MU thesis brief

• MU is the cleanest US-listed pure-play on AI memory scarcity. The business is tied to HBM, DRAM and NAND demand becoming bottlenecks in AI infrastructure rather than just commodity PC/phone cycles.
• Micron’s Q2 FY26 was not normal-cycle noise: $23.86B revenue, 74.9% gross margin, $12.20 EPS and $6.9B adjusted FCF. Q3 guide was even stronger: $33.5B revenue, ~81% gross margin and $19.15 EPS.
• The key phrase from Micron is “memory has become a strategic asset”. If true, this cycle has more durability than old DRAM boom/busts because hyperscalers and AI accelerator vendors need secure HBM supply.
• HBM4 is the centre: Micron began volume shipments of 36GB 12H HBM4 designed for NVIDIA Vera Rubin, has sampled 16-high / 48GB HBM4, and is building advanced packaging capacity for CY2027.
• Supply tightness is supported by cleanroom constraints, HBM trade-ratio, packaging bottlenecks and constrained DRAM/NAND industry supply. Industry reporting said Micron’s 2026 HBM supply was covered by pricing/volume agreements.
• The miss was process, not thesis. We correctly identified a strong business/StockWaves setup but treated RSI/extension as a block. L-035 now applies: strong thesis + high target = starter sizing, not no entry.
• Valuation is the fight. TTM sales/FCF multiples are extreme, but forward earnings can make the stock look cheap if FY27 EPS near $100 is credible and not peak-cycle fantasy.
• After the run to ~$928, the first $855 target is no longer opportunity; it is confirmation. The next decision is whether $1,124 remains realistic before/after Jun 24, and what risk level invalidates the chase.

Watch next

• Jun 24 earnings: Q3 delivery, Q4 guide, FY27 commentary, HBM mix, customer prepayments/visibility and capex plan.
• HBM share vs SK Hynix and Samsung, especially NVIDIA Vera Rubin / Rubin Ultra qualification signals.
• DRAM/NAND spot and contract pricing; any sign of supply moving from tight to oversupplied.
• Price behaviour after tagging $855: consolidation/retest versus blow-off toward $1,124.
• Whether analysts raise PTs after numbers or remain far below price — current average PT ~$653 is -30% below spot.

Main risks

• Peak-cycle EPS trap: memory stocks look optically cheapest near unsustainable earnings peaks.
• HBM share loss to SK Hynix/Samsung or weaker NVIDIA/hyperscaler qualification.
• Capex ramps faster than contracted demand and crushes FCF durability.
• AI capex slowdown at NVIDIA/Microsoft/Meta/Google/Amazon/Oracle/Broadcom/Cisco.
• After +894% in a year and RSI ~76, the technical risk is real even though it should not have blocked the original starter.