01 · Equity deep-dive — synthesized analyst desk
MU
$943.78 ▼ 24.8% off mid-2026 peak
NASDAQ · SEMICONDUCTORSMKT CAP ≈ $1.07T52-WK $117.29 – $1,255.00AS OF SEP 2, 2026

AI data centers made Micron a trillion-dollar foundry. The market is pricing a cyclical memory bust that may never come.

High Bandwidth Memory (HBM) for AI accelerators is completely sold out into late 2026, driving trailing net income past $50B. Yet the stock has pulled back nearly 25% because the market expects a classic DRAM oversupply bust. Has AI broken the brutal boom-bust memory cycle, or is this just the loudest boom in history? Four analyst lenses, three scenarios, four time horizons.

The verdict · TL;DR
One question decides the stock: Is HBM demand structural, or cyclical? TTM revenue surged past $90B and EPS is up 700% Y/Y, but the stock retreated on fears of a $25B capex hangover. The base case argues that the memory oligopoly and insatiable AI data center demand create a structurally higher profit floor. The bear case — a brutal 2028 supply glut — remains the primary hazard. A fundamental test of whether "this time is different."
5-yr · prob-weighted
$1,481
+57% vs $943.78
52-week playback · where the tape sits ❚❚ Pulling back from the AI premium
$943.78 · Sep 2, 2026 consensus $1,295 · +37%
$117.29 · 52-wk low $1,255.00 · 52-wk high
Price history + cone of outcomes · 2024 → 2031
HISTORICALBULLBASEBEARPROB-WTD
$3,200$2,560$1,920 $1,280$640$0 202420252026 202720282029 20302031 $285 · Early '24 $1,255 peak $1,481 $1,100$1,150$1,250 $3,000 $1,350 $225 TODAY · $943.78

Gray line = Micron's actual stock history into today (riding the generative AI wave to a $1,255 peak, before cooling); colored paths = synthesized scenario midpoints forward, probability-weighted (base 50% · bull 25% · bear 25%). Linear scale, mid-year marks. Wall Street 12-month consensus ≈ $1,295 (+37%).

Re-weight the scenarios

The $1.07T valuation hinges on a judgment call: is this a structural paradigm shift, or the peak of a cycle? Drag to set how likely the bear and bull cases are (base takes the remainder); the blended target below, the dotted line on the chart, and the prob-weighted row of the scenario cards all update live.

25% bear 50% base 25% bull
Blended 5-yr expected $1,481 +57% vs $943.78
+167%
TTM Revenue Growth ($90.3B)
72.6%
Gross Margin
+701%
TTM EPS Growth ($44.17)
13.0x
Forward P/E Multiple
$25.3B
TTM Capital Expenditures
1.2 TB/s
HBM3E Bandwidth Delivery
-30%
HBM3E Power Usage vs Rivals
~100%
HBM Capacity Sold Out Thru '25
02 · The panel — four ways to read the same tape

Four analyst lenses, four answers

The same silicon supercycle supports wildly different conclusions depending on your view of memory economics. Each lens below is a synthesized expert perspective with its own 12-month target.

Growth PM

The Cycle Breaker

AI is fundamentally reshaping data center architecture. High Bandwidth Memory (HBM) isn't a commodity; it's a tightly co-engineered bottleneck for GPUs. Micron locked in 5-year contracts at structural premiums. With forward EPS blowing past $150 and gross margins over 72%, applying historical "cyclical" multiples to a structural compounder is the biggest mistake in semis right now.

12-MO TARGET $1,650 · 11x forward EPS
Macro / Moat Analyst

The Capacity Gatekeeper

The memory market is a disciplined oligopoly (Samsung, SK Hynix, Micron). They learned from past gluts. HBM consumes roughly 3x the wafer capacity of standard DRAM. Because they are prioritizing HBM for AI, they are artificially starving standard DRAM supply, keeping overall pricing structurally elevated. The floor has been permanently raised.

12-MO TARGET $1,300 · aligned w/ consensus
Memory Cycle Skeptic

The Capex Hangover

Cyclicality is undefeated. Micron's TTM capex soared to $25B+. History proves that every time the industry spends this aggressively, a glut follows 18 months later. If AI data center build-outs pause even slightly in 2027/2028, all that depreciation hits the income statement against falling memory prices. Earnings will crater back to $15–$20.

12-MO TARGET $850 · multiple compresses on peak earnings
Value / FCF Analyst

The Cash Buffer

Net income topped $50B over the trailing twelve months. Even with heavy capex requirements, the sheer volume of operating cash flow provides a massive buffer. At a forward P/E of 13x, the market is already pricing in a moderate cyclical downturn. Downside is protected by enterprise solid-state drive (SSD) replacement cycles and long-term cloud contracts.

