Phase Space AI

Valuation

Western Digital [WDC]

WDC — Valuation

Spot $453.09 (2026-07-29). Diluted shares 385m (company guidance; NOT the 345m outstanding, NOT AV's 376m). Net cash +$450m (company-stated). EV $173,990m.

---

1. COMPANY STATE — declared first

STATE B — cyclical, with an argued upward reset of the mid-cycle level, PLUS a structural break in the entity itself

Test satisfied twice over:

1. Exogenous cycle. Gross margin range over the last eleven quarters: 3.4% → 50.2%. Operating margin range: −24.3% → +38.6%. FY2023–24 was a genuine solvency-adjacent downcycle carried on ~$7.5bn of debt.
2. Structural regime change in the reporting entity. The SanDisk separation (2025-02-21) removed the Flash business. Pre-2025 WDC is a different company — half NAND, differently levered, differently cyclical. Under valuation.md State A's test ("no structural regime change"), WDC fails it explicitly.

Consequence: mid-cycle normalisation on both revenue and margin, and a hard constraint that no pre-separation period may be used as a margin, growth or multiple anchor. Only the five quarters from FQ3'25 (2025-03-28) forward are on the current basis.

Dating the cycle position

• Margin: at or near peak. FQ3'26 non-GAAP operating margin 38.6% is an all-time record and FQ4 guidance implies ~40.8%. Incremental gross margins of +70–75% for four straight quarters cannot persist indefinitely.
• Volume: mid-cycle, with contracted visibility. LTAs into CY2029, 12-month manufacturing lead times, exabyte growth conviction >25% CAGR, no unit-capacity additions.
• Multiple: top of the distribution. Trailing P/E 55.7x = 97.3rd percentile of 2,288 daily observations of WDC's own ten-year history (median 7.7x, p75 13.0x). Measured to twelve months ago, the maximum ex-trough trailing P/E was 36.2x and the median 6.8x. WDC's own realised-forward P/E (price ÷ next four quarters of actual EPS, corrected for the 1.3230 SanDisk spin-off price adjustment — see §5) ran 8.6x–17.1x across 2024–early-2025. It is now 34.9x. Every observation in the top decile of the multiple distribution was created inside the last twelve months.

Forecast start point: a record margin, an early-to-mid volume cycle, and a multiple ~3x its own historical forward level. Stated up front, as required.

Why State B is not licence for reflexive mean reversion

The historical mid-cycle for HDD margin was produced by a supply regime in which each margin recovery funded unit-capacity additions that then broke price. That mechanism has been explicitly withdrawn by WDC's own management ("we are not making any investments in adding unit capacity") and replaced with exabyte-based LTAs to CY2029. Reverting to a 2016–2024 average operating margin (~6% on the combined entity, and not even measurable on the HDD-only basis) would be applying a bound calibrated on a regime that no longer exists and on a company that no longer exists. That is the error the brief names.

Equally, extrapolating 40.8% is extrapolating a peak, on the name that is behind on the technology transition.

---

2. TERMINAL MARGIN — derived, bridged, defended

Terminal (FY2031) operating margin: 34.5%

34.5% is 1.8pp below the FY2026E full-year non-GAAP actual and 6.3pp below the guided FQ4 exit rate.

Why a terminal margin below the trailing full-year actual is defensible here — the explicit justification the brief demands. Three named, causal reasons, not conservatism:

1. WDC is behind on HAMR. Its 44TB HAMR is in qualification while Seagate's Mozaic 4 is in revenue at two of the largest CSPs and targeted at 70% of nearline exabytes by FY2027-end. WDC's R&D runs at 8.8% of revenue vs Seagate's 6.2%, and it must carry a dual ePMR/HAMR roadmap. A terminal margin above the trailing actual would require assuming that gap costs nothing.
2. The LTA book is explicitly smaller than customer demand (CEO: "The LTA volume… does not meet their full requirement"). The marginal exabyte prices outside the contract. In an up-cycle that is upside; in the terminal, mid-cycle state, it is where margin normalises.
3. The FY2026 margin was produced by a four-quarter run of +70–75% incremental gross margins — a rate that by construction cannot be the steady state, since it implies gross margin rising without bound.

