Home

MU — Deck

Micron Technology Inc · MU · NASDAQ

The cheapest large-cap semi at 8x forward earnings — supercycle or peak trap?

$466
CMP
$526B
Market Cap
8x
Fwd P/E (FY2026E)
$12.07
Q2 FY26 EPS (was -$5.34 in FY23)
Up 556% in 12 months, 74% gross margin in Q2 FY26, HBM sold out through 2027
1 · Business

DRAM/NAND oligopolist reborn as an AI memory infrastructure play

  • Cloud Memory (CMBU). $13.5B in FY25, up 7x in two years, 45% operating margin — HBM3E for hyperscale AI is the profit engine.
  • Mobile & Client (MCBU). $11.9B, still the largest segment but grew just 2% as management starved it of supply to feed data center.
  • Automotive & Embedded (AEBU). $4.8B at 12% margin — the steadiest segment, content growth driven by EV and ADAS adoption.
One of three companies on earth that can make leading-edge DRAM — moat is real but narrow, dependent on parity at each node.
2 · Numbers

From a $5.8B loss to $12 quarterly EPS in three years

74.4%
Q2 FY26 Gross Margin (was -9% in FY23)
8x
Fwd P/E (peers trade 32-40x)
$1.7B
FY25 FCF (capex consumed 42% of rev)
32%
ROIC (through-cycle avg 10-15%)

~60% of COGS is fixed depreciation, so every dollar above the ~$20B breakeven falls through at high incremental margins. The central question: if mid-cycle gross margin settles at 50% (not the historical 35%), normalized EPS is ~$35 and the stock trades at 13x — still cheap.

3 · People

B+ governance — strong execution team, heavy insider selling

  • CEO Mehrotra. Co-founded SanDisk, 45 years in memory, holds $456M in MU stock — but sold $18M in 6 months and added Chairman title Jan 2025.
  • Insider selling breadth. All six NEOs sold a combined $124M in shares over 6 months. No executive bought during the March 2026 dip to $322.
  • Liu's contrarian buy. Former TSMC Chairman T. Mark Liu purchased $7.8M in open-market shares in Jan 2026 — the only informed insider buyer.
  • Board refresh. Liu (TSMC) and Simons (Deloitte semi audit) joined in 2025; 7 of 8 independent. Combined Chair-CEO is a governance negative.
4 · Story

Commodity bust to AI supercycle in 24 months

The Downturn (FY2023). Revenue fell 49%, gross margins went negative, and Micron wrote down $1.8B in inventory. Management cut 15% of headcount, suspended bonuses, slashed exec pay. China's CAC banned Micron from critical infrastructure. Supply discipline held.

The AI Rebirth (FY2024-present). HBM demand created pricing power Micron never had — hyperscalers now prepay under 5-year agreements. Revenue tripled from $15.5B to a $95B+ run rate. The company reorganized segments around AI adjacency, secured $6.4B in CHIPS Act grants, and is building fabs across four countries.

Credibility: 7/10. Hit every tech milestone (HBM3E, HBM4 samples) but Boise fab slipped from CY2025 to CY2027.
5 · Web Intel

Street consensus overwhelmingly bullish at $520 avg target, but insiders keep selling

  • Q2 FY26 blowout. $23.9B revenue beat estimates by 24%, EPS of $12.20 beat by 39% — fourth consecutive record quarter with a 30% dividend hike.
  • TurboQuant scare. Google's March 2026 algorithm demo showed 6x AI memory reduction; MU dropped 10% intraday before BofA called it a buying opportunity.
  • YMTC lawsuit. Chinese rival sued Micron in US court in June 2025 over spyware allegations — a new front in the China geopolitical battle.
EVP Sadana sold 24,000 shares ($10.1M) on April 10 — five days ago — while the stock hit all-time highs.
6 · Risks

Three risks that could break the thesis

  • Cycle reversion. Memory has humbled every 'this time is different' call. At 74% gross margin, the operating leverage that created $12 EPS works violently in reverse — FY23's $5.8B loss came on a smaller fixed-cost base.
  • AI memory efficiency. Google TurboQuant demonstrated 6x memory reduction per AI workload. If inference dominates and memory intensity flattens, demand disappoints without any supply overshoot.
  • Capex execution. Building fabs in Idaho, New York, India, and Singapore simultaneously on $20-25B annual spend. Boise slipped 2 years; Clay NY hasn't broken ground.
7 · What's Next

One earnings print and one pricing signal before the cycle verdict

  • May 2026. DRAM contract price negotiations set Q3 pricing tone — any softness is the earliest cycle-turn signal before earnings.
  • Mid-June 2026. Q3 FY2026 earnings — management guided ~81% gross margin, the highest in memory history. Beat or miss defines the next 6 months.
  • H2 2026. HBM4 volume ramp for NVIDIA Vera Rubin — execution vs Samsung/SK hynix determines 2027 market share.
  • Sep 2026. OECD Pillar Two takes effect in Singapore — Micron's $1.05B tax incentive at risk, could materially raise the effective tax rate.
Q3 FY2026 gross margin is the single datapoint that will confirm the supercycle or trigger the peak narrative.
8 · For & Against

Lean cautiously bullish — HBM physics outweigh the insider selling, but barely

  • For. HBM supply is physically constrained (3x wafer area per bit) and sold out through 2027 — this is fab physics, not narrative.
  • For. At 8x forward earnings with 49% revenue growth, Micron is the cheapest large-cap semi by far — PEG of 0.22 vs NVDA's 1.92.
  • For. Liu's $7.8M open-market buy and the 30% dividend hike signal informed confidence the cycle has legs.
  • Against. All six NEOs sold $124M in 6 months — no executive bought the March dip to $322, a tell worth noting.
  • Against. FY25 FCF was just $1.7B on $37B revenue; $20-25B capex in FY26 means cash generation depends on the supercycle continuing.
  • Against. Google TurboQuant and Samsung's aggressive HBM expansion threaten the demand and supply assumptions behind 74% margins.
My View — HBM supply constraint is physical reality, and 8x forward earnings offers genuine margin of safety. The flip condition: Q3 FY26 gross margin below 70% confirms the cycle is turning faster than bulls assume.

Watchlist to re-rate: Q3 FY26 gross margin vs 81% guide, DRAM contract pricing trend in May, Samsung HBM capacity timeline