Western Digital Corp
Western Digital's revenue drift baseline projects $17.31B across the next four quarters (+34.0% versus the preceding four quarters). It is refitted through 2026-07-03 using all 12 available contiguous revenue quarters. R&D is now extracted correctly but was not selected: expense predictors did not reliably improve on the strongest simple baseline.
Revenue forecast
| Quarter | Forecast | 95% model PI | YoY growth |
|---|---|---|---|
| Q1 FY2027 Projected end 2026-10-02 | $3.98B | $3.65B–$4.31B | +41.2% |
| Q2 FY2027 Projected end 2027-01-01 | $4.21B | $3.73B–$4.69B | +39.6% |
| Q3 FY2027 Projected end 2027-04-02 | $4.44B | $3.83B–$5.05B | +33.2% |
| Q4 FY2027 Projected end 2027-07-02 | $4.68B | $3.95B–$5.41B | +24.8% |
Separate management guidance: WDC guided Q1 FY2027 revenue to $4.0B–$4.2B on 2026-08-05. Company earnings release.
95% model prediction intervals assume independent, normally distributed quarterly revenue changes with constant mean and variance. They include uncertainty in estimated drift and widen with forecast horizon. They are not empirically calibrated by four backtests and omit structural breaks, demand shocks and model-selection uncertainty. Projected period ends assume 13-week quarters. Management guidance is shown separately and was not used to fit or select the model.
Why the simple baseline was retained
R&D extraction now includes the issuer's alternate SEC tag. The first two expense observations predate the Sandisk separation and do not match restated continuing-operations revenue; they are excluded from expense-model fits. All 12 revenue quarters remain available; expense models use 10 comparable quarters in the final sample. The best expense candidate has a 2.31% average error versus 2.32% for drift across four tests, but its selection-period error is 1.37% versus 1.35% for drift. It fails the prespecified requirement to beat every baseline in both selection and confirmation. That tiny retrospective average advantage is insufficient to select the extra predictors.
Choose the lowest selection-period MAPE. An expense candidate must beat every simple baseline in both selection and confirmation; otherwise retain the best selection-period baseline.
| Model | Q1–Q3 selection MAPE | Q4 confirmation error | All four MAPE |
|---|---|---|---|
| Revenue drift · Selected Baseline | 1.35% | 5.22% | 2.32% |
| Revenue trend + R&D + SG&A Candidate | 1.37% | 5.12% | 2.31% |
| Log trend + R&D + SG&A Candidate | 3.13% | 0.11% | 2.37% |
| Equal-weight global/local trend ensemble Candidate | 1.90% | 6.52% | 3.05% |
| Linear revenue trend Baseline | 2.06% | 7.59% | 3.45% |
| Last-four-quarter linear trend Baseline | 3.73% | 5.44% | 4.16% |
| Revenue trend + lagged R&D Candidate | 3.17% | 8.71% | 4.56% |
| Log revenue trend Candidate | 9.02% | 1.81% | 7.22% |
| Previous quarter Baseline | 7.91% | 10.94% | 8.67% |
| Same quarter last year Baseline | 24.30% | 30.48% | 25.85% |
Each test fits only preceding observations. Expense candidates hold future spending at its latest known level; they are not given future actual spending. Every candidate evaluated is shown above.
R&D extraction corrected
WDC reports R&D separately; the previous report incorrectly said otherwise. The corrected history includes FY2026 R&D of $1161M, including $284M in Q4; FY2026 SG&A totals $551M.
The expense comparison uses 10 quarters with post-separation filing provenance. Two older expense observations still include the former Flash business and are excluded from expense-model estimation. The corrected disclosure does not imply that R&D or SG&A caused revenue growth.