HBM4 and HBM4E Product Outlook

• HBM4 (current-gen): Mass production shipments began in Q2 2026. Ramp-up will accelerate in the second half of 2026. The company claims industry-leading power efficiency and cost competitiveness .
• HBM4E (next-gen): Samples of 12-layer HBM4E were shipped to a major customer in June 2026 . Mass production is planned for 2027 . SK Hynix is pulling forward its sample timeline to maintain leadership ahead of rivals like Samsung and Micron .
• Future roadmap: Management flagged that advanced process complexity and facility lead times are lengthening, so the company is investing aggressively in new capacity to meet structural demand .

Broader Memory Industry Tension

These results highlight a structural divide in the memory industry that goes beyond SK Hynix itself:

Premium AI chip focus (HBM and AI server DRAM):

• HBM is the primary profit driver, with nearly insatiable demand from hyperscalers building AI clusters .
• SK Hynix essentially owns this segment — it is Nvidia's lead HBM supplier and has locked in multi-year contracts that guarantee high prices and volumes .
• The 76% operating margin reflects a premium product with pricing power completely detached from the commodity memory cycle .

Conventional memory (PC, mobile, general server DRAM/NAND):

• The recovery in conventional memory is real (strong price gains contributed to the revenue surge) but remains tied to the broader supply-demand cycle .
• There is growing market concern that the conventional memory recovery may be peaking, and that non-AI end markets are still soft .
• The "miss" itself is partly a reflection of this tension: investors want to see that the AI boom is pulling up the rest of the memory business, not just HBM. The stock selloff indicates skepticism that the broader memory cycle can sustain the current pace .

In short, SK Hynix is printing money on HBM and has locked in demand years ahead, but the market is now asking whether conventional memory can keep up — and whether AI infrastructure spending can continue to grow at the exponential rate already priced into the stock.

These results were driven by surging AI infrastructure demand, tight memory supply, and strong DRAM and NAND price increases . DRAM average selling prices rose about 30% quarter-over-quarter, while NAND prices climbed in the mid-50% range .

Why They Missed Estimates — and Why the Stock Fell 9.6%

Despite these record numbers, both revenue and operating profit fell short of consensus expectations:

Source:

The reason is not company weakness but sky-high investor expectations. The AI hardware boom has pushed analyst forecasts to extremely aggressive levels. SK Hynix's own performance was historic, but the market had already priced in even more . The stock fell 9.6% on the day — as much as 15% intraday — and the miss also dragged down U.S. memory peers Micron and SanDisk .

"The results fell short of brokerages' consensus estimates of 83.94 trillion won in revenue and 63.99 trillion won in operating profit, " reported The Korea Times .

HBM Focus and Multi-Year Supply Agreements

SK Hynix's strategy centers on high-bandwidth memory (HBM), the critical component for training large AI models, and locking in demand years ahead through long-term agreements:

• HBM4 commercial debut: SK Hynix began mass production shipments of HBM4 in Q2, making this the first quarter with commercial HBM4 revenue . Yields and quality are already approaching those of mature HBM3E .
• HBM4E sampling: In June 2026, the company shipped samples of 12-layer HBM4E chips to a major customer, with mass production targeted for 2027 .
• Long-term agreements (LTAs): SK Hynix has signed or is negotiating multi-year supply contracts with about 10 key customers. These are typically 5-year frameworks covering volume commitments, pricing mechanisms, and deposits . Management said the agreements provide a "demand floor" and reduce performance uncertainty .
• Demand outstrips supply: Requests from clients for HBM chips over the next three years already "greatly exceed" SK Hynix's production capabilities, the company stated during its Q1 2026 earnings call .
• No AI slowdown seen: Management stated explicitly, "We currently do not see any signs of a slowdown in AI investment," expecting AI infrastructure buildout to remain robust beyond 2027 .

HBM4 and HBM4E Product Outlook

• HBM4 (current-gen): Mass production shipments began in Q2 2026. Ramp-up will accelerate in the second half of 2026. The company claims industry-leading power efficiency and cost competitiveness .
• HBM4E (next-gen): Samples of 12-layer HBM4E were shipped to a major customer in June 2026 . Mass production is planned for 2027 . SK Hynix is pulling forward its sample timeline to maintain leadership ahead of rivals like Samsung and Micron .
• Future roadmap: Management flagged that advanced process complexity and facility lead times are lengthening, so the company is investing aggressively in new capacity to meet structural demand .

Broader Memory Industry Tension

These results highlight a structural divide in the memory industry that goes beyond SK Hynix itself:

Premium AI chip focus (HBM and AI server DRAM):

• HBM is the primary profit driver, with nearly insatiable demand from hyperscalers building AI clusters .
• SK Hynix essentially owns this segment — it is Nvidia's lead HBM supplier and has locked in multi-year contracts that guarantee high prices and volumes .
• The 76% operating margin reflects a premium product with pricing power completely detached from the commodity memory cycle .

Conventional memory (PC, mobile, general server DRAM/NAND):

• The recovery in conventional memory is real (strong price gains contributed to the revenue surge) but remains tied to the broader supply-demand cycle .
• There is growing market concern that the conventional memory recovery may be peaking, and that non-AI end markets are still soft .
• The "miss" itself is partly a reflection of this tension: investors want to see that the AI boom is pulling up the rest of the memory business, not just HBM. The stock selloff indicates skepticism that the broader memory cycle can sustain the current pace .

In short, SK Hynix is printing money on HBM and has locked in demand years ahead, but the market is now asking whether conventional memory can keep up — and whether AI infrastructure spending can continue to grow at the exponential rate already priced into the stock.