Micron Trades at 6 Times Next Year's Earnings. The $38 Billion of Capacity That Ends This Cycle Doesn't Open Until 2028.

Micron reported fiscal Q3 revenue of $41.5B (up >4x YoY) and gross margin of 84.6% (vs. 74.4% in fiscal Q2), with operating cash flow rising to $25.4B (up >5x YoY). Management guided fiscal Q4 revenue to about $50B, gross margin ~86%, and EPS ~ $30.73, supported by data-center demand and expectations that DRAM/NAND supply remains below demand. At the same time, SK Hynix approved ~$38B for two new fabs with first clean rooms in Dec 2028 (NAND M17) and Jun 2029 (DRAM Y2), while Micron has 16 take-or-pay customer contracts covering ~20% of DRAM and ~1/3 of NAND volume through end-2030, helping extend revenue durability beyond the near-term cycle.

Analysis

The market is still treating memory like a mean-reverting commodity, but the supply response is now so delayed that the near-term setup looks closer to a duration trade than a cycle trade. New wafer capacity approved today does not pressure the 2026-27 pricing stack; it mainly validates that customers and suppliers expect elevated AI/storage demand to persist, which supports MU’s forward earnings power and keeps gross margin volatility unusually low for a memory name.

The second-order winner is MU’s balance sheet and capital return story: when a cyclical business can lock a larger share of revenue via take-or-pay structure, the equity deserves a higher quality multiple than the market typically grants a peak-earnings name. The losers are the system integrators and server OEMs with the least pricing power on bill of materials inflation, especially names where memory is a large share of build cost and end-demand is already sensitive to enterprise IT budgets.

The key risk is not fab timing; it is AI demand digestion. If hyperscaler capex slows or inventory turns normalize faster than expected, spot pricing can break before the new fabs matter, which would hit MU’s uncontracted volume first and compress the multiple quickly. The contrarian point: the consensus is focused on the next downcycle, but may be underestimating how much of MU’s revenue is being structurally de-risked before that downcycle arrives; still, only part of the book is protected, so this is not a full annuity story.

Key Decisions for Investors

• Add to MU on pullbacks over the next 1-3 months; favor buying weakness after earnings or sector de-ratings rather than chasing strength. Base case is continued forward estimate revisions through 2026, with risk/reward skewed to the upside as long as data-center demand stays intact.
• Pair trade: long MU / short SOXX for a 3-6 month horizon to isolate the memory-duration thesis from broad semiconductor beta. This should work if MU keeps outperforming on forward guidance while the rest of semis remain more fully valued to the same AI narrative.
• Short DELL or HPE against MU on a 1-2 quarter horizon as a second-order margin-pressure trade. Memory cost inflation tends to squeeze integrators before it shows up in final demand, and these names have less ability to absorb BOM pressure than the component supplier.
• Set a falsifier alert on any evidence of AI capex digestion: a guide-down in hyperscaler spend, DRAM spot prices falling sharply, or MU commentary shifting from tight supply to normalization. If that happens, reduce MU exposure quickly because the stock can re-rate faster than the fabs can impact supply.