SK Hynix Delivers Record Quarter Fueled by AI Boom, But Revenue Falls Short of Expectations
South Korean memory chip powerhouse SK Hynix has once again posted stellar financial results, with record profits and revenues for the first quarter of the year. The surge is largely attributed to skyrocketing demand for its products, driven by the insatiable appetite of artificial intelligence (AI) applications. While earnings largely met analyst projections, the company’s top-line revenue slightly missed forecasts.
SK Hynix reported first-quarter revenue of 52.58 trillion won (approximately $35.55 billion), falling short of the LSEG smart estimate of 53.55 trillion won. However, operating profit came in strong at 37.61 trillion won, closely aligning with the estimated 37.92 trillion won.
The March quarter saw SK Hynix’s revenue nearly triple compared to the same period last year, marking a significant milestone as it surpassed the 50 trillion won threshold for the first time. Operating profit surged fivefold year-over-year and nearly doubled from the preceding quarter, propelling the company’s operating margin to an all-time high of 72%. This impressive performance saw SK Hynix’s shares initially rise as much as 3.6% in early trading on the South Korean market before settling down to a 0.9% decrease.
SK Hynix specializes in memory chips, crucial components for data storage across a wide spectrum of devices, from high-performance servers powering AI infrastructure to everyday smartphones and laptops. The company explicitly cited the escalating prices of memory chips and the burgeoning demand from AI as key drivers for its record earnings. As a leading global supplier of High Bandwidth Memory (HBM), essential for AI data centers, SK Hynix is strategically positioned at the forefront of this technological revolution.
“Despite the fact that the first quarter is typically a seasonal downturn, strong demand persisted due to expanded investments in AI infrastructure,” the company stated in its earnings release. Furthermore, SK Hynix anticipates continued growth in memory demand as AI evolves from large-scale model training to more sophisticated agentic AI, which requires real-time inference across diverse service environments.
“The importance of memory has become greater than ever… as this supply-demand imbalance persists, customers are prioritizing procurement over price,” an SK Hynix executive emphasized during the earnings call. HBM technology is a critical subset of Dynamic Random-Access Memory (DRAM), a fundamental semiconductor memory type utilized for data storage and program execution in PCs, workstations, and servers.
**The Fierce DRAM Market Duel**
Industry data from Counterpoint Research indicates that the DRAM market has experienced a robust 30% quarter-over-quarter growth for two consecutive periods, largely propelled by the upward trend in memory prices. The surging demand for HBM has been a primary catalyst, constraining manufacturing capacity and contributing to a broader memory shortage in recent quarters.
SK Hynix has carved out a significant competitive advantage over rivals such as Micron and Samsung Electronics in the DRAM market. This edge stems from its early leadership in HBM technology and its pivotal role as a key supplier to Nvidia, the world’s preeminent AI processor manufacturer.
However, the competitive landscape remains dynamic. Counterpoint Research data suggests that Samsung reclaimed the top spot in DRAM revenue from SK Hynix in the final three months of 2023. Despite this, SK Hynix maintained its dominance in the HBM segment, commanding an impressive 57% market share. Samsung’s stock reached a new intraday record of 227,000 won on Thursday.
In a significant development, Samsung announced in February its commencement of shipping its first HBM4 chips to undisclosed clients, nearly a year after SK Hynix began delivering its HBM4 samples. HBM4, representing the sixth generation of HBM technology, is the most advanced iteration to date and is expected to be the primary AI memory chip powering Nvidia’s next-generation Vera Rubin architecture, engineered for demanding AI workloads in data centers. SK Hynix has indicated its plans to begin supplying samples of its seventh-generation HBM4E in the latter half of this year, with mass production targeted for 2027.
**Navigating Capacity Constraints and Geopolitical Risks**
The global chip wafer shortage is projected to persist until 2030, as demand for HBM continues to outstrip supply and strain manufacturing capabilities. This outlook was echoed by SK Group Chairman Chey Tae-won in March 2023, who suggested that expanding wafer capacity could take at least four to five years, with a potential shortfall exceeding 20%.
To address the escalating demand, SK Hynix has been proactively expanding its production capacity. The company recently announced plans to invest approximately 19 trillion won in a new manufacturing plant in South Korea.
MS Hwang, a research analyst at Counterpoint Research, commented on the first-quarter results, stating that they “show strong profitability and reveal that a lot more memory is needed for AI inference than expected, with companies rushing to secure supply.” Hwang anticipates that even if memory price gains moderate in the latter half of the year, SK Hynix’s profitability could continue to grow.
However, potential headwinds could arise if the conflict in the Middle East intensifies and prolongs, further disrupting the supply of essential materials for semiconductor manufacturing. While not expected to be a long-term issue, such a scenario could critically impact the entire AI supply chain. SK Hynix has indicated that it has diversified its suppliers for key industrial inputs like helium, bromine, and tungsten, and has built up sufficient inventory, anticipating limited production impact. The company also highlighted its long-term liquefied natural gas agreements as a measure to mitigate potential increases in energy costs.
Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/20931.html
