SK hynix eyes record Q1 as AI memory prices surge

SK hynix is set to report Q1 results with analysts expecting record sales of 47.7 trillion won (US$32.4 billion) and operating profit of 32.7 trillion won (US$22.2 billion) as AI demand lifts memory prices.

The estimates would tops its previous quarterly highs from Q4 and mark year-on-year rises of 170% in sales and 339% in operating profit, driven by high-bandwidth memory shipments and higher prices for DRAM and NAND products.

Investors are weighing whether the cycle is near a peak, though analysts said recent spot price drops for PC DRAM in China do not reflect contract pricing, with long-term supply deals seen as supporting the upcycle.

🔗 Source: The Korea Times

🧠 Food for thought

Implications, context, and why it matters.

SK hynix’s profits keep rising on strong AI memory leadership

Goldman Sachs expects SK hynix to hold more than 50% of the total High Bandwidth Memory (HBM) market through at least 2026 1.
UBS says SK hynix will supply HBM3E first for Google’s latest Tensor Processing Units (TPUs), its AI chips called v7p and v7e. UBS also expects SK hynix to take about 70% of the HBM4 market for NVIDIA’s next-generation Rubin platform, its upcoming AI chip system, in 2026 1.
SK hynix has committed US$15 billion to expand next-generation memory manufacturing capacity, with emphasis on HBM and advanced DRAM technologies 2.
SK hynix has finished development of HBM4 and says it is ready for high-volume manufacturing. The company adds that its HBM4 stacks beat standards set by JEDEC, the semiconductor industry standards body, by 25% in performance 3.

AI memory demand is changing how chips get built and supplied

Rising demand for high-margin HBM is shifting attention across the memory sector.
As chipmakers move capacity toward HBM, the supply-demand balance for general-purpose DRAM tightens, which lifts profitability across the industry 1.
The ramp also strains advanced packaging, since each HBM stack needs complex integration with processors, and packaging limits how fast AI hardware output can grow 2.
Goldman Sachs expects HBM use to spread beyond graphics processing units (GPUs). The firm projects demand from custom-ordered, application-specific integrated circuit (ASIC)-based AI chips will rise 82% and reach one-third of the HBM market in 2026 1.