Micron — Revenue Through the Cycle

$0B

$20B

$40B

$60B

$80B

$100B

$16.2B

FY15−0.5%

$12.4B

FY16−23%

$20.3B

FY17+64%

$30.4B

FY18+50%

$23.4B

FY19−23%

$21.4B

FY20−8%

$27.7B

FY21+29%

$30.8B

FY22+11%

$15.5B

FY23−49%

$25.1B

FY24+62%

$37.4B

FY25+49%

~$105B E

FY26E+181%

FIG. 001-A · Micron FY revenue, FY15–FY26E

Reported revenue Prior cyclical peak Cyclical trough FY26E (annualised from H1A + Q3 guide)

The cycle, in three acts

FY18 peak → FY19-20 reset. DRAM ASPs collapsed 30%+ as hyperscaler capex paused; revenue fell from $30.4B to $21.4B over two years. Classic memory cycle.

FY22 peak → FY23 wipeout. Post-COVID double-ordering reversed violently. Revenue halved YoY to $15.5B — the worst print since FY16. Gross margin went negative.

FY24 recovery → FY26 super-cycle. HBM3E for NVIDIA Hopper drove FY25 to $37.4B. Q2 FY26 alone printed $23.86B — more than full-year FY23. Q3 guide of $33.5B at ~81% gross margin is unprecedented for a memory company.

What's structurally different this time

HBM is allocation-priced, not spot-priced. Multi-year supply agreements with NVIDIA, AMD, and the hyperscalers de-risk the historical ASP volatility that defined every prior cycle.
Capacity is constrained, not glutted. HBM consumes ~3× the wafer area per GB vs standard DRAM — fab capacity is the binding constraint through CY2027.
Capex is now $25B+/yr (FY26 guide). Reinvestment intensity has stepped up from ~$10B historical run-rate; the question is whether the cycle persists long enough to earn an ROIC>WACC return on this build.
Market share is the swing variable. Micron sits at ~24% HBM share vs SK Hynix 43% and Samsung 33% — incremental share gains are the asymmetric driver from here.

Sources: Micron 10-K filings (FY15–FY25), Q2 FY2026 earnings release (Mar 2026), Micron IR. FY26E built from Q1A $13.64B + Q2A $23.86B + Q3 guide $33.5B + implied Q4 ~$34B. Chart for analytical discussion only — not investment advice.