Micron Technology, Inc.

Four independent signals converge on the pivot interpretation. (1) Micron retired the 29-year-old Crucial consumer brand on 2025-12-03 with wind-down completion by end-February 2026 (ev_050, ev_051); the alternative reading — Crucial is a margin-driven discontinuance rather than a strategic pivot — is undermined by the simultaneous CapEx reallocation to HBM/data-center lines. (2) The Manassas 1α DRAM start (May 22 2026) is explicitly framed as AI-data-center and national-security memory with "no consumer application" (ev_065, ev_076). (3) Q2 FY2026 revenue grew very likely driven by an HBM mix that locked 60-70% of DDR5 demand under multi-year contracts including NVIDIA (ev_073, ev_086). (4) Micron crossed the $1T market cap on 2026-05-26 (ev_064, ev_071). Confidence is HIGH because the four signals span independent source types and no surviving alternative reconciles all four.

NVIDIA's ~80% data-center AI GPU share (ev_040) makes it the demand-side monopoly for HBM3/HBM3E/HBM4. Micron's entire 2024 HBM3 output was reportedly pre-committed (ev_042) and a multi-year supply agreement was disclosed in Q2 FY2026 (ev_073). The competing hypothesis — that the EBU automotive segment under ent_058 and continued enterprise DRAM diversify the book sufficiently — is unlikely to materially offset NVIDIA exposure on a 12-24 month horizon, because Crucial revenue is being intentionally retired (ev_050) and SK Hynix's reported 27% revenue concentration to NVIDIA (ev_060) is a comparator suggesting industry-wide HBM revenue is structurally concentrated. Confidence is HIGH on the direction; the exact magnitude is MODERATE because Micron has not disclosed customer concentration in detail.

Two 2026 academic sources directly contradict on Q2 2025 share: SSRN 6504361 reports a yield-driven decline from ~40% to 17% (ev_034); SSRN 6395479 reports 21% on a recovery trajectory (ev_035). A Q3 2025 reading places Micron at 11% (ev_034). TrendForce reported a 24% year-end-2025 target with HBM scaling to four GPU/ASIC clients (ev_047); AstuteGroup reports Micron overtook Samsung for second place by September 2025 (ev_048); Counterpoint Research reports Micron at ~33% of Q1 2026 DRAM (ev_075). The June 9 2026 MU-stock drop of 11% after NVIDIA's HBM4 "nod" to SK Hynix/Samsung (ev_083) is consistent with a market that is also uncertain about the trajectory. Confidence is MODERATE because the disagreement is between sources of comparable grade.

The $3.5B Idaho subsidy on an $8.6B fab cost (ev_037) and the NY direct-funding agreement (ev_029, EX-10.8) signed December 2024 represent a federal-financing dependency that the 2025-11-07 NY fab delay (ev_053) and the contemporaneous Idaho-second-fab acceleration with reallocated funds (ev_052) confirm is being actively renegotiated. The alternative — that Micron would proceed on the same timeline absent CHIPS Act funds — is very unlikely given the disclosed $100B headline investment is explicitly framed as conditional on federal support. Confidence is HIGH because both the dependency (SEC exhibits EX-10.7 through EX-10.10) and the political sensitivity are documented in independent source types. Premortem: an administration accelerating CHIPS Act funding instead of curtailing it would yield the symmetric upside scenario — the dependency is bidirectional.

CXMT capacity increased in 2025 despite export controls (ev_041); KrAsia reports the most aggressive memory expansion ever from CXMT and YMTC amid the global supply crunch (ev_054); YMTC is building an all-Chinese-tools production line and targeting 15% global NAND by late 2026 (ev_055); a third Wuhan fab is in progress (ev_056). CXMT holds 1,308 EPO-registered patents in HBM-adjacent classes (H10B/H10D/G11C) including stacking IP (WO2026118617, June 2026) (ev_087). On HBM specifically, Wong/Yeung/Lee characterize CXMT as likely to follow Samsung/SK Hynix in DRAM rather than lead (ev_046). The competing hypothesis — that CXMT's patent count signals near-term HBM competitiveness — is unlikely because patent count is a noisy proxy for production-grade HBM yield. Confidence is MODERATE.

SK Hynix and Samsung shipped paid HBM4 samples to NVIDIA ahead of Q1 2026 contract finalization for Vera Rubin (ev_077), while Micron showcased HBM4 at COMPUTEX 2026 (ev_083). The 11% decline on 2026-06-09 occurred against a stock that had tripled in 2026 (ev_064, ev_071), implying the market priced an expectation Micron had not yet earned. NVIDIA's confirmation Vera Rubin entered full production while HBM4 competition intensifies (ev_081) and the HBM4 mass-production delay to end-Q1 2026 driven by NVIDIA spec changes (ev_061) suggest the supplier set remains in flux. The alternative — that Micron has been strategically displaced from the HBM4 cycle — is unlikely because no source describes Micron as fully out of contention; the framing is "lost priority" rather than "lost qualification." Confidence is MODERATE because actual contract allocations are not disclosed.

Hunter.io enumerates 13,510 indexed Micron emails and confirms the deterministic pattern {f}{last}@micron.com with accept_all=true (ev_039) — a high-confidence input to spear-phishing operations targeting ent_057 (Jon Dickinson, VP Global Government & Public Affairs and CHIPS Act primary liaison), ent_060 (David Daycock, VP DTG Process Integration — directly accountable for HBM3E yield), and ent_069 (Prakash Rau Mokhna Rau, Director TD NAND Process Integration). MDN Observatory grades micron.com at D / 30 (ev_044, ev_088), with 2 of 10 web-security tests failing — consistent with absent or misconfigured Content-Security-Policy or X-Content-Type-Options headers. The TXT inventory confirms 22 enterprise SaaS dependencies (Atlassian x2, Adobe, Anthropic, Perplexity, OneTrust, MongoDB, Twilio, Palo Alto Networks, Reejig, Jamf, Salesforce, Microsoft 365, Marketo, etc.) (ev_020), any of which is an upstream attack path. The CAC 2023 ban (ev_094) was predicated on alleged cybersecurity risk; a D-grade public surface is rhetorical ammunition for adversaries even when the underlying risk is modest. Confidence is MODERATE because attack feasibility depends on internal controls Corvus cannot observe.