Micron: New Street upgrades to buy, says memory is no longer a cyclical story

New Street Research upgraded Micron Technology to buy, setting a $1,250 price target, arguing the memory cycle is structurally different from past boom-bust patterns. The note points to Micron’s stock up >10x since April 2025 lows while production value (COGS) is up only ~25%, and projects peak outcomes by 2030 of >$600B cash and >$150B annual free cash flow, with a milder downcycle (free cash flow burn trough of ~$18B). It also forecasts memory demand growth where AI drives ~two-thirds of the mix and implies a ~$2T–$3T memory market cap by 2030, supporting the premium valuation case for high-bandwidth memory.

Analysis

MU is being re-priced from a cyclical supply story into a scarce-asset AI infrastructure story. That is a real mechanism, but it also creates its own antidote: once returns on incremental capacity look durable, competitors will chase them, and the bottleneck can shift away from wafer supply toward packaging, substrates, and test equipment. In that regime, the cleaner second-order winners may be AMAT, LRCX, and KLAC over the next 6-18 months if the industry keeps reinvesting rather than rationing supply.

The near-term risk is that the market is extrapolating peak scarcity into a permanent margin structure. The first falsifier is any sign that AI capex growth moderates or that memory lead times normalize; that would hit the multiple before earnings do, because the stock is already discounting a much smoother path than history. The longer-term falsifier is classic industry behavior: if every player sees a premium, capacity expansion follows, and the “less cyclical” thesis gets tested by a delayed but still forceful oversupply cycle.

Contrarian view: the upside case may be less about MU becoming a perpetual compounder and more about memory becoming an acceptable way to express AI demand, which is a narrower claim. If that is right, the best risk/reward may be relative exposure rather than outright beta — long the highest-quality memory name versus legacy storage or less differentiated semi names. The market may also be underestimating how much of the eventual value accrues to the equipment and advanced packaging stack rather than to the memory vendors themselves.

Key Decisions for Investors

• Prefer a relative long MU / short WDC or STX pair over an outright momentum chase; thesis is that AI mix supports HBM pricing while legacy storage remains slower-growth and more exposed to enterprise budget normalization. Review after next earnings cycle for any guide-down in storage pricing or HBM lead times.
• If we want directional exposure, use MU call spreads rather than stock into strength; the stock already prices a long-duration multiple expansion, so defined-risk upside is better than paying for all of the secular story upfront.
• Watch AMAT, LRCX, and KLAC as secondary longs over 6-18 months if memory capex stays elevated; their earnings leverage can be stronger than MU’s if the industry responds to high returns by expanding capacity and advanced packaging spend.
• Set a downside alert on MU if AI capex chatter or memory pricing turns lower; any evidence of lead-time normalization or softer cloud spending should be treated as a multiple-compression trigger, not just an earnings issue.
• If MU continues to rerate, consider trimming into strength rather than adding after a multi-session extension; the key risk/reward inflection is when the market stops viewing the name as cheap cyclically and starts pricing it as perfect structurally.