The gap is 1.8pp, not 20pp. It is not a mean reversion to the old cycle; it is a small, causally argued step down from a guided peak. And 34.5% remains 8.5pp above the last same-basis pre-boom quarter (26.0%) and ~30pp above the trough.

Opex bridge

m_EBIT,T  =  m_gross,T − R&D − SG&A − other
    34.5% =    45.0%   −  6.5%  −  3.5%  −  0.5%

• m_gross,T = 45.0%. FQ3'26 actual 50.5% non-GAAP; FQ4 guide 51–52%; FQ3'25 40.1%. 45.0% sits between the last pre-boom quarter and the current level, ~5.5pp below the trailing quarter and ~6.5pp below the guided exit rate. Justified by permanent mix (60% of exabytes on UltraSMR by FY2027-end; 40TB ePMR then HAMR) net of the pass-through of some density gain at LTA renewal.
• Opex 10.5% of revenue. FQ3'26 actual 11.9% ($397m), FQ4 guide $385–395m on $3.65bn = 10.7%. R&D held at 6.5% (vs 8.8% today) reflects the dual-roadmap cost normalising once HAMR qualifies; SG&A at 3.5% vs 4.4% today reflects revenue scale. Deliberately set above Seagate's 9.0% because WDC's opex ratio is structurally 2.4pp higher.
• Hard constraint: m_EBIT,T (34.5%) ≤ m_gross,T (45.0%) ✓
• Sufficiency: 45.0% gross leaves 10.5pp for all R&D, selling and administration. Actual FQ4'26 guide is 10.7pp at a smaller revenue base. Feasible with headroom. ✓

terminal_margin_source: company's own same-basis quarters FQ3'25–FQ4'26E (26.0% → 40.8%), set at 34.5% — below the FY2026E full-year with three named causal reasons (HAMR lag, sub-demand LTA book, unsustainable incremental margins). No industry percentile, no peer median, no cap.

---

3. EXIT MULTIPLE — from the identity

EV_T/EBIT_T = (1−t)(1−g/ROIC)/(WACC−g)

(1 − 0.16) × (1 − 0.04/0.35) / (0.10 − 0.04) = 0.84 × 0.8857 / 0.06 = 12.4x

Warranted terminal EV/EBIT = 12.4x.

Sanity check: WDC's current EV/EBIT is 37.3x on FY2026E non-GAAP EBIT and 29.2x on the FQ4'26 guided annualised run-rate. Implied compression from today's forward multiple to the warranted exit: −57.5%. The comparator set spans 4.7x–34.9x forward P/E with no growth ordering and is declared UNIDENTIFIED for multiple-setting purposes.

---

4. IMPLIED-PATH TEST

Solved: revenue CAGR. Fixed: terminal margin 34.5%, exit EV/EBIT 12.4x, WACC 10.0%, 5 years, FCF margin 26.0%, revenue base $12,822m (FY2026E), net cash +$450m, diluted shares 385m. Terminal value ≈ 84% of EV, so the reverse DCF is mandatory and primary.

Use 33.3%. WDC's FQ3'26 FCF margin was 29.3% and it is guiding to >30%; ignoring five years of that overstates the required CAGR by 5.3pp.

The margin: demonstrated − required

This is genuinely closer than Seagate, and the reason must be stated rather than smoothed. WDC's +42.4% FY2026 revenue growth is measured off a cyclically depressed FY2025 base (FY25 continuing ops $9,006m at a 26% operating margin, coming out of a −24% trough). Recovery growth off a trough is not evidence of sustainable growth, and treating it as the demonstrated rate is precisely the "trailing-peak extrapolation" State B forbids.

Against the sustainable construction — management's own >25% exabyte CAGR plus the high-single-digit price it has guided — the required 33.3% sits at the top edge of the achievable range. Against exabyte growth alone it is −8.3pp short.

