SK Hynix posts record Q2 profit, driven by AI memory (HBM) boom, despite stock decline from missing consensus forecasts; reaffirms AI demand strength in supply chain.
SK hynix reported record second-quarter results on Wednesday as surging demand for high-bandwidth memory (HBM) chips used in AI servers reshaped the semiconductor industry. The South Korean chipmaker reported revenue of 79.3187 trillion won and operating profit of 60.5426 trillion won, up 257% and 557% respectively year-over-year, with first-half revenue crossing 100 trillion won for the first time in company history. Despite record earnings, SK hynix shares dropped post-announcement as investors had expected even stronger results, reflecting how elevated expectations have become in the AI chip sector.
High Bandwidth Memory has emerged as essential infrastructure for AI. By vertically stacking memory chips, HBM provides significantly more bandwidth and better power efficiency than regular DRAM, enabling Nvidia and other chipmakers' AI accelerators to operate with vastly improved efficiency. As the leading HBM supplier for Nvidia, SK hynix serves as a strong indicator of AI infrastructure demand. The company reported growing demand for its products as AI advances toward more sophisticated "agentic" systems capable of performing higher-level functions. To meet future demand, SK hynix has signed long-term contracts with approximately 10 major clients, granting them the right to reserve production capacity years in advance.
The global memory market reflects this same trend. TrendForce anticipates DRAM contract prices rising 58% to 63% for the third quarter, with NAND flash prices rising 70% to 75%. HBM and server memory are being prioritized over consumer applications, with price increases driven by limited market availability, hyperscale cloud investments, and long-term contracts.
SK hynix is not alone in benefiting. Samsung Electronics expects operating profit of 89.4 trillion won and revenue of approximately 171 trillion won for the second quarter, gaining from strong AI memory demand. Samsung has taken an aggressive stance on HBM4, becoming the first company to produce it commercially in large quantities after beginning mass production in February. The company has already generated more than $1 billion in HBM4 sales within four months of product release. Samsung's commercial HBM4 offers data transfer performance of 11.7 Gbps per pin, exceeding the JEDEC minimum of 8 Gbps, with superior power and thermal efficiency. Micron Technology has also published record quarterly results, with CEO Sanjay Mehrotra calling memory "a strategic asset" in the age of artificial intelligence and noting that hyperscale customers are signing more multi-year supply contracts to ensure future supplies.
This AI boom is creating shortages elsewhere. IDC warns of an "unprecedented memory chip shortage" potentially extending to 2027, as manufacturers devote increasing production to HBM for AI accelerators rather than PCs and smartphones. Global DRAM supply is projected to increase roughly 16% and NAND supply about 17%, both below historical averages. As supply tightens, smartphone manufacturers face a choice: absorb higher component costs, lower device specifications, or pass increases to consumers. In IDC's downside scenario, worldwide smartphone shipments could drop as much as 5.2%.
Competition in next-generation AI memory is intensifying. SK hynix began mass shipments of HBM4 during the quarter and plans to expand production in the second half, with chips meeting customer performance requirements while improving power efficiency. The company has completed sample shipments of HBM4E for future AI accelerators. Samsung, having begun HBM4 commercial production first, has gained a competitive edge; TrendForce predicts Samsung will earn more than $1.2 billion from HBM4 sales this year, becoming the first supplier to exceed $1 billion. SK hynix's strategy prioritizes the HBM3E market before expanding HBM4 production, with plans to ramp up from Q3 onwards. Looking ahead, SK hynix intends to expand advanced chip packaging while improving NAND production through strict budget planning.
The advent of AI has fundamentally transformed memory chip manufacturers, shifting focus from mere production capacity to delivering the next generation of faster, better-performing AI memory chips to market.