Is Micron Stock Expensive at $725? Here's the Math on Why Shares Are Still Dirt Cheap.

Micron shares have surged from roughly $448 to $804 in the past month, driven by AI memory demand and a fiscal Q2 revenue jump to $23.9 billion, nearly triple year over year. EPS rose to $12.07 from $1.41, while Wall Street sees full-year EPS around $58 and roughly $100 by fiscal 2027. Despite the sharp run-up, the article argues Micron still screens attractively at about 8x forward earnings.

Analysis

The market is treating MU less like a cyclical memory supplier and more like a scarce infrastructure layer for AI compute. That rerating can persist because the earnings inflection is being driven by mix, not just volume: high-bandwidth and advanced memory content per accelerator creates a more durable pricing umbrella than a normal DRAM cycle. The second-order implication is that the entire AI stack becomes more memory-constrained, which should support upstream suppliers while pressuring hyperscalers’ capex efficiency over the next 2-4 quarters.

The key risk is that the current multiple assumes the memory supercycle extends cleanly into FY27 without a supply response. Historically, once pricing visibility improves, capacity additions and inventory rebuilding can turn an earnings air pocket into a sentiment reversal within 2-3 quarters, especially if AI capex pauses or customers begin second-sourcing aggressively. In other words, the stock is not vulnerable because it is "expensive" on today’s forward earnings; it is vulnerable if forward earnings estimates stop moving up.

Consensus is likely underestimating how crowded the "AI memory scarcity" trade has become. The marginal buyer is probably not fundamental; it is momentum and quant flows chasing upward revisions, which can amplify upside in the near term but also create a sharp downside if one guide disappoints. On the other side, NVDA benefits from the same memory intensity, but sustained memory inflation can eventually compress accelerator ROI and slow enterprise adoption, making this less purely bullish for the whole AI complex than the market is pricing.

Best risk/reward is to stay long MU into positive estimate revisions but hedge with a basket or pair that benefits if AI capex broadens without memory margin expansion. The setup looks strongest over days-to-weeks on continuation, but the cleaner expression over months is to own MU against a less levered AI beneficiary or against a maker of the end-demand hardware if memory pricing outruns final system economics.

Key Decisions for Investors

• Maintain a tactical long MU position for 2-6 weeks while revisions momentum is accelerating; use a trailing stop if the stock loses the post-gap breakout zone, since flow-driven names can de-rate fast once upside estimates flatten.
• Pair trade: long MU / short a lower-beta AI beneficiary basket over 1-3 months to isolate memory pricing upside; this is attractive if investors keep paying up for scarcity while the rest of the AI stack reverts to average multiples.
• Buy MU call spreads 2-4 months out instead of outright equity to capture continued multiple expansion with defined downside; the thesis works best if another guide-up lands before supply concerns surface.
• For conservative exposure, trim NVDA into strength if memory-cost inflation starts outrunning accelerator ASP leverage; if memory becomes the bottleneck, NVDA can still rise, but upside may narrow as end customers push back on total system ROI.
• Set a review trigger on the next two quarterly prints: if gross margin guidance stops improving or inventory days rise, reduce MU exposure quickly, since the trade depends more on estimate revisions than on the headline valuation multiple.