SK Hynix shares rose 25 percent and Samsung Electronics gained more than 20 percent Friday, as investors increased bets on sustained demand for AI infrastructure. The surge in Seoul-listed chipmakers tracked a rally in U.S. technology stocks, reflecting renewed confidence in AI spending.
High-bandwidth memory, or HBM, remains a critical component for training and deploying large language models. These specialized DRAM chips are stacked vertically, delivering significantly higher data throughput than conventional memory—a capability essential for AI accelerators like those produced by Nvidia.
SK Hynix holds the leading position in HBM. The company has locked in significant supply agreements; CEO Kwak Noh-jung previously said SK Hynix has sold out its HBM production through the end of 2025. That sold-out posture, more than anything else, explains Friday's move.
Samsung is expanding its HBM production and investing in the advanced packaging technologies required for HBM integration. Both firms supply the global AI buildout, but SK Hynix's execution lead remains the market's focal point.
The broader technology market also gained. Nvidia rose 2.6 percent to $195.04, while Microsoft climbed 15.5 percent to $451.10. Tesla gained 3.5 percent to $308.85 and Amazon increased 3.9 percent to $235.50. The Nasdaq Composite advanced 2.8 percent to 25,122.
The valuation gap is worth noting for anyone doing the math. Samsung trades at 8.2 times forward earnings—a 30 percent discount to its five-year average. SK Hynix trades at 6.8 times, a striking number for a company with dominant share in the market's most constrained component. Both multiples suggest the market is still pricing in meaningful cyclical risk despite the AI demand signal.
The competitive moat in HBM production is real and durable. Advanced manufacturing processes, heavy capital expenditure and deep semiconductor packaging expertise keep the supplier base narrow. That concentration of market power is exactly what the valuation discount obscures.
Both companies are allocating capital to expand HBM fabrication capacity and fund research and development. Those spending decisions will directly determine future revenue and market share as AI infrastructure orders scale.
Upcoming earnings from major cloud providers and AI companies will sharpen the picture on infrastructure buildout pace. Any acceleration or deceleration in hyperscaler spending flows directly into memory and logic chip order books.
Risks remain. Oversupply in commodity DRAM could pressure overall memory pricing, and a broader economic slowdown could slow enterprise AI adoption—though demand at the frontier model level shows no signs of easing.
For South Korea's export-driven economy, the chipmakers' performance carries macro weight. Semiconductor exports are a primary driver of gross domestic product, and continued growth in this sector supports national economic targets.
