Verdict: KEEP

WDC is 0.55% of the selectable universe. Factor profile is below threshold — 57.2% idio variance, β_SPY = 1.85, MTUM beta +1.61 — but every factor loading runs WITH QQQ, not against it. Pro-momentum in a momentum index is structural alignment. The confirmed removes (MAR -0.57, CSX -0.46, SBUX -0.45) all had anti-momentum signals. WDC is the opposite.

Fundamentals are among the strongest in the selectable universe: revenue accelerating (+40% YoY guided), gross margins expanding (46.1% → 47-48%), sold out through CY2026 with multi-year LTAs to CY2028, deleveraging at escape velocity ($1.5B in notes redeemed in February alone). Generating $653M quarterly FCF at 21.6% margin while buying back $615M/quarter.

Earnings May 6 inside the window. Four consecutive beats. Street expects $2.35 vs company guide of $2.30. Setup favors another beat on 40% YoY revenue growth with 75% incremental margins.

No actionable mispricing identified in either direction. The central tension — structural vs. cyclical — is well-understood by both sides. 26 analysts, $94.6B cap, consensus is roughly right. At 0.55% weight, max plausible drag from a wrong keep is 11 bps. Cannot prove structural underperformance. Keep.

Factor Profile

250-day regression against SPY + MTUM + XLK:

Alpha (ann, trailing): +173.2%. Idio vol: 49.6%. Total vol: 65.6%. R²: 42.8%. n=251.

Critical caveat on trailing α: The +173.2% is backward-looking during a +562% 1Y run. Not a forward alpha estimate. The regression is measuring a historic inflection, not steady-state alpha generation. Forward α depends on whether the AI/HDD demand thesis sustains through the window.

Comparison to confirmed removes:

The idio% is comparable to removes, but the direction of factor exposure is reversed. Our factor-based removes share a pattern: they load AGAINST the QQQ momentum factor. WDC loads WITH it. In a momentum index, that's the difference between drag and thrust.

The Business: HDD Pure-Play Post-Separation

WDC completed the separation of its flash/NAND business (now Sandisk/SNDK) in February 2025. What remains is a pure HDD business — specifically, nearline enterprise drives for hyperscale data centers. Cloud = 89% of revenue. Client 6%, Consumer 5%.

This is not the Western Digital of 2020. This is a storage infrastructure company riding the AI inference wave, with a customer base of 7 hyperscalers who have signed multi-year purchase commitments.

Income Statement

All Q2 results above high-end guidance. Four consecutive quarterly beats: +22.7%, +12.1%, +12.9%, +10.5%.

Margin mechanics: CFO Sennesael confirmed 75% incremental gross margin. Two drivers: (1) ASP/TB up 2-3% YoY — pricing power in tight supply, not commodity pass-through; (2) Cost/TB down ≈10% YoY — manufacturing efficiency plus UltraSMR mix shift (software-based, margin-accretive). Each incremental dollar carries 75% gross margin.

Operating leverage is acute: OpEx $372M, declining as % of revenue (120bps sequentially). Q3 OpEx guided $380-390M on $3.2B revenue — OpEx grows ≈4% while revenue grows 40%.

Balance Sheet: Escape Velocity Deleveraging

Debt schedule as of January 2, 2026 → post-February 2026 redemptions:

Three 8-Ks in February — all balance sheet cleanup. Interest savings from redemptions: ≈$53.5M annually (≈$0.03/quarter EPS accretion), eliminating roughly 25% of run-rate interest cost.

Convertibles are effectively equity. Conversion price $37.74 vs stock at ≈$277 = 7.3x in the money. ≈42.4M diluted shares already in count (378-385M guided). WDC can force redemption after Nov 15, 2026. These will convert, eliminating the $1.6B "debt" label. Real debt = Term Loan ≈$1.55B.

