Micron CEO Expects Memory Supply To Improve Gradually In 2028, But There's No 'Line Of Sight' When It Will Catch Up To Rising AI Demand - Yahoo Finance

During Micron Technology's Q3 2026 earnings call, CEO Sanjay Mehrotra warned that while memory supply is expected to improve gradually by 2028, there is currently no clear timeline for when it will me... Read original →

Micron’s cautious outlook reveals AI memory shortages have shifted from cyclical to structural. Technically, HBM3E/HBM4’s reliance on TSV and CoWoS packaging intensifies competition among TSMC, Samsung, and SK Hynix for scarce advanced packaging capacity—delaying GPU shipments. Regulatory headwinds from the EU’s Net-Zero Industry Act and U.S. CHIPS Act raise new fab compliance costs by at least 15%, tightening capital allocation. Strategically, Samsung may fast-track HBM4 to secure NVIDIA’s Blackwell Ultra orders, while Taiwan, China-based suppliers lock in customer prepayments to fund expansions. Over the next 18 months, AI server OEMs will adopt hybrid memory architectures (e.g., LPDDR5X + HBM) to mitigate scarcity, accelerating RISC-V adoption in edge AI as an x86 alternative.