MU

FY2023: Micron's revenue halved, DRAM prices fell in the high-40s percent, net income was a loss of $5.83 billion, and adjusted free cash flow was negative $5.45 billion while capex continued. Now FY2026 capex is guided at $20 billion. Why is this time different?

The sheet does not assert this time is different. FY2023 revenue was $15.54B (−49% year over year), GAAP net income was $(5.83)B, DRAM ASPs fell in the high-40s percent, NAND ASPs fell in the low-50s percent, and adjusted FCF was $(5.45)B — per the FY2023 Q4 press release and 10-K risk factors. FY2025 revenue rebounded to $37.38B (+49%), net income $8.54B, gross margin 40%, with CMBU up 257% year over year to $13.52B. HBM calendar 2025 and 2026 supply is under binding price/volume agreements including HBM4. FY2026 capex is guided at approximately $20B — approximately 2.4 times FY2025 depreciation of $8.28B. The 10-K risk factors explicitly state that memory ASP cycles have ranged from +35% to −high-40s% annually since 2017.

Your own 10-K says if HBM demand weakens, suppliers may shift capacity to conventional DRAM and create oversupply. You are spending $20 billion in FY2026 on that scenario. How do you read that risk factor?

The language is in the FY2025 10-K risk factors: generative AI drove HBM demand but "demand may fluctuate," and if it weakens, suppliers can shift capacity to conventional DRAM, creating oversupply. That is Micron's primary filing. HBM requires a 3-to-1 wafer trade ratio versus DDR5 for the same supply volume, per the FQ1-26 investor presentation — meaning each HBM bit consumes disproportionate fab output. Idaho first wafers are guided for 2H calendar 2027 at the earliest; New York and Singapore HBM packaging buildouts follow. The gap between when that new capacity opens and whether AI-driven HBM demand holds through 2027–2028 is the capex-cycle risk the sheet scores at red-amber (~68–78).

One customer is 17% of Micron's total revenue. You have $14.58 billion in debt — more than double what it was in FY2022 — and a $20 billion annual capex commitment. What happens to fab utilization if that customer pulls back?

Per the FY2025 10-K Note 28, one customer represented 17% of FY2025 total revenue; the customer is not named in the filing. Total debt carrying value was approximately $14.58B (current $560M plus long-term $14,017M), more than double the approximately $6.8B from FY2022 per the same 10-K. CHIPS Act grants of up to $6.4B plus a 35% investment tax credit provide partial offset, but grants are conditional with clawback risk if milestones are missed — $1.02B in unearned government incentives sits on the balance sheet. One customer at 17% of revenue combined with a $20B build cycle and $14.6B in debt is the concentration-leverage intersection the sheet flags.

CEO Sanjay Mehrotra sold over $65 million in Micron shares in 2026 alone — all on pre-set 10b5-1 plans. CFO Murphy sold $28.4 million in late 2025. Is that a red flag?

The sheet scores Indicator 3 at amber (~38–48) and is explicit that 100% of disclosed selling is via 10b5-1 plans. Mehrotra's 2026 sales include a May 1 tranche of 40,000 shares at approximately $511–$545 per share (~$21–22M) and a May 29 tranche of 37,439 shares at approximately $942–$969 per share (~$35M+), both on a plan adopted January 30, 2026. CFO Murphy sold 126,000 shares for approximately $28.4M in October 2025 on a plan adopted July 31, 2025. The honest call: pre-set plans adopted before a parabolic stock move are systematic diversification, not a standalone red flag. The score is amber, not red. The context — $65M+ in CEO 2026 sales executed at prices ranging from roughly $511 to $969 or higher — is worth stating plainly alongside the plan disclosure. | FY2025 Q4 earnings press release (SEC ex-99.1) | 2025-09-23 | Capex, cash flow, FY2025 totals | | FY2025 Q4 investor presenta…