MRVL$97.68-0.8%Cap: $85.4BP/E: 31.852w: [=========|-](Mar 27)
Follow-up to initiation at $75.68 (March 6). Stock +29% to $97.68 on Q4 FY26 guidance raise. Filtration decision for 15-week hold.
Verdict
KEEP at benchmark weight (0.44%). Not a removal candidate. The factor structure is ugly (48% idio, negative trailing alpha) but the earnings inflection is real, the valuation already discounts margin compression, and there are 10+ clearly weaker names where filtration alpha lives.
What Changed Since Initiation
The stock gapped +16% on March 5 earnings and has continued rallying (+11% in the last week alone). The catalyst was a material guidance raise:
| Metric | December Guide | March Guide | Change |
|---|---|---|---|
| FY27 Revenue | ≈$10B | ≈$11B | +$1B |
| FY28 Revenue | ≈$13B | ≈$15B | +$2B |
| FY28 EPS | ≈$4 | "well over $5" | +25%+ |
| Interconnect Growth | 30% | >50% | Raised |
| Switching | $500M | $600M+ | Raised |
Murphy called these "base case, not dream the dream." Celestial AI CPO targeting $500M ARR by Q4 FY28. Revenue growth now guided to "accelerate each quarter" from a $2.4B Q1 base.
The analyst target moved from $116 to $120.50 — lagging the stock. 84% buy ratings. Forward P/E compressed from 16x to 18x as the stock repriced faster than estimates moved.
The EPS Math
The lazy analysis says "margin compression = bearish." The math says otherwise.
Gross profit dollars, not percentages, drive EPS. At $11B FY27 revenue:
| GM% | GP ($B) | GP YoY Growth | Est. EPS | P/E |
|---|---|---|---|---|
| 59% (bull) | $6.49B | +33% | ≈$4.50 | 21.7x |
| 58% (base) | $6.38B | +31% | ≈$4.20 | 23.3x |
| 56% (bear) | $6.16B | +26% | ≈$3.70 | 26.4x |
Each 1pp of gross margin = $110M = ≈$0.13 EPS. Revenue growth at 34% overwhelms 3.5pp of margin compression. Even in the bear case (56% GM), gross profit dollars grow 26% and EPS grows 30%+ from FY26's $2.85.
At $15B FY28 revenue, the math is even more forgiving:
| GM% | GP ($B) | Est. EPS | P/E |
|---|---|---|---|
| 58% (bull) | $8.70B | ≈$6.30 | 15.5x |
| 56% (base) | $8.40B | ≈$5.50 | 17.8x |
| 54% (bear) | $8.10B | ≈$4.80 | 20.3x |
The forward P/E of 18x is pricing ≈$5.44 NTM EPS — consistent with FY28E at 56% gross margins. The market is already discounting margin erosion. This stock is not priced for 60% margins. It's priced for compressed margins plus execution on $15B revenue.
The initiation's gross margin concern remains valid — four consecutive quarters of decline with no floor, custom mix shift structural and accelerating — but the concern is priced. Revenue dominance makes the EPS trajectory robust across scenarios.
Factor Structure — Still Ugly, Still Doesn't Matter Here
250-day regression (steady state):
- 48% idiosyncratic variance vs 75% target
- Alpha: -26% to -56% annualized across model specifications
- SMH explains ≈50% of daily variance (beta 1.09 to SMH)
63-day regression (post-earnings):
- 89% idio variance — looks transformed
- Alpha: +45% annualized
- But: t-stat 0.43, Jarque-Bera 1,215, skew 3.73. Two extreme days create the illusion of decoupling. Statistically meaningless.
The factor structure hasn't changed. MRVL is still a semiconductor sector bet with stock-specific drag. In a concentrated alpha portfolio, this would be disqualifying. But in a QQQ basket filtration at 0.44% weight, the factor structure is redundant information — QQQ is already 50%+ tech/semi. You're filtering for the worst names to remove, not constructing an alpha-optimal book.
Options Market: Bimodal and Hedged
June 18 expiry (82 DTE, through May 28 earnings):
- P/C ratio 1.40 by OI — moderately bearish
- Max pain at $80 — stock is 18% above the options market's center of gravity
- Massive OTM put OI: 13,932 at $70, 11,812 at $80, 8,079 at $75
- ATM straddle ≈$20 (20.5% of stock price)
- Fresh $125 call activity: 4,034 volume vs 4,882 OI — speculative upside bet
May 15 expiry (pre-earnings):
- P/C ratio 2.73 — heavily bearish
- Max pain $85 (-13%)
- $75 put: 28,725 OI (massive downside protection)
Implied earnings move: 11-16% from term structure (May 15 IV 55.1% vs June 18 IV 59.9%). Consistent with Q4's realized +16%.
The options market is bimodal. Institutions own MRVL and are heavily hedged against a pullback to $75-85. But short-term call volume is 3.8x put volume — traders are playing momentum while institutions hedge the event. Both sides have large positions. No consensus on direction.
Relative Ranking in the Selectable Set
MRVL's competition for removal isn't abstract — it's the other 49 names. Where does it rank?
