AI-Driven Memory Shortage Could Last 4 to 5 More Years as HBM Demand Tightens Wafer Supply

Key takeaways box

• Chey Tae-won warned the shortage could last another four to five years, saying wafer supply is still more than 20% below demand as AI pushes up HBM requirements.
• Micron expects tight market conditions to persist beyond calendar 2026 and is already discussing multiyear contracts with customers.
• Samsung said memory demand remained strong despite limited supply availability, while SK hynix said demand for conventional server memory also increased sharply alongside HBM.
• IDC, TrendForce, and Gartner all point to downstream pressure on pricing, PCs, smartphones, and broader electronics availability as memory supply stays constrained
  
  

The global memory market is moving into a tighter and potentially longer shortage cycle as AI demand for high-bandwidth memory, or HBM, absorbs more wafer and packaging capacity. At Nvidia GTC 2026, SK Group Chairman Chey Tae-won said wafer supply remains more than 20% below demand and could take four to five years to catch up, warning the shortage could persist until 2030.

This direction is reinforced across the industry, with suppliers and analysts pointing to prolonged tightness and slower supply recovery.

Why SK Group Warns the Memory Shortage Could Last Years

Chey’s warning reflects the shift in the memory market from a routine cyclical rebound to a capacity problem. Speaking on the sidelines of Nvidia GTC in San Jose, he said the shortfall is rooted in wafer supply, not just demand strength, and that the industry may need at least four to five years to add enough capacity.

The industry cannot quickly manufacture its way out of the imbalance, especially when the fastest-growing part of the market is also the most capacity-intensive.

HBM Demand Is Tightening Wafer Supply

At the center of the squeeze is HBM, which has become critical to AI accelerators and advanced server platforms. SK hynix, one of Nvidia’s key HBM suppliers, said 2025 demand was driven by AI memory and other high-value products, and that large-scale HBM4 production is already underway.

HBM consumes more advanced manufacturing and packaging resources than conventional memory, which means the mix shift matters as much as total demand. As suppliers prioritize HBM because of both strategic importance and higher margins, more wafer starts and advanced packaging slots are pulled toward AI-linked products.

Conventional DRAM and NAND Are Also Feeling the Pressure

As more capacity is directed toward HBM, less flexibility remains for mainstream DRAM and NAND, increasing the risk of tighter allocations, longer lead times, and firmer pricing across the broader electronics market.

Micron said fiscal Q2 2026 results were driven by a “strong demand environment” and “tight industry supply,” while its earnings materials said it expects tight conditions in DRAM and NAND to persist beyond calendar 2026.

TrendForce said supply for notebook DRAM and NAND has tightened significantly since the start of 2026 and warned that a mainstream notebook’s retail price could rise by nearly 40% if memory and CPU costs keep climbing.

Samsung is signaling the same direction from a different angle. In its FY2025 results, the company said its memory business delivered record highs despite limited supply availability. Separately, Reuters reported that co-CEO Jun Young-hyun said Samsung is working with major customers on three-to-five-year contracts and described the current cycle as an “unprecedented supercycle.”

SK hynix’s own expansion plans also point to a multi-year response. The company said demand for conventional server memory increased sharply alongside HBM and outlined capacity additions tied to the M15X fab in Cheongju, the Yongin Semiconductor Cluster, and new advanced packaging facilities.

Taken together, those signals matter because they come from the companies closest to the memory supply chain. None of them are describing a market that looks ready to ease quickly.

Analysts See Rising Memory Prices and Broader Device Impact

Independent researchers are also pointing to a broader downstream effect. IDC said in December that the global memory shortage could persist well into 2027 as AI data-center demand continues to outstrip supply.

Gartner said soaring memory costs are projected to cut global PC shipments by 10.4% and smartphone shipments by 8.4% in 2026, showing that rising DRAM and SSD costs are already affecting end-market demand.

TrendForce added that memory inflation is now large enough to reshape notebook bill-of-material economics, underscoring that this is no longer just a hyperscaler or AI-server issue. It is increasingly a broader electronics procurement issue.

Memory Is Becoming a Planning Constraint for Buyers

Memory is no longer just a volatile line item that can be managed with short-term buying or by waiting for the next downcycle. As supply tightens and more capacity is locked into higher-value applications, memory is becoming a planning constraint across servers, notebooks, smartphones, industrial systems, and embedded designs.

That means the old assumption that pricing will normalize quickly may no longer hold. If suppliers keep prioritizing HBM, and if more customers move toward longer contracts, the spot market may offer less relief than in prior cycles.

Procurement teams should be preparing for a market defined by tighter allocations, narrower flexibility, and longer visibility longer visibility windows and earlier planning around lead-time risk before shortages show up in production schedules.

Need better visibility into DRAM, NAND, and memory availability?

Use the IBS BOM Tool or request a quote to compare sourcing options, secure supply faster, and reduce lead-time risk before constraints tighten further.