Sandisk retained interest: $2,068M on balance sheet. Already used 21.3M SNDK shares + $4M cash to retire $800M of Term Loan (June 2025, tax-free exchange). Plans to monetize remaining stake within 1 year. The $1,714M gain in H1 FY2026 is excluded from non-GAAP EPS — operational numbers are clean.

Effective balance sheet: Cash $2.0B + SNDK stake $2.07B = $4.07B liquid. Post-redemption debt ≈$3.15B (of which $1.6B converts). Real debt ≈$1.55B. ≈$2.5B net cash after SNDK monetization. Transformed from leveraged cyclical to fortress in 12 months.

Capital returns: $615M buybacks in Q2, $1.4B total since program launch. $0.125/quarter dividend. At ≈$2.6B annualized FCF, can simultaneously deleverage and return capital.

Demand Visibility: Sold Out and Locked In

Purchase commitments: Firm POs with top 7 customers through CY2026. Three of top 5 signed longer-term agreements — two through CY2027, one through CY2028.

Volume ramp: Shipped 3.5M+ drives in Q2 (215 exabytes, +22% YoY), ramping to ≈4M in Q3. Capacity is the constraint, not demand.

NAND structural shift (cross-ticker confirmed): SNDK CEO (LR 4.5): NAND market transitioning from "quarterly auction" to multi-year supply agreements with data centers. MU, STX, LRCX all confirm tight supply through 2026+. Industry-wide shift from commodity to strategic infrastructure.

Technology roadmap:

• UltraSMR: >50% nearline mix. Top 3 customers fully adopted, 2-3 more in process. Software-based, margin-accretive, zero incremental cost.
• HAMR: Qualification pulled forward 6 months. Started with 1 hyperscaler Jan 2026, 2nd "imminent." Innovation Day (Feb 3): roadmap to 100+ TB by 2029. $73M Thailand R&D investment March 2026.
• ePMR yields: "Low 90s percentage" — no manufacturing issues.

The AI inference thesis: CEO Tan: "As the AI value changes from model training to inference, more data is going to get created... more data needs to get stored... a lot of that data will be stored on hard drives." Confirmed by customer purchasing behavior — multi-year LTAs with volume commitments.

Market Consensus

Sell-Side

4 Strong Buy, 16 Buy, 6 Hold, 0 Sell. 77% bullish, 0% bearish.

Mean $321 (+16%), median $325, range $170-$440. The $270 spread on a $277 stock tells you: unanimous on direction, wide disagreement on magnitude.

Goldman at Neutral/$250 is the tell. Raised from $220 on Feb 4 but stayed Neutral. They see the cycle but won't chase +563% 1Y. This is the "great quarter, closer to peak than trough" view. The 9.2% short interest says real money agrees with Goldman.

Options Market

Someone is buying downside protection, not upside participation. Consistent with long holders hedging a high-beta name into earnings, not informed selling.

Insider Behavior

$7.4M in officer selling in March (Gubbi $4.9M, Tregillis $2.5M) vs ≈$100K in plan-based CEO acquires (49-92 shares — payroll, not conviction). Nobody with material information is accumulating with their own money. Rational monetization after +563%, not alarming — but not confirming either. 0.008% of market cap.

Short Interest

9.2% of float, 3.4 days to cover. Non-trivial. Likely structural short from pairs/sector-neutral funds (high-beta momo HDD name is a natural pair against NAND longs). Enough shorts to fuel a squeeze on a beat, enough to add pressure on a miss.

Mispricing Assessment

The central tension is structural vs. cyclical. Both sides are well-represented, well-funded, and well-informed.

If the structural thesis is correct (consensus view) → 20x forward is cheap

Multi-year LTAs to CY2028, NAND transitioning from "quarterly auction" to strategic supply agreements, AI-driven data center demand making HDD mission-critical. Secular growers with contracted revenue and margin expansion get 30-40x. WDC at 20x forward on $2.35 est (still accelerating at +40% YoY) would be undervalued by 30-50%. But this IS the consensus — 20/26 analysts are Buy. Not a novel insight.

