Macronix halts NOR expansion, phases out ROM, bets NT$30B expansion on eMMC NAND

Macronix's pivot is survival, not strategy. NOR Flash growth has flatlined into single digits, and game ROM is a dead category displaced by eMMC and UFS. Freezing expansion and exiting ROM signals management concedes the legacy stack can no longer sustain ROIC. The critical technical risk is process discontinuity. 3D NAND's charge-trap stacking shares virtually zero process overlap with planar NOR floating-gate structures. Two decades of planar know-how depreciates by over 60% in the 3D domain. Of the NT$30B capex, a substantial share will be consumed by equipment qualification and yield ramp rather than product iteration. With output not ramping until H2 2027, 2025–2026 becomes a pure cash-burn window that will stress the balance sheet. The competitive landscape is unforgiving. Samsung, Micron, Kioxia, and SanDisk control over 95% of eMMC, with 3D layer counts past 300. Macronix enters as a latecomer with no scale to amortize depreciation and no advanced-node yield moat. The contract price increases for specific customers almost certainly refer to locked-in long-term agreements with local OSATs in Taiwan, China or industrial and automotive accounts—not global consumer pricing power. The 12–24 month tail effect: NOR pricing stays elevated as Winbond, GigaDevice, and others absorb freed capacity. On the NAND/eMMC side, if 2027 new supply collides with a downcycle, the capacity-as-liability scenario from the 2019 NAND price war will repeat. History does not forgive late entrants.