Global memory chip inventories at Samsung Electronics and SK hynix have plummeted to critical lows of less than ten days of operating supply, driven by an unprecedented surge in artificial intelligence infrastructure demand, according to data released by KB Securities and reported by Business Korea.
The severe inventory drain creates an acute bottleneck across the semiconductor market. According to KB Securities head of research Kim Dong-won, the depletion primarily affects crucial components like DDR5 and enterprise solid-state drives (SSDs). Manufacturers are aggressively reallocating their fabrication lines to satisfy heavy orders for high-bandwidth memory required by data centers powering AI models.
Production Shifts and the HBM4 Crunch Ahead
The manufacturing crunch stems directly from a structural shift toward complex high-bandwidth memory architectures. According to reporting from Business Korea, the production of high-speed memory utilizes a multi-layered DRAM configuration that requires significantly more silicon wafers and manufacturing capacity than traditional memory chips.
Analyst projections indicate the pressure will intensify as the industry transitions toward next-generation HBM4 production. According to Business Korea, global data center operators are projected to pour $1.3 trillion into AI infrastructure, representing a 60 percent year-over-year increase. Semiconductor storage components are expected to capture 57 percent of those total expenditures by 2027, up sharply from prior cycles and exacerbating the supply deficit across standard DRAM and NAND markets.
Downstream Pricing and Industry Impact
The memory shortage is already translating into higher equipment costs for end users. According to Wccftech reporting, hardware vendor NVIDIA has raised prices on AI servers by 15 percent or more to counter surging memory costs.
Nvidia’s corporate leadership has repeatedly warned that component scarcity will likely remain a limiting factor for company growth well into 2028.