Podcast Episode
The root cause is structural. Manufacturing high-bandwidth memory chips for AI applications consumes wafer capacity at three to four times the rate of standard memory. With the world's biggest technology firms on track to spend an estimated six hundred and fifty billion dollars on data centre infrastructure this year alone, chipmakers have been forced to divert production toward lucrative AI server memory at the expense of conventional DRAM for smartphones, PCs, and vehicles.
SK Hynix, the dominant supplier of high-bandwidth memory to Nvidia with roughly fifty-seven percent market share, reported that its DRAM and NAND inventory covers just four weeks of supply. Data centres are now consuming around seventy percent of all memory chips produced globally, up from thirty-two percent five years ago.
New DRAM fabrication plants require three to five years to build, and major producers have already sold forward capacity through twenty twenty-seven and twenty twenty-eight. For hardware makers and consumers alike, the era of constrained supply and elevated prices shows no sign of easing.
Global Chip Wafer Shortage Set to Last Until 2030 as AI Devours Memory Supply
March 17, 2026
0:00
3:55
SK Group chairman Chey Tae-won warned that the global semiconductor wafer shortage will persist until at least 2030, with a projected shortfall exceeding twenty percent. The crisis, driven by insatiable AI demand for high-bandwidth memory, is already pushing DRAM prices up sharply and forcing consumer device makers to cut production forecasts.
AI Hunger Creates a Wafer Famine
The chairman of South Korea's SK Group has delivered a stark warning to the semiconductor industry: the global chip wafer shortage is not going away any time soon. Speaking at Nvidia's GTC twenty twenty-six conference in San Jose, Chey Tae-won said the shortfall could persist until twenty thirty, far beyond earlier industry estimates that pointed to twenty twenty-seven as a turning point.The root cause is structural. Manufacturing high-bandwidth memory chips for AI applications consumes wafer capacity at three to four times the rate of standard memory. With the world's biggest technology firms on track to spend an estimated six hundred and fifty billion dollars on data centre infrastructure this year alone, chipmakers have been forced to divert production toward lucrative AI server memory at the expense of conventional DRAM for smartphones, PCs, and vehicles.
Consumer Devices Caught in the Crossfire
The consequences are already being felt across the consumer electronics industry. DRAM prices have surged between eighty and ninety percent in recent quarters, with some DDR4 kits doubling in price since late twenty twenty-five. Research firm IDC now forecasts global PC sales could shrink by nearly five percent this year, while smartphone sales are expected to decline by at least two percent as manufacturers struggle to secure affordable memory.SK Hynix, the dominant supplier of high-bandwidth memory to Nvidia with roughly fifty-seven percent market share, reported that its DRAM and NAND inventory covers just four weeks of supply. Data centres are now consuming around seventy percent of all memory chips produced globally, up from thirty-two percent five years ago.
Strategic Moves and Market Positioning
Alongside the shortage warning, Chey revealed that SK Hynix is reviewing a potential listing of American Depositary Receipts on US exchanges, a move that could help narrow the valuation gap with American peers like Micron Technology. The company is also showcasing its latest AI memory technologies at GTC, including sixth-generation HBM4 and HBM3E products.New DRAM fabrication plants require three to five years to build, and major producers have already sold forward capacity through twenty twenty-seven and twenty twenty-eight. For hardware makers and consumers alike, the era of constrained supply and elevated prices shows no sign of easing.
Published March 17, 2026 at 11:29pm