If the cycle peaks in CY2027 (Goldman/shorts view) → $277 is 15-25% rich

Trailing EPS trajectory: $1.36 → $1.66 → $1.78 → $2.13 → $2.35 est. ≈$8 trailing, maybe $10-12 forward annualized. At $277, that's 23-28x what might be peak cycle earnings. MU traded at 5-8x at its last peak. WDC gets a structural premium — but if NAND supply normalizes in 2027, you're paying 30x+ mid-cycle earnings for a memory company.

The market is splitting the difference — correctly

WDC trades above historical peak memory multiples (structural premium) but below secular growth multiples (cycle discount). Forward P/E 20.58x is between the 5-8x cyclical peak and 30-40x secular growth endpoints. That's a roughly efficient price given genuine uncertainty about regime.

No actionable mispricing identified. Both sides have merit, both sides have real money behind them. Our P ≈ market's P. Edge = 0.

Tariff Exposure

Manufacturing in Thailand (HDD assembly) and Malaysia. 10-Q: "business and results not materially impacted through Q2 2026."

Actions over words: $73M HAMR R&D investment in Thailand (March 2026). You don't invest $73M in a jurisdiction you're about to flee. Seagate also manufactures in Thailand — industry-wide event, not WDC-specific. Broad tariff escalation would hit entire QQQ tech supply chain.

Assessment: Low probability of WDC-specific tariff damage. Industry-wide escalation is a factor bet, not a filtration signal.

Counterparty Analysis

If we removed WDC and it outperformed, our counterparty would be momentum and tech-sector longs buying a company with 40% YoY revenue growth, 75% incremental margins, sold out through CY2026, and $2.5B net cash. That's an informed, structurally advantaged buyer with 26 analysts and $94.6B of institutional attention. Our edge over that buyer: zero.

If we kept WDC and it underperformed, our counterparty would be Goldman-aligned cycle skeptics and pairs traders who correctly called peak earnings on a stock up 563% in a year. That's also an informed buyer. But at 0.55% weight, max drag is 11 bps.

The asymmetry of errors favors keeping: wrong-keep costs 11 bps max, wrong-remove forfeits participation in a name with Q3 beat setup, HAMR catalyst optionality, and momentum alignment with the index.

Catalyst Calendar (15-Week Window: Mar 30 – Jul 13)

What Would Change the Verdict

• Q3 earnings miss or guide-down — first in 5+ quarters would signal demand cracking → WATCH
• Tariff escalation on Thai electronics — direct hit to manufacturing cost structure → WATCH
• NAND supply crunch reversal — would collapse pricing power industry-wide → FILTER
• HAMR qualification failure — delays capacity roadmap, disappoints hyperscaler timelines → WATCH
• Momentum factor crash — β_MTUM +1.61 means disproportionate pain, but this would affect entire QQQ, not just WDC → monitor, not remove

None present in current data.

Sources: WDC 10-Q (SEC, 2026-01-30) — primary [Tier 1]. WDC Q2 FY2026 earnings call (2026-01-29) [Tier 2]. WDC 8-Ks: 2026-02-09, 2026-02-18, 2026-02-24 [Tier 1]. Factor regression: 251d trailing SPY + MTUM + XLK [Quant]. yfinance market data (2026-03-27) [Tier 2]. Options chain (yfinance, 2026-03-27) [Tier 2]. Worldview evidence: (SNDK NAND structural shift, LR 4.5), (WDC Q2 beat, LR 1.8), (CEO structural shift framing, LR 1.5), (SNDK guide vs street, LR 3.0) [Tier 2]. Innovation Day (Feb 3, 2026): Morgan Stanley PT $369 [Tier 3]. X/social: Thailand HAMR R&D investment [Tier 3-4]. Cross-ticker NAND supply confirmation: MU, STX, LRCX transcripts [Tier 2].