Samsung Electronics and SK Hynix have officially notified major clients of a substantial price increase for server DRAM products in Q1 2026, with the hike generally ranging between 60% and 70%.
Confirmed by multiple industry sources, the two South Korean memory giants—Samsung Electronics and SK Hynix—have formally proposed a sharp price rise for server DRAM products in the first quarter of 2026, with the increase rate typically falling within the range of 60% to 70%. This adjustment not only sets a new high for single-quarter memory chip price increases in recent years, but also signals that AI-driven computing infrastructure construction is profoundly reshaping the entire supply chain landscape.
This price hike is not an isolated incident, but rather the result of multiple converging factors. On one hand, the explosive growth of generative AI has continuously driven up demand for high-performance computing hardware. AI acceleration platforms represented by NVIDIA H200 and AMD MI300X have raised higher requirements for memory bandwidth and capacity, which has not only spurred a surge in orders for High Bandwidth Memory (HBM), but also indirectly boosted the consumption of traditional server DDR4/DDR5 DRAM.
Data shows that a single AI server requires 8 times the DRAM of a regular server. Meanwhile, the total investment in AI infrastructure by the four major North American cloud providers is expected to exceed 600 billion US dollars in 2026, further widening the demand gap for general-purpose server DRAM. In response, leading manufacturers such as Samsung and SK Hynix have shifted their wafer production capacity to higher-margin high-bandwidth memory products like HBM3E. These high-end products, designed for AI accelerators such as NVIDIA's, consume three times the wafer resources of traditional DRAM, directly squeezing the production capacity for general-purpose server DRAM. With new production capacity taking 1.5 to 2 years to come online, it will be difficult to fill the gap in the short term.
On the other hand, after the United States relaxed some restrictions on exports of high-end AI chips to China at the end of 2025, Chinese cloud service providers and large tech enterprises promptly launched large-scale procurement plans, further exacerbating the tight global DRAM production capacity situation.
Faced with a supply shortage in the market, Samsung and SK Hynix have adopted an extremely aggressive supply strategy. The two companies have explicitly refused to sign 2- to 3-year long-term supply agreements, insisting on quarterly price negotiations and contracts to lock in the quarterly price increase rhythm. Sources indicate that while international cloud providers including Google, Microsoft, and Amazon have expressed concerns about rising costs, most have accepted this round of price hikes due to the urgency of deploying AI infrastructure. There are even rumors that some clients who failed to lock in production capacity in advance are facing internal accountability over delivery risks, highlighting the strategic value of memory resources at present.
The enthusiastic market reaction has also confirmed the industry's recognition of the price hike logic. On January 5, the day the news broke, Samsung Electronics' stock price soared nearly 7.5% to a record high, while SK Hynix's stock price rose nearly 3% in tandem. This directly drove South Korea's KOSPI to close up 3.43%, hitting a new closing high, and the global semiconductor sector also strengthened in response.
Notably, the shift in production capacity allocation has also intensified structural shortages. To meet the manufacturing demands of advanced products such as HBM3E, the two South Korean manufacturers have redirected a large portion of their 1α nm and more advanced process capacity to high-margin product lines, leading to constrained output of general-purpose server DRAM. This "crowding-out effect" has made the already recovering server memory market even worse off.
Market research institutions predict that the average annual price of server DRAM in 2026 may surge by more than 140% year-on-year, with Q1 set to be a key inflection point. In terms of supply and demand, global DRAM bit supply is expected to grow by only 15% to 20% in 2026, while demand growth will reach 20% to 25%, making the widening supply-demand gap irreversible.
Affected by this, downstream terminal manufacturers are facing mounting cost pressures. The proportion of memory costs in consumer electronic products such as PCs and smartphones has risen significantly, and some brands have begun to consider price increases or adjust product configuration strategies. Meanwhile, the operating profits of Samsung and SK Hynix are expected to reach 155 trillion KRW and 148 trillion KRW respectively in 2026, surging more than 2.5 times compared with 2025. Their gross profit margins are expected to exceed 60%, surpassing TSMC for the first time in seven years.
This memory price surge ignited by AI not only reflects the deep dependence of technological evolution on hardware infrastructure, but also reveals the fragility and gaming dynamics of the global semiconductor supply chain in the face of highly certain demand. In the coming quarters, the price trends, production capacity allocation, and client response strategies in the DRAM market will continue to serve as important indicators for observing the competitive landscape of the global technology industry.