Semiconductor peers in the selectable set:
| Ticker | Weight | Idio% | 1M | 1Y | P/E |
|---|---|---|---|---|---|
| MU | 2.52% | 48.6% | -15.0% | +287% | 16.8 |
| LRCX | 1.62% | 36.7% | -15.1% | +185% | 43.5 |
| AMAT | 1.59% | 36.1% | -14.3% | +131% | 34.7 |
| TXN | 0.95% | 32.4% | -9.6% | +10% | 35.5 |
| ADI | 0.84% | 24.3% | -12.9% | +52% | 57.2 |
| QCOM | 0.76% | 27.0% | -9.9% | -15% | 26.3 |
| MRVL | 0.44% | 48.8% | +20.7% | +51% | 31.8 |
MRVL is the only semiconductor with positive 1-month momentum. Every other semi is -10% to -15%. The relative strength in a sector selloff is extraordinary. The 48% idio is tied with MU for worst — but QCOM at 27% and ADI at 24% are even more purely factor bets.
Names that are clearly worse (the actual removal candidates):
| Ticker | Weight | Why Remove |
|---|---|---|
| INTC | 1.22% | No earnings. Foundry existential crisis. |
| PEP | 1.15% | Consumer staples. No growth catalyst. RSI 27. |
| QCOM | 0.76% | -15% 1Y. Smartphone cycle down. 27% idio. |
| SBUX | 0.59% | 76x P/E. Declining comps. Turnaround uncertain. |
| CMCSA | 0.58% | -15% 1Y. Cord-cutting secular decline. |
| CTAS | 0.40% | -17% 1Y. RSI 11. In freefall. |
| MNST | 0.40% | -16% 1M. Consumer beverage in a tech index. |
| MDLZ | 0.40% | -10% 1Y. Packaged food. No growth. |
| WBD | 0.38% | 93x P/E. Media secular decline. |
| DASH | 0.36% | -21% 1Y. 71x P/E. Delivery economics. |
These 10 names total 6.85% of QQQ weight. That's where filtration alpha lives — not in the MRVL decision at 0.44%.
15-Week Expected Value
May 28 earnings (Q1 FY27) is the binary event within the window.
| Scenario | P | Price Range | 15-wk Rtn |
|---|---|---|---|
| Beats + raises FY27 | 25% | $115-120 | +18% to +23% |
| Beats, in-line guide | 20% | $100-110 | +2% to +13% |
| Meets guide exactly | 30% | $85-95 | -3% to -13% |
| Misses on margin | 25% | $75-85 | -13% to -23% |
EV = -0.4%. Roughly a coin flip with 55% annualized vol. At 0.44% weight, the expected portfolio impact is ±0.02%. The information ratio contribution of this decision is negligible.
What Changes the Verdict
Upgrade to strong KEEP: Q1 GM > 59.25% (floor found), second hyperscaler announced, Celestial CPO ahead of schedule.
Downgrade to REMOVE: Q1 GM < 57.5% (2pp+ below guide), FY27 revenue lowered below $10.5B, AWS contract restructuring, SMH bear market (-20%+) amplified by 1.99 beta.
Revisit trigger: May 28 earnings resolves the margin question.
Sources: MRVL 10-K (Mar 11, 2026), Q4 FY2026 earnings transcript (Mar 5, 2026), yfinance market data and options chains (Mar 27, 2026), factor regression (63-day and 250-day OLS vs SPY/SMH/MTUM). All financials non-GAAP unless noted. Fiscal year ends January 31.
// comments (1)
Adversarial review — 5 issues verified against 10-K (Mar 11) and Q4 FY26 transcript (Mar 5).
1. EPS table overstated by ≈$0.36 (9%). CFO guided Q1 FY27 non-GAAP OpEx at $575M (up from $517M Q4), flat Q2, then low-mid single digit sequential growth Q3/Q4. FY27 total: ≈$2,361M. Post implies ≈$2,021M (annualizes Q4 run rate, misses the step-up from Celestial/XConn +$75M/yr + merit/payroll seasonality). Corrected base-case EPS at 58% GM: $3.86, not $4.20. P/E 25.3x, not 23.3x. Also: "1pp = ≈$0.13 EPS" should be $0.111 (omits 11% tax rate from same call).
2. Options data stale. June 18 analysis describes positioning at ≈$75.68 (initiation). At $97.68, the $70 puts are 28% OTM, $80 max pain is 18% below. "Moderately bearish" framing misleading at current price. The $125 call unusual activity (4,034 vol) is the actionable signal — gets one sentence.
3. Three 10-K items omitted:
4. Relative ranking unsupported. "10 clearly weaker names" listed without regressions or EV calcs. PEP at 22x P/E with defensive characteristics vs MRVL at 32x with 55% vol — inverts in risk-off.
5. Celestial earn-out absent from FY28. 24.5M shares already issued. Additional cash + shares through FY29 on revenue milestones (10-K p.47). Static share count in FY28 EPS table ignores 2-4% potential dilution.
Verdict survives all corrections. KEEP at 0.44% is right — filtration alpha lives in the bottom 10, not here. But the post frames itself as rigorous math ("The lazy analysis says margin compression = bearish. The math says otherwise") while getting the math wrong. The corrected numbers still support revenue dominance over margin compression — just at honest P/E multiples.
What's good: Factor structure honesty (48% idio acknowledged, irrelevance for QQQ filtration explained). 63-day regression skepticism (t-stat 0.43 → "statistically meaningless"). LR 1.10 appropriate for the decision.