Verdict: KEEP

QQQ weight 0.29%. Cumulative LR 0.11 (strongly bearish fundamentals). Doesn't clear the bar — maximum alpha 6-9 bps, and 67% of variance is semiconductor sector, not idiosyncratic. Removing MPWR is a short-SMH bet at beta 1.33, not a stock bet. Would remove at 0.8%+ weight.

Factor Decomposition

6-month daily regression vs SPY + SMH:

FactorBetap-value% Variance
SMH1.33<0.00167.3%
SPY-0.50n.s.—
Idiosyncratic——32.7%

R² = 0.673. Annualized alpha = -5.0% (not statistically significant). Idiosyncratic variance 32.7% — well below 75% target. This stock moves with semiconductors, not on its own fundamentals.

Trailing excess return vs QQQ: +25.1% (6mo), +21.3% (3mo), -14.5% (1mo reversal).

Bearish Evidence (7 signals, cumulative LR 0.11)

1. Enterprise Data Structural Decline (LR 0.5)

Largest end market declined -2% YoY ($708M → $694M) while overall revenue grew +26%. Enterprise Data fell from 32.5% to 25.2% of revenue as Storage & Computing overtook it at 26.3%. Every other segment grew 26-46%. CFO Blegen on Q4 call: "largest revenue end market enterprise data. 2% yet overall grew 26%." Source: 10-K lines 297-303, Q4 2025 transcript.

2. Inventory Overbuilding + Double-Ordering Risk (LR 0.4)

Total inventory $565M, up 35% YoY from $420M. Finished goods led at $236M (+50% from $158M), WIP $220M (+30%), raw materials $108M (+17%). Inventory cash outflow was -$145M (2025) vs -$35M (2024) — 4x increase. Management explicitly warned: "high-level issue figure out real demand, double ordering on part customers as secure capacity" — citing 2020/2021 semiconductor shortage precedent. If double-ordering unwinds, writedown risk materializes and the low-inventory-write-down tailwind to margins reverses. Source: 10-K Note 6 lines 4183-4186, Q4 2025 transcript.

3. Insider Selling — $220M+ in One Week (LR 0.65)

All four named officers sold simultaneously on Feb 6, 2026: CEO Hsing ($141M across Feb 6-12), CFO Blegen ($25.4M), SVP Sciammas ($26.8M), GC Tseng ($26.8M). All under Rule 10b5-1 plans adopted Nov 13-28, 2025 (Sciammas/Tseng/Xiao) — CEO plan adopted prior quarter. Restatement filed Feb 27, three weeks after the selling cluster. 10b5-1 structure reduces the signal vs discretionary selling, but adoption timing (90 days before the sell window, shortly before restatement became certain) is notable. Source: 10-K Item 9B lines 5539-5571, Form 4 filings.

4. Material Weakness + CFO Departure (LR 0.6)

Ernst & Young issued adverse ICFR opinion for BOTH Dec 31, 2024 and Dec 31, 2025. Scope is narrow — deferred income tax accounting only, no impact on operating results or cash flows. Remediation expected during 2026. CFO Blegen departed immediately upon filing. Narrow scope limits severity, but adverse opinion spanning two fiscal years plus immediate CFO exit is a governance signal the market may be underweighting. Source: 10-K lines 5470-5518.

5. China Revenue Concentration — Increasing (LR 0.6)

China: $1,544M (55.3% of 2025 revenue), up from 53.3% (2024) and 51.3% (2023). Trend is toward MORE China dependence, not less. Asia total: 92.4%. Long-lived assets in China: $333M (53% of total). New risk factor language added in 2025 10-K: AI-specific export control risk for power management components (lines 1070-1074). The market prices China risk as binary (tariff on/off) but the trend line shows structural dependence deepening annually. Source: 10-K Note 3 lines 3966-3974.

6. Gross Margin Compression — 3 Consecutive Years (LR 0.7)

56.1% (2023) → 55.3% (2024) → 55.2% (2025). Management attributes 2025 decline to "higher warranty expenses" offset by "lower inventory write-downs." That offset is fragile — if the inventory buildup (#2 above) leads to excess, the write-down tailwind reverses and margins compress faster. Management guides "10 to 20 basis points quarter over quarter" improvement, which requires the inventory situation to resolve cleanly. Source: 10-K lines 2905-2916, Q4 transcript.

7. Securities Litigation (LR 0.7)

Class action: Waterford Twp. v. Monolithic Power Systems (W.D. Wash. No. 25-cv-220). Derivative: Miller v. Hsing (No. 25-cv-527) — stayed pending securities action resolution. Standard restatement follow-on. Management calls it "meritless." Likely settles for immaterial amount, but creates headline risk and management distraction through 2026-2027. Source: 10-K lines 2552-2567.

Market Consensus vs Evidence

The equity market and options market disagree:

MarketSignalImplied P(bear)
Sell-side (16 analysts)15 Buy, 1 Hold, 0 Sell. Median PT $1,350 (+26%)≈6%
Options (Apr 17 exp)P/C ratio 1.41, put IV premium +3% over calls, OTM put skew +10.1%≈25-30%
Short sellers4.7% SI, 4.2 days to cover≈15%
Our evidence (LR 0.11)7 bearish signals, primary-source verified≈90%

Forward P/E 40.7x implies earnings roughly doubling from trailing ≈$17 to ≈$26. That requires Enterprise Data recovery, sustained Storage & Computing growth without double-ordering unwind, and margin expansion — all while navigating CFO transition, material weakness remediation, and deepening China dependence.

The potential mispricing is in growth durability. The options market partially validates this (bearish positioning, elevated put skew). But at $51.7B cap with 16 analysts, every bear signal is public. Our disagreement with consensus is about probability weighting of known information, not information asymmetry.

Why KEEP

  1. Weight kills it. 0.29% x 20% underperformance = 5.8 bps. Even 30% underperformance = 8.7 bps. Not worth a remove slot when average remove weight is 0.51%.
  2. 67% semiconductor sector. Removing MPWR is short SMH at beta 1.33. No semiconductor sector view in this basket.
  3. No counterparty edge. $51.7B cap, 16 analysts, all evidence is public 10-K/Form 4/PACER. Counterparty is informed and disagrees. Edge = weighting difference on public info, not information asymmetry.
  4. Trailing momentum against us. +25.1% excess vs QQQ 6-month. Recent 1-month reversal (-14.5%) but shorting into a momentum name with 67% sector variance is a sector timing bet.
  5. Remove slot opportunity cost. 10 slots at ≈0.51% avg weight = ≈5.1% total short. Spending one on 0.29% is -43% below average slot value.

Conditional Thesis (Would Remove at 0.8%+)

If MPWR weight were 0.8%+, the remove case is: Enterprise Data structural share loss to TI/ADI during AI boom + inventory overbuilding with management-acknowledged double-ordering risk + governance overhang (restatement + adverse ICFR + CFO departure + litigation) + deepening China/AI export control exposure. Factor type: DEMAND, 180-365d half-life. Not urgent catalyst — would wait for price.

Watch Triggers

  • Apr 30 Q1 earnings: Enterprise Data trend, inventory writedowns, margin guidance, new CFO commentary. Beat cadence narrowing (+0.9% to +2.2% last 4Q). Options pricing ≈10% implied move.
  • Weight drift: Mathematically unlikely to reach 0.8% within 15-week window absent a major reweighting.
  • Export controls: Any new BIS rules targeting AI power management components would crystallize the China risk.