12-MO TARGET $1,100 · fair value floor
03 · Wall Street's read

Wall Street 12-month price targets

What the sell-side expects over the next year. Bars are sorted low to high; the dashed line is today's $943.78.

Consensus ≈ $1,295 (+37%) · selected names, range $850–$2,000
BUYHOLDSELL
Cautious Bear $852 Citigroup $1,150 New Street Research $1,250 Mizuho $1,300 BMO Capital $1,300 KeyCorp $1,500 Raymond James $1,500 Needham $1,650 DA Davidson $2,000 TODAY · $943.78

Sell-side 12-month targets — a selection of the ~45 firms covering Micron; the full consensus is ≈ $1,295, about +37% above today, with a Strong-Buy skew. The dashed line marks today's $943.78: practically the entire street views the current price as a mispriced dip, believing peak earnings are further out than the market fears. Firms, ratings, and targets are illustrative combinations of recent coverage.

04 · Price scenarios — 1 / 2 / 3 / 5 years

Where the wafers land

Synthesized scenario midpoints (mid-year). Returns shown vs. today's $943.78. These are illustrative frameworks — five-year outcomes hinge entirely on whether AI data centers pause their builds or expand into AGI infrastructure.

1 Year

Mid-2027
Bull$1,400+48%
Base$1,100+17%
Bear$600−36%
Prob-wtd$1,050+11%

2 Years

Mid-2028
Bull$1,800+91%
Base$1,150+22%
Bear$400−58%
Prob-wtd$1,125+19%

3 Years

Mid-2029
Bull$2,400+154%
Base$1,250+32%
Bear$300−68%
Prob-wtd$1,300+38%

5 Years

Mid-2031
Bull$3,000+218%
Base$1,350+43%
Bear$225−76%
Prob-wtd$1,481+57%
Bull case — show the assumptions & math
The AI memory supercycle is structural. Data centers shift permanently to HBM-heavy architectures, eradicating the boom-bust cycle. Earnings compound through 2030 as margins stabilize at ~70%.
EPS scales to ~$150 by 2031 × structurally repriced 20× multiple → ≈ $3,000 · 5-yr price CAGR ≈ +26%/yr
Base case — show the assumptions & math
Cyclicality remains, but the trough is much higher. HBM disciplines the market and prevents a standard DRAM glut. Peak earnings moderate, but average annual EPS hovers strongly around $90.
Mid-cycle EPS ≈ $90 by 2031 × normalized 15× multiple → ≈ $1,350 · 5-yr price CAGR ≈ +7%/yr
Bear case — show the assumptions & math
A brutal capex hangover. AI hyperscaler budgets exhaust by 2028, just as $25B+ in annual memory capex hits the market. Oversupply crushes margins, driving earnings back to pre-AI norms.
EPS craters to ~$15 by 2031 × depressed 15× trough multiple → ≈ $225 · 5-yr price CAGR ≈ −25%/yr
05 · Follow the cash

Revenue, capex, free cash flow & debt ($B)

Where the money goes. The staggering rise in revenue and free cash flow is funding a massive, $25B+ capex buildout. The bear thesis worries the olive bar will vanish while the clay bar stays high.

Annual revenue, capex, operating/free cash flow & debt · 2023 → 2026
REVENUECAPEXFREE CASH FLOWTOTAL DEBT
$0$20$40$60$80$100 2023202420252026 (TTM)

Micron's AI inflection in one view: trailing revenue exploded past $90B (sky), generating enough operating cash to fund an unprecedented $25.3B capex buildout (clay) while still leaving ~$26B in free cash flow (olive). Debt levels (slate) remain utterly flat — meaning this expansion is purely cash-funded. Figures illustrative based on trailing 12-month data into late 2026.

06 · Earnings power

EPS path underpinning the targets ($)

Valuations are anchored to earnings. Here is the earnings trajectory that drives the base case targets, assuming a moderated cycle rather than a total collapse.

Adjusted EPS · reported vs. estimated, 2024 → 2031E
REPORTEDBASE ESTIMATE
$0$32$64$96$128$160 202420252026 TTM2027E2028E2029E2030E2031E $4.00 $15.00 $44.17 $72.00 $85.00 $90.00 $95.00 $90.00

Adjusted EPS — Gray = actual trailing figures; Sky = base case forward estimates. Note that the base case expects earnings to peak and moderate structurally higher, avoiding the catastrophic sub-$10 trough that the bear case fears. $90 in 2031 EPS at a normalized ~15× multiple yields the $1,350 base-case target.

07 · Growth scorecard

The business is scaling at unprecedented speeds

Year-over-year figures exiting Q3 2026. The extreme pace of growth is what makes valuing MU on historical metrics so dangerous.

