SK Hynix reported a record 76% operating margin for the second quarter, an extraordinary result for a memory maker. Revenue rose 257% from a year earlier, while operating profit climbed 557%.
Those figures show that demand for AI memory is strong and that SK Hynix has unusual pricing power. They do not settle the more valuable question for chip investors: can the cloud companies buying AI systems generate enough revenue from them to keep spending at this pace? Microsoft and Meta report after the U.S. market closes today. Their results will be the next test of whether the AI hardware boom has a durable business case beyond the companies supplying it.
Record Results Did Not End the Debate
Tuesday offered a sharp reminder that strong supplier earnings do not erase broader concerns about technology valuations. South Korea's KOSPI fell 10.84%, and the Korea Exchange halted trading for 20 minutes as the selloff deepened.
That is the pressure point for SK Hynix. Advanced memory has become a key part of AI systems, and current demand is clearly profitable. Yet a memory cycle can change quickly when supply rises or large customers adjust their buying plans.
The 76% operating margin is more useful than the company's headline net-profit figure. SK Hynix said net profit exceeded revenue in the second quarter, an unusual outcome that makes it a poor starting point for judging recurring chip economics. The company did not provide enough detail in the reported facts to show how much of that net-profit figure came from operations versus other items.
Operating margin measures profit from the company's core business after operating costs. At this level, it indicates that sales prices and product mix are producing far more profit per dollar of revenue than in a normal memory market.
Operating Profit Shows the Current Demand Picture
SK Hynix's operating profit grew more than twice as fast as revenue in the quarter. Revenue growth of 257% is already remarkable. The 557% increase in operating profit shows that the company kept a much larger share of each sales dollar after production and selling costs.
That gap points to operating leverage, where revenue rises faster than costs. It also fits a market where the most advanced memory is scarce and customers are willing to pay heavily to secure supply.
Memory has long been a cyclical industry. Suppliers can earn outsized margins when demand rises faster than output, then face falling prices when new capacity reaches the market or customers work through inventories. SK Hynix's quarter does not show how much future HBM4 supply will enter the market, what customers have agreed to pay, or how long current pricing can last.
HBM4 Adds Evidence, With Limits
SK Hynix began mass shipments of HBM4 during the second quarter. HBM4 is a high-bandwidth memory product used in AI systems. Moving from development into mass shipments gives the demand case more support than a single quarter of strong pricing.
The company plans to ramp HBM4 production in the second half of 2026. That plan matters because a production ramp will test whether demand can absorb more supply at attractive economics, but customers must keep buying enough advanced memory to absorb that supply.
SK Hynix has also finalized long-term agreements with around 10 customers and remains in talks with other major clients. Those agreements give the company more visibility than relying only on short-term orders.
SK Hynix did not disclose the prices, volumes, or duration of those agreements. A long-term contract can secure access to a customer without proving that shipment volume will remain high or that pricing will preserve today's margin. Investors cannot yet tell how much of the current profitability is protected as HBM4 production rises.
The Buyers Must Now Show Their Returns
Microsoft reports fiscal fourth-quarter 2026 results after the close today. Meta also holds its second-quarter earnings call after the close. Their reports move the AI discussion from the supplier's income statement to the buyer's capital budget.
Microsoft's cloud commentary will carry particular weight. Cloud computing offers a direct route for AI capacity to become revenue, as customers pay for computing services and related software. Strong cloud growth and concrete AI revenue commentary would strengthen the case that data-center investment can earn a return beyond internal use.
Meta faces a different test. Its AI spending is tied more closely to improving its existing products and building future services. Management's explanation of how AI investment supports its business will matter alongside any discussion of future capital spending.
But a wide gap between rising spending and weak signs of revenue or product gains would make the supplier earnings story less secure.
What Tonight Can Confirm
Investors should separate two questions that can look similar in a chip earnings release. The first is whether AI memory demand is strong today. SK Hynix's quarter answers that clearly.
Higher planned investment would support the near-term demand case for SKHY. Clear signs of cloud growth, AI-related revenue, or business improvement tied to AI would provide a stronger foundation for the longer cycle. Thin monetization commentary or a slower spending outlook would leave SK Hynix's record quarter looking more dependent on present scarcity.
After the close, watch for Microsoft to quantify AI-related cloud revenue and for Meta to pair its 2026 capital-spending outlook with measurable advertising, engagement, or product gains tied to AI.
