Micron (MU) shares are down 30% from the 52-week high despite “blowout” results last month, with the Roundhill Memory ETF also down 23% over the past month, suggesting rotation out of memory. The article argues AI-driven DRAM/NAND demand plus a structural supply crunch (peers expect worsening through 2027 and demand exceeding supply beyond 2030) should keep pricing power intact. Analysts project EPS to jump 785% to $73.37 in fiscal 2026, with further gains to $198.26 per share in ~4 years, implying substantial upside even under a discounted valuation assumption.
MU’s setup is driven less by today’s fundamentals than by positioning: memory is the cleanest way to express AI enthusiasm, so when semis de-risk, this name tends to get hit harder than the broader complex. That creates a tradable dislocation, but only if pricing remains tight; if contract DRAM/NAND trends flatten, the market will discount the next cyclical roll-over before it appears in reported EPS.
Second-order, higher memory prices are a tax on the AI capex stack. MU captures the upside, but server OEMs, module assemblers, and hyperscalers absorb the cost or delay deployments, which can pressure weaker suppliers and slow AI unit growth at the margin. NVDA is not the direct loser, but memory inflation can raise system bill-of-materials enough to delay some server purchases, especially outside the very top-tier cloud customers.
Over 6-18 months, the key risk is supply response: if SK Hynix/Samsung et al. add capacity faster than expected, the market can move from scarcity to normalization while earnings still rise. The consensus is likely underestimating reflexivity in memory: the better the pricing environment, the more aggressively competitors will reinvest, which is why the long-term multiple may deserve to stay below the broader index even in an AI-driven upcycle.
AI-powered research, real-time alerts, and portfolio analytics for institutional investors.
Overall Sentiment
mildly positive
Sentiment Score
0.20
Ticker Sentiment
