How Micron reversed the global tech selloff, exposing AI's 'memory tax' - Fortune

Micron Technology delivered a stunning quarterly performance that not only exceeded Wall Street expectations but also highlighted a fundamental shift in how memory chips are valued in the AI era. With... Read original →

Micron’s results underscore a structural shift: memory is no longer a commodity but the AI stack’s performance bottleneck. Technically, surging HBM demand accelerates adoption of EUV and 3nm-class cleanroom processes, cannibalizing NAND and legacy DRAM capacity—jeopardizing edge and automotive supply. From a compliance view, five-year take-or-pay SCAs stabilize revenue but inflate hyperscalers’ ‘memory tax,’ inviting antitrust scrutiny amid growing regulatory wariness of Big Tech capex. Competitively, Samsung and SK Hynix must mimic long-term contracts, yet lag Micron by 6–9 months in HBM3E yields and TSV stacking; NVIDIA and AMD are countering with custom memory interfaces. Over the next 18 months, capex will pivot toward storage, blurring foundry-IDM boundaries, while geopolitical constraints ensure that neither Taiwan, China nor mainland China can resolve the advanced HBM supply gap.