Verdict: the implied-path test is FAIL, but by a narrower and more debatable margin than Seagate's. The honest statement is: WDC's price requires exabyte growth at the top of management's guided range plus continued high-single-digit annual price increases for five consecutive years — i.e. it requires the structural thesis to be fully true, with no year of price give-back. It does not require acceleration beyond anything management has described. It requires no disappointment.

Sensitivity over the exit multiple

Holding growth, solving the multiple:

At management's own >25% growth number, the price requires a 17.8x terminal EV/EBIT against a warranted 12.4x — a 1.4x gap. That is the smallest gap in this cluster and the single strongest quantitative argument for WDC over the memory names on a five-year view.

---

5. TWELVE-MONTH TARGET

Multiple anchor: UNIDENTIFIED

Two independent disqualifications of WDC's own trading history:

1. Regime change in the entity. Pre-2025-02-21 multiples are on a half-NAND company with $7.5bn of debt. Only ~5 quarters of same-basis history exist — below any reasonable identification threshold.
2. Regime change in the industry. Realised-forward P/E ran 8.6x–17.1x across 2024–early-2025 and is 34.9x today. Current trailing P/E is at the 97.3rd percentile of the ten-year distribution, with the entire top decile created in the last twelve months.

These figures are corrected for a defect found late in this run. AV SPLITS records a 1.3230 "split" effective 2025-02-24 — this is not a stock split, it is the SanDisk spin-off price-adjustment factor. Alpaca bars fetched with adjustment=all divide all pre-2025-02-24 prices by 1.3230, while AV reportedEPS is as-reported and unadjusted. Pairing the two understates every pre-separation multiple by 24.4%. The first-pass figures in this analysis were 6.3x–7.2x; corrected they are 8.6x–17.1x. The conclusion is unchanged (today's 34.9x is ~3x the historical level) but the numbers are now right, and this is the same class of error as the KLA 10-for-1 split pairing.

Declared UNIDENTIFIED. Reported, not used. No peer median is substituted.

Target built forward, multiple stated as an assumption

In twelve months (July 2027) the market prices FY2028.

Base multiple justification: 24.0x is (a) a 31% de-rating from today's 34.9x forward multiple, which is the correction the FY2028 comparison alone implies; (b) set equal to the multiple used for Seagate, so the cluster is internally consistent and the relative call is driven by earnings rather than by a multiple assumption; and (c) above the 100th percentile of WDC's own same-basis history and declared as a regime-change assumption.

12-month target: $480 (+5.9%).

Cycle-position cross-check: WDC's 252-day peak close was $746.23 on 2026-06-18 and it is −38.1% from that high today, with the memory names down 9.5–12.8% today alone. The base target is 35.7% below the June high. The near-zero upside is the honest output: on the same 24x FY2028 anchor, Seagate offers +12.5% and WDC +5.9%, and the difference is entirely WDC's 33% forward-P/E premium on a weaker technology position.

Sanity band against the external professional target

Explained, not tuned:

1. Staleness. WDC peaked at $746.23 on 2026-06-18 and is −38.1% from there. $655.50 is 12% below that high — a target set into the June melt-up.
2. Multiple, not earnings. $655.50 on FY2028E EPS of $20.00 is 32.8x — essentially today's 34.9x forward multiple held flat. The house uses 24.0x. The entire 26.8% gap is 8.8 turns of multiple; the two sides are not materially apart on the earnings path.
3. The house is below Street but above spot, so the B16 failure mode (every target below spot) does not apply.

The one thing the Street number does usefully flag: it prices WDC at a lower forward multiple premium than the house does relative to STX (Street: WDC 32.8x vs STX 28.4x on FY2028E, a 15% premium; house: identical 24.0x on both). If anything the house is harsher on WDC's relative position than the Street is, and the reason is explicit — the HAMR lag (Research §4, §8).

---

6. Criteria summary

No position verdict is issued. The book decides.