Should You Buy the Dip on Micron?

Micron reported Q2 FY2026 revenue of $23.9B, up 75% Q/Q and 196% Y/Y. Management now expects capital expenditures of at least $25B (about $5B above prior guidance), and the stock has dropped roughly 20% from its $471 peak amid concerns over increased HBM competition and Alphabet's TurboQuant AI model compression lowering memory requirements. Micron says memory is sold out through 2026, expects substantial revenue and gross margin gains in coming quarters, and signed its first five-year strategic customer agreement, indicating strong demand despite near-term spending and competitive headwinds.

Analysis

Micron sits at a structural inflection where product mix (HBM vs commodity DRAM/NAND) and contract structure (short-term spot vs multi-year SCAs) will drive differentiated economics. Because HBM is a bandwidth/latency product with severe manufacturing scale economies, any entrant that wins design wins from hyperscalers can take meaningful share but cannot instantaneously expand capacity — that preserves pricing optionality for incumbents over the next 12–24 months even if unit growth decelerates.

Google’s model-compression announcement creates an important demand-fracture: it reduces per-inference capacity needs but increases the value of lower-cost, higher-utilization infrastructure. Hyperscalers will likely reallocate budget from raw-capacity purchases toward system upgrades (NICs, interconnect, software stacks) and longer-term platform deals; that shift favors vertically integrated suppliers who can bundle memory into system-level offerings. Expect hyperscaler procurement cadence to become more “utilization-sensitive” over the next 6–18 months, increasing quarter-to-quarter revenue volatility for pure-play memory names.

The biggest tail risk is an acceleration of software-led obsolescence: if model compression adoption crosses a threshold inside major clouds within a single procurement cycle, ASPs for premium HBM could compress rapidly while wafer-level capacity from recent capex comes online 12–36 months later, creating a squeeze on near-term ROIC. Conversely, a hardware counter-response (architectural changes from accelerator vendors that push bandwidth back upward) could restore pricing power quickly — monitor accelerator roadmaps and hyperscaler testbeds as leading indicators.

Net positioning should reflect asymmetry: ownership for structural AI demand but hedges for policy-rate/convexity and software disruption. Volatility will persist; the market is pricing a regime where cyclical capex and structural software gains battle for share over multiple years rather than inside a single quarter.

Key Decisions for Investors

• Initiate a core-long MU position (2–3% net portfolio) with a 12–24 month horizon; size to risk budget and set tactical protection: buy 9–12 month puts to cap downside to ~15% cost while keeping upside uncapped. Rationale: captures idiosyncratic upside if HBM tightness and design-win momentum persist; cost of protection limits drawdown if rapid software-led demand erosion occurs.
• Construct a calendar/vertical on MU to express asymmetric upside: buy a 12–18 month LEAP call (long-dated) and sell 3–6 month out-of-the-money calls to fund premium. Time to entry: post next hyperscaler procurement cadence release; risk/reward: limited funded premium today vs convex upside if HBM ASPs hold, break-even materially lower than straight equity exposure.
• Trade software-compression gamma: sell short-dated implied vol on GOOGL (3–6 month strangle) sized to 0.5–1% portfolio vega with active delta-hedging. Rationale: market has overreacted to initial model-compression headlines; earnings and ad-revenue durability provide a hedgeable short-VOL trade. Risk: large upside moves in GOOGL require disciplined delta hedging or defined-risk converts (call spreads).
• Reduce directional exposure to CPU/legacy memory cyclicity: underweight INTC into the next 6–12 months and redeploy into names with direct HBM content or software capture. Rationale: procurement shift to utilization-sensitive buys favors accelerators and hyperscaler software stacks over incremental CPU-led spend; risk/reward asymmetric if hardware roadmaps pivot.