The global RAM shortage could persist until 2030, according to SK Group chairman Chey Tae-won, who issued a stark warning at Nvidia’s GTC conference in San Jose on March 17, 2026. The memory crisis is no longer a near-term pinch—it is a structural problem that will reshape hardware pricing and availability for years.
Key Takeaways
- Global RAM shortage expected to last until 2030 with over 20% industry-wide supply shortfall
- Securing additional wafers requires 4-5 years of construction due to capacity, power, and water constraints
- Server memory chip prices surged 60-76% in Q4 2025 and will continue rising
- SK Hynix controls 57% of the HBM market and 32% of overall DRAM supply
- PC and smartphone shipments projected to fall 10.4% and 8.4% respectively in 2026 due to memory price inflation
How severe is the global RAM shortage?
The global RAM shortage is severe and structural, not cyclical. Chey stated that the memory industry expects more than a 20% shortage of wafers persisting through 2030, driven by AI data center demand that has outpaced manufacturing capacity. This is not a temporary bottleneck—it reflects a fundamental mismatch between what the market needs and what fabs can produce. Previous statements from SK Group leadership in Washington, D.C. last month pegged the current AI memory shortage at exceeding 30%, meaning demand has already overshot supply by nearly a third.
The math is grim. Even if manufacturers sprint to expand capacity, the physical constraints are immovable. Building a new wafer fab takes 4 to 5 years minimum, and that timeline assumes no delays in securing construction capacity, electrical power, water resources, or other critical inputs. SK Hynix, which controls 57% of the global HBM market and 32% of overall DRAM supply, is racing to build a $13 billion HBM packaging and testing facility at its Cheongju complex in South Korea, with construction beginning in April 2026 and completion targeted for the end of 2027. Yet even this massive investment will not fully close the gap by 2030.
What is driving memory chip prices higher?
AI demand is the primary culprit. Data centers are consuming HBM and DRAM at rates that outstrip production, and the shortage is translating directly into price inflation. Server memory chip prices surged 60 to 76 percent in the fourth quarter of 2025 and are expected to rise further in the first quarter of 2026, according to Counterpoint Research. Looking further ahead, Gartner projects that DRAM and SSD prices will climb 130 percent by the end of 2026 compared to 2025 levels.
This is not merely a data center problem—it cascades into consumer hardware. PC shipments are forecast to fall 10.4 percent in 2026 versus 2025, while smartphone shipments are expected to drop 8.4 percent, both driven partly by the 130 percent surge in memory and storage costs. Manufacturers face a brutal choice: absorb the cost and shrink margins, or pass the price to consumers and risk demand destruction. Many will do both.
When will memory prices stabilize?
SK Hynix CEO Kwak Noh-jung is expected to announce DRAM price stabilization measures soon, though the specifics remain undisclosed. However, any relief is likely to be temporary and partial. The structural shortage persists through 2030, so stabilization measures will probably focus on smoothing volatility rather than returning prices to pre-AI levels. Expect elevated memory costs as the new normal, not a return to the pricing regimes of 2024 and earlier.
SK Hynix is also reviewing a potential American Depository Receipt listing in the US, though the company has no plans to build a manufacturing plant in the United States. Instead, SK Hynix is focusing factory expansion on South Korea, where existing infrastructure allows faster deployment than greenfield sites, which would require 5 to 7 years regardless of location. This strategic choice underscores the reality: even with unlimited capital, physics and geography impose hard limits on how quickly supply can scale.
What about competing memory suppliers?
SK Hynix is not alone in facing capacity constraints, but its dominance in HBM—the memory type most critical for AI—gives it outsized influence over the shortage narrative. The company is planning to ship final HBM4 samples to Nvidia to maintain its AI memory leadership. Competitors like Micron and Samsung are also expanding capacity, but none can match SK Hynix’s current market share or the speed of its expansion timeline. The shortage will likely persist across the entire industry, not just at one supplier.
Will the shortage extend beyond 2030?
Chey’s 2030 forecast is a best-case scenario assuming no major geopolitical disruptions, no construction delays, and successful execution of all announced fab expansion projects. Delays are common in semiconductor manufacturing—power outages, water shortages, supply chain hiccups, or regulatory changes could push the shortage further into the 2030s. The chairman is essentially saying: if everything goes right, relief arrives around 2030. Betting on everything going right is unwise.
How will the RAM shortage affect hardware makers?
Manufacturers of PCs, smartphones, servers, and appliances will face sustained margin pressure. Either they absorb higher memory costs—crushing profitability—or they raise prices and risk losing customers to competitors who absorb costs temporarily. Some smaller manufacturers may exit the market entirely. The shortage is a structural reshuffling that will consolidate the hardware industry around players with deep pockets and pricing power.
FAQ
Is the global RAM shortage affecting consumer devices right now?
Yes. Server memory prices have already surged 60-76 percent in Q4 2025 and are rising further, and consumer-facing devices like PCs and smartphones are experiencing reduced shipments due to memory cost inflation. The shortage is not hypothetical—it is already reshaping the market.
Can SK Hynix solve the shortage on its own?
No. Even with SK Hynix’s $13 billion facility coming online by end-2027, the company cannot single-handedly close a 20 percent industry-wide supply gap. The shortage is structural and will require capacity increases across the entire semiconductor ecosystem.
What does the 2030 timeline mean for AI adoption?
AI data center buildout will continue, but at higher cost. Companies will optimize memory usage, shift workloads, or accept slower scaling. The shortage will not halt AI progress, but it will make it more expensive and selective, favoring well-capitalized players over startups and smaller enterprises.
The global RAM shortage is not a crisis that will resolve in quarters—it is a multi-year structural challenge that will reshape hardware pricing, manufacturer profitability, and AI infrastructure economics through the end of the decade. SK Group’s warning is bleak because the math is bleak. Demand has outpaced supply by design constraints that cannot be overcome quickly, and the gap will persist until 2030 at minimum.
Edited by the All Things Geek team.
Source: TechRadar