Year-over-year growth by metric · Trailing Twelve Months
COREAI / FRONTIER
Standard DRAM Rev. +40% Total Revenue +167% Operating Income +345% EPS Growth +701% HBM Capacity Output +900%+ (Sold out)

The fundamental disconnect: earnings grew 700% Y/Y while the stock is down 25% from its highs. High Bandwidth Memory (clay) is ramping exponentially off a small initial base to serve NVIDIA/AMD accelerators, dragging operating leverage higher alongside it.

08 · The debate

Bull vs. Bear

The entire valuation argument compresses into one disagreement: has structural data-center demand killed the memory super-cycle, or are we just at the very top of it?

▲ THE BULL CASE

  • HBM is structurally different. HBM3E is a complex, co-packaged logic/memory stack with high defect barriers, not a plug-and-play commodity. It commands sticky, long-term contracts.
  • Supply discipline is real. With HBM taking ~3x the wafer capacity of normal DDR5, the "Big Three" (Micron, Samsung, SK Hynix) are severely constrained from accidentally oversupplying legacy memory.
  • Data center math. 2026 McKinsey analysis models AI demanding ~70% of total data center capacity by 2030; language models scale memory bandwidth linearly.
  • Best-in-class power metrics. Micron's HBM3E draws ~30% less power than rivals, a critical advantage for hyper-scalers constrained by grid power.
  • Valuation reset. At ~13x forward EPS, the market is pricing Micron like a cyclical steel mill at peak earnings, ignoring its $26B FCF generation.

▼ THE BEAR CASE

  • $25B in Capex will haunt them. Micron's trailing capital expenditure equals its entire revenue from just two years ago. High fixed costs mean catastrophic margin compression if demand softens.
  • AI CapEx cannot compound forever. Hyperscalers are front-loading NVIDIA GPU purchases to win the AI arms race; if inference revenue fails to match the infrastructure cost, GPU/HBM orders will evaporate in 2028.
  • Samsung wakes up. Samsung stumbled on HBM3, handing market share to SK Hynix and Micron. When Samsung's immense foundry capacity inevitably qualifies and scales HBM, the supply shortage ends.
  • China's CXMT wildcard. Legacy DRAM pricing could crash as China heavily subsidizes local fabs (CXMT) to capture low-end memory, flooding the market.
09 · Risk map

Risk map — likelihood × impact

Where each risk sits. The severe threat to the stock is an AI infrastructure pause (Possible × High), which would immediately trigger the classic cycle bust.

Low impact
Medium impact
High impact
Likely
  • Legacy DRAM softening
  • Samsung HBM oversupply
Possible
  • NAND price pressure
  • GPU packaging bottlenecks
  • AI CapEx pause / cycle bust
Tail
  • Geopolitical / Taiwan shock

AI CapEx pause / cycle bust

Possible × High

Hyperscalers slow their data center builds, leaving Micron holding billions in unutilized HBM fabrication capacity.

Geopolitical / Taiwan shock

Tail × High

Disruption in the South China Sea cripples TSMC and global supply chains, paralyzing the AI hardware ecosystem.

Samsung HBM oversupply

Likely × Medium

Samsung perfectly scales its HBM yields by late 2027, turning a high-margin supply shortage into a price war.

GPU packaging bottlenecks

Possible × Medium

TSMC CoWoS packaging limits the amount of GPUs NVIDIA can actually ship, indirectly capping Micron's ability to sell HBM.

Legacy DRAM softening

Likely × Low

Consumer PCs and smartphones experience sluggish upgrade cycles, depressing standard memory prices.

NAND price pressure

Possible × Low

Storage (NAND) doesn't enjoy the same AI premium as memory (DRAM), keeping ~30% of Micron's business in commodity cycles.

10 · Plain-language glossary

The jargon, decoded

Hover the dotted terms in the metrics, or scan the desk's working definitions here.

HBM (High Bandwidth Memory)
Memory chips stacked vertically like a skyscraper, sitting directly next to a GPU to feed it data insanely fast. Critical for AI.
DRAM
Dynamic Random Access Memory. The standard short-term working memory in PCs, phones, and servers.
Memory Cycle (Boom/Bust)
The historical pattern where high chip prices cause over-investment (Capex), leading to oversupply, price crashes, and massive losses.
Hyperscalers
Massive cloud providers (Microsoft, Amazon, Google, Meta) building the multi-billion dollar AI data centers driving MU's demand.
Yields
The percentage of functional chips on a silicon wafer. HBM has notably lower yields than standard DRAM, making it harder to manufacture.
NAND
Flash storage memory (like what's in a solid state drive). Retains data when power is off, unlike DRAM.
Capex (Capital Expenditure)
Cash spent buying physical assets — in Micron's case, ultra-expensive ASML lithography machines and fab buildings.
Prob-weighted
Each scenario's price × its probability, summed into a single expected value across bear, base and bull.