SK Hynix delivered a record quarterly profit and issued a stark warning on supply constraints, signaling that demand for AI-critical memory chips is set to outstrip production capacity for years, countering concerns that spending by big tech firms may be peaking.
The world’s second-largest memory chipmaker after Samsung Electronics reported operating profit of 37.6 trillion won for the January–March quarter, a five-fold jump from 7.4 trillion won a year earlier and broadly in line with expectations. Revenue surged 198% to 52.6 trillion won, underscoring the scale of the current AI-driven cycle in semiconductors.
Demand for high-bandwidth memory, a critical component in AI chipsets supplied to firms such as Nvidia, has been buoying the surge. These chips are essential for handling the massive data throughput required by advanced AI models, making them one of the most supply-constrained segments of the semiconductor market.
“Client requests for (HBM) chip supplies over the next three years already far exceeds our production capacity,” said Ki Tae Kim, head of HBM sales and marketing, during an earnings call, highlighting the depth and duration of the supply-demand imbalance.
The statement rings a bell in the context of recent market anxiety that AI infrastructure spending could slow. Instead, SK Hynix’s outlook suggests that demand visibility remains strong over a multi-year horizon, with customers effectively pre-booking capacity amid fears of prolonged shortages.
Pricing dynamics underline that imbalance. According to TrendForce, contract prices for certain DRAM chips jumped nearly 83% in the first quarter from the previous quarter, while some NAND products saw price increases of around 160%. Prices are expected to rise again in the current quarter, pointing to another period of robust earnings.
Analysts caution that while the pace of price increases may moderate after the second quarter, the underlying constraint will persist. Semiconductor fabrication capacity takes more than a year to come online after construction begins, meaning supply cannot quickly adjust to demand spikes driven by AI infrastructure buildouts.
SK Hynix said it expects favorable pricing conditions to continue “for the time being,” with AI demand more than offsetting weaker demand from traditional end markets such as PCs and smartphones.
The company’s outlook aligns with broader industry warnings. Chey Tae-won recently said a global chip wafer shortage could persist until 2030, as AI-related demand continues to outpace supply expansion.
In response, SK Hynix is accelerating capital expenditure. The company said investment this year will rise significantly from 30.2 trillion won last year, with spending focused on expanding infrastructure at its Yongin semiconductor cluster, ramping up production at the M15X fabrication plant, and securing critical equipment.
A key part of that strategy is access to advanced lithography technology. In March, SK Hynix announced plans to purchase 11.95 trillion won worth of extreme ultraviolet (EUV) lithography systems from ASML by 2027. These machines are essential for producing next-generation chips, particularly in high-performance memory.
Beyond operations, the company has signaled a more shareholder-friendly stance, saying it is reviewing capital return measures including dividends, share buybacks, and share cancellations, with plans to finalize its approach within the year.
Geopolitical risks appear, for now, to be contained. SK Hynix said it expects limited impact from the conflict in the Middle East, citing diversified supply chains for key chemicals, secured energy through long-term contracts, and sufficient inventory buffers to manage price volatility.
Investor response has been positive. Shares of SK Hynix rose 2.1% in morning trade, outperforming the broader KOSPI, which gained 1.2%. The stock has surged nearly 90% so far this year, lifting the company’s market value to around $590 billion — surpassing ASML, Europe’s most valuable company.
The rally is seen as a reflection of a broader re-rating of memory chipmakers, which had historically been viewed as cyclical businesses tied to consumer electronics demand. The rise of AI has begun to reshape that perception, positioning high-performance memory as a structural growth segment tied to data center expansion and advanced computing.
What is emerging is a supply chain under sustained pressure. Hyperscale cloud providers and AI developers are competing for limited chip supply, locking in orders years in advance. That dynamic is shifting pricing power toward manufacturers like SK Hynix, at least in the near to medium term.
The key question is duration. While new capacity is being built, the long lead times involved mean that shortages, particularly in specialized segments like HBM, are likely to persist.
