Micron, SanDisk get a number Wall Street rarely writes

Susquehanna raised its Micron price target to $1,750 from $600 and its SanDisk target to $3,250 from $2,000, both new Wall Street highs, on expectations that AI-driven memory demand will keep DRAM and NAND supply tight through 2027. The note cites second-quarter DRAM pricing up 50% to 60% QoQ and NAND up 75% to 100%, with memory revenue forecast by TrendForce to jump 134% to $552 billion in 2026 and another 53% to $843 billion in 2027. The article argues that AI infrastructure spending is fundamentally re-rating memory chips from cyclical commodities to core AI picks-and-shovels.

Analysis

The important shift is not that memory is strong; it is that the market is beginning to treat memory as a gated utility rather than a free-floating commodity. That matters because pricing power is no longer tied to broad handset/PC demand, but to a structurally inelastic AI buildout where HBM, server DRAM, and enterprise NAND all compete for the same constrained wafer base. The second-order winner is not just SNDK/MU, but any supplier with clean exposure to data-center storage and any OEM that locked supply early; the loser is the marginal server buyer that now has to ration deployments or accept longer lead times.

The setup is still earlier than the earnings revisions imply. The real catalyst is not the price target, but whether July/August supply commentary confirms that Q2 price moves are feeding through into Q3 contract resets without customers churning away. If the 50%-100% ASP assumptions prove even directionally correct, margin expansion will outpace sell-side models for multiple quarters; if they miss, the market will likely re-rate these names down violently because the equity story has become a leverage-on-confidence trade.

What the consensus may be missing is that shortages eventually become demand destruction, and in semis that tends to show up first in capex pauses and delayed server builds, then in inventory correction. That creates a late-cycle asymmetry: the stocks can keep running while customers absorb pain, but once procurement teams start signaling budget reallocation or spec downgrades, the tape can break fast. The market is pricing a multi-year scarcity regime; the risk is a much shorter window where scarcity is real but already fully capitalized in valuations.

Key Decisions for Investors

• Stay long SNDK and MU into the next 4-8 weeks, but only via defined-risk structures: buy 1-2 month call spreads rather than outright shares if chasing here, because the market is now discounting multi-quarter scarcity and implied upside can compress quickly.
• Pair trade: long SNDK / short WDC or STX for 1-3 months. The cleaner AI-storage exposure should outperform the broader HDD/legacy storage complex if pricing remains tight; use the short leg as a hedge against a general memory fade.
• On NVDA and AMD, reduce add-on exposure on strength over the next earnings cycle. These are upstream beneficiaries, but if memory prices stay elevated, the bottleneck shifts to system integrators and can cap accelerator unit growth before it hits chip demand.
• Sell downside in MU only after the next quarter’s pricing commentary. A 6-12 week put-selling or put-spread-financing strategy makes sense if guidance confirms contract repricing, since the fundamental reset would likely lag by at least one quarter.
• Watch for a tactical short setup in SNDK/MU if server OEMs start flagging shipment delays. That would be the first sign of demand destruction; a 10%-15% drawdown can happen fast if buyers begin deferring deployments rather than absorbing higher input costs.