The decisive print arrives in 41 days — Q3 FY2026 either confirms the franchise or breaks it.

• The Q2 print that drove the re-rate. March 18 brought $23.9B revenue (+196% YoY), 75% non-GAAP gross margin, $12.20 non-GAAP EPS — and a Q3 guide of $33.5B and ~81% gross margin. Forward P/E (on FY27 consensus ~$102) collapsed to ~8× and the stock ran from ~$462 on print day to an $812 all-time high eight weeks later.
• Three non-recurring lifts sit inside that 75%. Days payable stretched from 105 to 137 (≈$2B working-capital tailwind), the Singapore tax incentive cut FY25's effective rate to 11.6% from 36.4% ($1.05B / $0.93 EPS), and an FY24 inventory-recycle benefit has now lapped. Strip them and Q1's clean 56% gross margin is the truer level.
• Q3 FY2026 prints on or about June 24. A clean ≥80% gross margin without a footnoted one-time would corroborate HBM franchise economics. A sub-78% print, or 81% with a disclosed settlement, would ratify the peak-quarter read and force material downward revisions to FY27 consensus.

Record everything — and the cash to fund the next leg sits behind $25B+ of capex.

Cyclical earnings power has rebuilt to a higher peak than FY2018, driven by HBM mix shift and AI-tight DRAM. Capex is the gating valve: FY25's $15.9B absorbed 91% of operating cash flow, FY26 guides above $25B, and FY27 takes another $10B+ on top. Through-cycle free cash flow has averaged ~$2B per year for eight years. Defending $880B of equity here requires consensus FY26 EPS near $58 and FY27 near $102 to land and to convert to cash through the fab cycle.

HBM is a franchise. The other two-thirds of the company is still commodity memory.

• One segment carries the re-rate. CMBU (HBM plus hyperscale DRAM) earns 45% operating margin on 36% of FY25 revenue. The other 64% — mobile/client at 17%, OEM data center at 30%, auto/embedded at 12% — does not carry the moat. The variant view is that a franchise multiple is being applied to the whole company while the franchise lives in one of four segments.
• The peer benchmark — SK hynix's HBM-rich mix prints higher margins. SK hynix earned 72% operating margin in calendar Q1 2026 vs Micron's 69% in Q2 FY26 (a ~3-point peak-quarter gap), and 49% vs Micron's 26% on FY2025 full-year (a 22-point annual gap). SK hynix HBM share is 62% versus Micron's 21% (up from 9% a year ago). Either Micron closes the annual gap, which is the entire bull case, or the peer comparison proves the franchise lives in Icheon, not Boise.
• Qualification resets every generation. Micron took share when Samsung stumbled on HBM3E in 2024. Samsung announced HBM4 mass production in February 2026; one industry report had the Vera Rubin Ultra HBM4 split 70/30 SK hynix/Samsung with Micron excluded, partly neutralized by Micron's March 27 mass-production release. CY2027 allocation is still open.

Every memory glut since 2014 followed combined capex above $70B. CY2026 will print above $70B.

• Three negative-gross-margin troughs in fifteen years. FY2008–09, FY2012, FY2023. The most recent cut revenue in half, took a $1.83B inventory write-down, fired 15% of headcount, and printed a $5.8B net loss. Micron's own 10-K still names this risk in unchanged language.
• The Big Three are building the next round of capacity now. Combined CY2026 capex above $70B — Micron above $25B, SK hynix ~$27B, Samsung also expanding. Micron's FY27 capex steps up another $10B+. New wafer capacity lands 2027–2028, exactly as the first multi-year HBM contracts approach renewal.
• The bull's structural answer is still unfinished. CY2026 HBM is pre-sold under multi-year strategic customer agreements and HBM die-area economics tighten the rest of DRAM. Both are real. Neither has been tested through a full cycle — and a second SCA, the disconfirming evidence for the bear, has not yet been signed.