Micron Just Crossed $1,000 a Share. Here's the Math on Where It Goes Next.

Micron’s rally is being driven by surging AI-related demand for HBM and DRAM, with management saying its entire 2026 production capacity is sold out and DRAM shortages may persist beyond 2027. TrendForce expects the global memory market to reach $1.3T in 2027, with DRAM revenue rising 303% this year to $619B and then to $903B in 2027. The article argues Micron’s forward P/E of 9.5 remains attractive versus AI chip leaders, implying room for further upside despite cyclicality risks.

Analysis

The market is starting to re-rate memory from a cyclical input to a scarcity-priced AI substrate, and that matters more for second-order winners than for the headline stock. If HBM remains sold out while DRAM stays tight through 2027, the real transmission is margin discipline across the entire server stack: GPU vendors, networking, and OEMs will face higher bill-of-materials costs, but hyperscalers will likely absorb it because memory is a smaller share of total AI server economics than compute. That makes the supply chain less about outright demand destruction and more about who has contractual visibility and who is still exposed to spot pricing.

The risk is that the market is extrapolating a one-way supply deficit into a quasi-permanent regime before the capacity cycle has had time to respond. Memory history says pricing power is most dangerous when it feels most durable; the next inflection is not demand collapsing, but capex getting funded by the current windfall and new supply arriving with a lag of 12-24 months. If hyperscalers slow incremental builds even modestly, the earnings denominator can shrink fast because this business still has extreme operating leverage.

Consensus is also underestimating how valuation dispersion could broaden inside AI semis. If memory rerates toward infrastructure-like multiples, the relative multiple gap versus compute and networking leaders may compress, not because those names deserve lower prices, but because MU’s forward earnings are still more cyclical than the market is admitting. The cleaner expression of the thesis is not simply owning MU outright, but owning MU while fading the names most vulnerable to AI capex moderation or memory-cost pass-through friction.

The contrarian setup is that the best trade may be on the duration of the supercycle, not its existence. A multi-year sold-out book is bullish, but it also creates a textbook trap: the stock can keep going until the market starts pricing in normalized margins two years ahead. That makes the path-dependent opportunity high, but the window to add aggressively likely narrows once forward estimates stop rising at triple-digit rates.

Key Decisions for Investors

• Long MU on any 5-10% pullback over the next 2-4 weeks; target a 12-18 month horizon with upside driven by estimate revisions rather than multiple expansion alone. Use a trailing stop if forward EPS growth decelerates for two consecutive quarters.
• Pair trade: long MU / short a basket of lower-quality semiconductor equipment or memory-adjacent names with weaker contract visibility for the next 6-9 months. The thesis is that contracted supply and scarcity pricing accrue to the producer with the best balance sheet, while laggards face margin compression when supply normalizes.
• Buy MU Jan-2027 call spreads to express the supercycle with limited downside. Favor strikes that monetize continued earnings upgrades but cap premium paid if the market starts discounting 2028 oversupply risk.
• Trim or avoid chasing NVDA/AVGO on days of strong AI-memory headlines; if hyperscaler budgets remain fixed, higher memory content can become a tax on server economics and eventually compress ordering elasticity. Use them as relative shorts only against MU strength, not outright directional bearish bets.
• Set a catalyst watch for next hyperscaler capex prints over the next 1-2 quarters; if spend guidance inflects lower, reduce MU exposure quickly because the stock's multiple is now highly sensitive to any sign of demand digestion.