The memory chip shortage will persist until 2030, according to SK Group Chairman Chey Tae-won, who warned that global wafer supply continues to trail demand by more than 20%. Speaking at Nvidia’s GTC conference in San Jose on March 17, 2026, Tae-won predicted the shortage will last another four to five years, extending well beyond previous industry forecasts.
Key Takeaways
- Memory chip shortage expected to continue until 2030, lasting 4-5 more years from March 2026.
- Wafer supply trails demand by more than 20%, according to SK Group leadership.
- AI-driven demand for high-bandwidth memory (HBM) is the primary driver of the shortage.
- Server memory prices surged 60-76% in Q4 2025 due to supply constraints.
- SK Hynix, Samsung, and Micron control the majority of global memory chip supply.
Why the Memory Chip Shortage Persists Until 2030
The memory chip shortage remains entrenched because AI demand for high-bandwidth memory has overwhelmed manufacturing capacity. Wafer supply currently lags demand by more than 20%, creating a structural imbalance that cannot be resolved quickly. Building new fabrication plants and ramping production takes years, not months, which is why industry leaders expect the shortage to extend through the end of the decade.
SK Group’s forecast contradicts earlier predictions that supply would stabilize in 2025 or early 2026. Instead, the company’s chairman argued that demand growth—particularly from AI applications—continues to outpace production increases. This gap between supply and demand has already driven severe price inflation: server memory chip prices surged 60-76% in the fourth quarter of 2025 alone. That trajectory suggests continued upward pressure on memory costs throughout 2026 and beyond.
AI Demand Driving the Memory Chip Shortage
High-bandwidth memory (HBM) chips, essential for AI accelerators and large language models, have become the primary constraint in the semiconductor supply chain. Data centers racing to deploy AI infrastructure require massive quantities of HBM, which only a handful of suppliers can produce at scale. SK Hynix, Samsung, and Micron control the vast majority of global memory chip supply, yet even their combined output cannot keep pace with AI demand.
This concentration of supply among three manufacturers creates a bottleneck. Unlike commodity DRAM or NAND flash, which multiple companies produce, HBM remains a specialized product with limited manufacturing footprint. Expanding HBM capacity requires significant capital investment and time. The shortage will therefore persist as long as AI adoption drives memory demand faster than fabs can increase output.
Market Impact and Pricing Outlook
The extended memory chip shortage has immediate consequences for data center operators, cloud providers, and enterprises building AI infrastructure. Higher memory costs translate directly into higher server costs, which ripple through cloud pricing and AI service expenses. With prices already up 60-76% in Q4 2025, further increases are likely if supply constraints worsen or demand accelerates.
For consumers, the impact is less direct but still real. GPU and AI accelerator prices remain elevated partly due to memory component costs. Servers using the latest processors command premium prices because memory components are scarce and expensive. This dynamic will likely persist through 2027 and 2028, potentially delaying broader AI adoption in cost-sensitive segments of the market.
Can New Manufacturing Capacity Solve the Shortage?
Expanding memory chip production requires both capital and time. New wafer fabs take three to five years to build and qualify, meaning any facility announced today would not contribute meaningfully to the shortage until 2029 or later. Existing manufacturers are investing in capacity, but their expansion timelines align with SK Group’s 2030 forecast—not earlier relief.
The memory chip shortage will therefore remain a structural constraint on AI infrastructure growth through the end of the decade. Only after 2030, when new capacity comes online and demand growth moderates, should the market rebalance toward equilibrium.
How long will the memory chip shortage last?
SK Group Chairman Chey Tae-won predicts the memory chip shortage will persist until 2030, lasting approximately four to five more years from his March 2026 statement at Nvidia’s GTC conference. This timeline assumes continued strong AI demand and no major disruptions to manufacturing capacity.
Why are memory chip prices so high right now?
Server memory chip prices surged 60-76% in the fourth quarter of 2025 due to constrained supply and surging AI demand. Wafer supply trails demand by more than 20%, creating an acute shortage that drives prices upward.
Which companies control memory chip supply?
SK Hynix, Samsung, and Micron control the majority of global memory chip supply. Their limited combined capacity relative to AI demand has created the current shortage and will likely sustain it through 2030.
The memory chip shortage will reshape the technology industry for the next four years. Higher costs for memory will slow AI adoption in price-sensitive markets, delay enterprise infrastructure upgrades, and force cloud providers to optimize workloads more aggressively. For manufacturers and investors, the extended shortage signals sustained pricing power through 2029, but also the urgency of building new capacity before the market shifts. By 2030, the shortage should finally ease—but only if no major disruptions occur and manufacturing expansion proceeds on schedule.
This article was written with AI assistance and editorially reviewed.
Source: Tom's Hardware
