A global shortage of DRAM memory chips is expected to stretch well into the late 2020s, with analysts projecting that manufacturers will meet only 60 percent of demand by the end of 2027 — a shortfall driven largely by the surging appetite for AI-optimised memory that is reshaping how the world's largest chipmakers allocate production capacity.
The world's three dominant memory manufacturers — Samsung, SK Hynix, and Micron — are all investing in new fabrication facilities, but virtually none of that additional capacity is expected to come online before 2027 at the earliest, according to reporting by Nikkei Asia. In the meantime, demand continues to outpace supply at a rate that industry observers describe as historically unusual.
SK Group chairman Chey Tae-won warned in March that chip wafer shortages could persist as late as 2030, a timeline that would make this one of the most prolonged semiconductor supply crises in recent memory. SK Hynix opened a new fabrication plant in Cheongju, South Korea, in February 2026, but that represents the only meaningful production increase among the three major players for the current year.
Analysts at Nikkei estimate that production would need to grow by 12 percent annually in both 2026 and 2027 to keep pace with rising demand. Research firm Counterpoint Research has noted that the shortage will have broad cost implications across industries, from consumer electronics to enterprise computing.
A key driver of the imbalance is the pivot toward high-bandwidth memory (HBM), a specialised form of DRAM engineered to support artificial intelligence workloads. Major chipmakers are prioritising HBM production to supply AI accelerator manufacturers — most notably Nvidia — at the expense of conventional DRAM used in laptops, servers, and smartphones. Because HBM is more complex and resource-intensive to produce, each unit of factory capacity yields fewer chips, further constraining overall supply.
For consumers, the consequences are likely to manifest as higher prices for devices containing memory chips, including personal computers, smartphones, and networking equipment. Enterprises building out data centre infrastructure may face longer procurement lead times and elevated costs for server memory modules.
Manufacturers are racing to expand capacity, but semiconductor fabrication plants take years and billions of dollars to construct and commission. Even with aggressive investment timelines, the structural gap between supply and demand is unlikely to close quickly, leaving buyers across the market exposed to elevated prices for the foreseeable future.
Analysis
Why This Matters
- Higher memory prices will filter through to consumers in the form of more expensive laptops, smartphones, and other devices, potentially delaying hardware upgrade cycles for individuals and businesses alike.
- Enterprises and cloud providers building AI infrastructure face constrained supply of the high-bandwidth memory that underpins modern AI accelerators, which could slow the pace of AI deployment and raise operational costs.
- The shortage highlights a structural vulnerability in global semiconductor supply chains, where a small number of manufacturers control the vast majority of production capacity for a critical component.
Background
The global semiconductor industry has experienced cyclical shortages before — most notably the broad chip shortage of 2020–2023 triggered by pandemic-era demand spikes and supply chain disruptions. That crisis affected everything from automobiles to gaming consoles and prompted governments in the United States, Europe, and Asia to invest heavily in domestic chip manufacturing through legislation such as the US CHIPS and Science Act.
However, the current memory shortage has a distinct character. Rather than being caused primarily by supply chain disruption, it is being driven by a structural shift in what memory chipmakers are producing. The explosive growth of large language models and AI training workloads has created enormous demand for high-bandwidth memory, which commands higher margins and has become the strategic priority for Samsung, SK Hynix, and Micron. This has effectively crowded out conventional DRAM production.
Semiconductor fabrication plants, known as fabs, require years of planning, construction, and qualification before they can produce chips at volume. Investments made today in response to the current shortage will not yield meaningful output until 2027 or 2028, meaning the industry has limited tools to rapidly rebalance supply in the short term.
Key Perspectives
Memory manufacturers (Samsung, SK Hynix, Micron): The three companies are investing heavily in new capacity and argue that the shift toward HBM reflects rational responses to market signals — AI memory is more profitable and in higher demand. They maintain that new fabs will eventually close the gap, though timelines remain uncertain.
AI infrastructure buyers (cloud providers, Nvidia partners): These companies are competing aggressively for limited HBM supply and have generally accepted higher prices as a cost of building out AI capabilities. Their demand is the primary force reshaping memory market dynamics.
Critics and consumer advocates: Some analysts warn that the concentration of memory production among just three global players — and the further concentration of that output toward a single use case — leaves the broader technology ecosystem exposed. Consumer electronics makers and enterprise IT buyers have little leverage to push back against price increases when supply is this constrained.
What to Watch
- Monthly DRAM spot prices, which will serve as the clearest early indicator of whether supply is beginning to catch up with demand.
- Commissioning timelines for new fabs announced by Samsung, SK Hynix, and Micron — any delays will extend the shortage further.
- Shifts in AI hardware demand, particularly if a slowdown in data centre investment reduces appetite for HBM and frees up conventional DRAM capacity.