The Memory Shortage That Isn't One

The DRAM market is not running short. The world's three memory makers are routing output to whoever pays the highest margin — in what Chang describes as a deliberate choice to prioritize AI and cloud customers.

Samsung, SK Hynix, and Micron are directing the lion's share of their DRAM output toward high-bandwidth memory and server products sold directly to AI and cloud customers, where the margins justify the engineering. The independent module makers, the companies that turn wafers into the DIMMs, SSDs, and embedded storage that show up in consumer hardware, get whatever is left.

Apacer's CEO C.K. Chang made that arithmetic visible at the company's July 24 investor conference. The projected 30%-of-2026 figure is not a total industry number. It is an allocation number, the share that independent module makers can expect to receive in 2027. His company's response reads less like a business plan and more like a survival plan: inventory up roughly 48% in a single quarter to about $383 million, and a five-year syndicated loan of up to about $124 million arranged specifically to buy more wafers whenever the big three release them.

When a constrained commodity has two customer tiers, capacity flows to the higher-margin tier first, and downstream buyers learn to finance survival rather than negotiate price. TrendForce and IDC both expect tightness through at least mid-2027. That is not a chip shortage. It is a margin allocation, and the consumer hardware market is what gets routed around.