SK Hynix misses profit forecasts, shares slump 10% despite AI demand

The memory chipmaker posted a 557% surge in operating profit, but slower HBM4 shipments left analysts wanting more

SK Hynix just delivered one of the most impressive earnings reports in semiconductor history, and the market punished it anyway. The South Korean memory giant posted Q2 2026 operating profit of 60.5 trillion won, a 557% increase from the 9.2 trillion won it earned in the same quarter last year, and investors responded by dumping shares nearly 10%.

The problem wasn’t the profit. It was the gap between what SK Hynix delivered and what Wall Street expected. The LSEG SmartEstimate consensus had pegged operating profit at 64 trillion won, meaning the company came in roughly 3.5 trillion won short. Revenue told a similar story: 79.3 trillion won represented a 257% year-over-year jump, but it still missed the 84 trillion won estimate.

The culprit behind the shortfall was slower-than-expected shipments of high-bandwidth memory, specifically the next-generation HBM4 chips that AI infrastructure builders have been clamoring for. Revenue recognition got delayed even though underlying demand remained strong.

Think of it like a restaurant with a packed reservation book but a kitchen that’s running behind on orders. The customers are there. The appetite is real. The food just isn’t hitting the table fast enough to ring up the sales.

For crypto investors watching from the sidelines, this matters more than it might seem. HBM chips are the critical memory component in the GPU and accelerator systems that power everything from AI model training to high-performance computing workloads. When HBM shipments slow, it signals potential bottlenecks in the broader AI infrastructure buildout, the same infrastructure that increasingly intersects with crypto mining, on-chain AI applications, and the GPU compute economy that protocols like Render and Akash depend on.

Shares of SK Hynix closed down 9.6% on July 29, 2026. The stock is now more than 50% below its high from the previous month, though it remains up approximately 115% year-to-date. That kind of volatility in a blue-chip semiconductor name tells you something about how fragile sentiment around AI infrastructure spending has become.

The long game looks different from the short game

Here’s the thing. If you zoom out past the quarterly miss, SK Hynix is positioning itself for multi-year dominance in the AI memory market. The company disclosed that it has signed approximately 10 long-term supply agreements, typically spanning five years, with major clients. That’s the kind of revenue visibility most companies would trade a kidney for.

The capex numbers reinforce that confidence. SK Hynix raised its 2026 capital expenditure guidance to the high-40 trillion won range, roughly $27.6 billion or more. That represents an increase of approximately 50% compared to the 30.2 trillion won allocated for 2025. You don’t spend that kind of money unless you’re expecting demand to stay elevated for years, not quarters.

The balance sheet backs up the aggression. SK Hynix reported a net cash position of 88 trillion won, bolstered by gains from a recent sale of its stake in Kioxia, the Japanese memory manufacturer formerly known as Toshiba Memory. That cash pile gives the company significant runway to invest in next-generation manufacturing capacity without taking on the leverage risks that have historically plagued cyclical chipmakers.

No shareholder return policy has been announced yet, though the company indicated plans to address that later in 2026. For now, the message is clear: capital goes to building, not buying back stock.

The SK Hynix earnings miss highlights a tension that’s defining the entire AI trade right now. Demand is real, but execution timelines are slipping. Hyperscalers, the Microsofts and Googles and Metas of the world, are still spending aggressively on AI infrastructure. But the supply chain from advanced packaging to memory to cooling systems is struggling to keep pace.

For crypto markets specifically, the AI infrastructure narrative has been a meaningful driver of sentiment. Tokens tied to decentralized compute, AI agents, and GPU marketplaces have traded as proxies for the broader AI buildout. When a bellwether like SK Hynix signals that the buildout is happening slower than expected, it introduces uncertainty into those correlated trades.

The 50% drawdown from last month’s high in SK Hynix shares also serves as a reminder about what happens when a consensus trade gets crowded. AI memory was one of the most popular institutional bets in global equity markets. The same dynamic plays out constantly in crypto, where narratives attract capital until the slightest disappointment triggers violent repricing.

Watch the HBM4 ramp closely over the next quarter. If SK Hynix can accelerate shipments and close the gap to analyst estimates, the stock’s pullback could look like a buying opportunity in retrospect. If delays persist, it may signal a broader slowdown in the pace of AI infrastructure deployment, one that would ripple across both traditional semiconductor stocks and the crypto tokens riding the same narrative wave.

The long-term supply agreements and massive capex commitment suggest SK Hynix’s management sees this as a timing issue, not a demand problem. Markets, as usual, aren’t feeling that patient.