Theta Research
  • Home
  • Graph
  • CompaniesPro
  • Pulse
  • Pricing
  • Shipped
  • Engine

For serious investors who think in systems.

partners@theta.md

Xiaohongshu · 26469493287

ProductAtlasCompaniesPulseEventsEnginePricing
CompanyAboutHow it worksChangelog
LegalTermsPrivacyRefunds

© 2026 theta.md

Theta Research provides research and analytics for informational purposes only. Nothing on this site is investment advice.

← Back to Events
COMPLETEDTier 1earnings-single

Marvell Q1 FY2027 Earnings — Custom Silicon Pipeline Test + DSP/CPO Crossover

2026-05-27 → 2026-05-27·15 companies·12 predictions·Brier: 0.215
?MRVLdirect?AVGOcompetitive?TSMsupply-chain?NVDAsupply-chain?AMDcompetitive—AMZNsupply-chain—MSFTsupply-chain?METAsupply-chain—GOOGsupply-chain?CRDOcompetitive?ANETcompetitive?LITEsupply-chain?COHRsupply-chain?CIENsupply-chain?AMKRsupply-chain

Scenario Comparison

Bull CaseS1
20%

Q1 revenue ≥$2.50B (4%+ beat), Q2 FY27 guide ≥$2.65B (≥10% QoQ), non-GAAP GM ≥61%. Custom silicon revenue called out as 25%+ of DC, with named Maia + Arke contribution disclosed. Management explicitly addresses Trainium 3 status as 'on track' with concrete program detail.

↑ 8↓ 211 companies
+MRVLmoderate
−AVGOmild
+AMZNmild
+MSFTmild
+7 more
Base CaseS2
45%

Q1 revenue $2.40-2.50B (in-line to slight beat), Q2 FY27 guide $2.50-2.60B (4-8% QoQ). FY27 full-year guide reaffirmed at >$11B. Custom silicon commentary positive but no new design win disclosed. Trainium 3 referenced with continued 'working with AWS on multiple programs' language. CPO discussion remains defensive ('we participate via 3D SiPho').

↑ 1↓ 39 companies
−MRVLmoderate
+AVGOmild
−CRDOmoderate
~ANETmild
+5 more
Bear CaseS3
25%

Q1 revenue $2.35-2.40B (in-line on revenue but missed Street ~$2.44B), Q2 FY27 guide ≤$2.50B (≤5% QoQ). Management confirms 'partial Trainium 3 award' (not sole-source) OR non-GAAP GM falls to 58-59% on custom silicon mix. Management reaffirms FY27 >$11B but acceleration shifts to H2.

↑ 1↓ 710 companies
−MRVLstrong
+AVGOmoderate
−AMZNmild
−CRDOstrong
+6 more

All Scenarios

7

Positioning Suggestions

→ PRE-EARNINGS: Stock at $168 is extended (179% YoY, 57x fwd P/E, 19x P/S). Probability-weighted expected return into earnings is asymmetrically negative given setup. If holding from below, consider trimming 25-50% to lock gains; reinvestment of trimmed amount can wait for post-print clarity.

→ DO NOT add aggressively at $168 — base case 12-month target $124 implies ~26% downside even if base scenario plays out. Bull case $160 already breached.

→ If using options for asymmetric exposure: long straddle/strangle (~10-13% IV) is fair-priced but expensive; consider put credit spread $130/$120 to collect premium with bear-case cushion.

→ Pair trade idea: long AVGO / short MRVL into print as 'custom ASIC mean reversion' — AVGO at ~28x fwd vs MRVL at 57x fwd; if MRVL fades, AVGO benefits from cross-share rotation.

→ If Scenario 3-4 (miss/cut) plays out: defer adds until stock tests $128-145 zone (50DMA / base-case midpoint). DO NOT catch the knife at $150-160 — momentum traders will compound the de-rate.

→ If Scenario 1-2 (beat/in-line) plays out: hold existing position; only add on confirmed Maia + Arke revenue disclosure (Scenarios 5-7 acting as accelerants).

→ Watch list reads: LITE/COHR/CRDO as smaller-cap proxies — they typically move 1.5-2x MRVL's % move on AI infra catalysts. Use as 'beta amplifier' or 'tape risk' positions depending on direction.

→ Hedge against MRVL/NVDA same-week paired risk: MRVL prints May 27 AC, NVDA prints May 28 AM (~16 hours later). If MRVL beats but NVDA disappoints, MRVL gains can reverse Thu; if MRVL misses but NVDA beats, MRVL bounce is tradable Thursday AM.

Predictions

12
✓

Marvell Q1 FY27 revenue ≥ $2.40B (i.e., meets or beats guidance midpoint).

85%P1

If below $2.40B → reduce position 50%+, expect 15%+ gap-down; do not add until $128-145 zone tested.

✗

Marvell Q1 FY27 revenue beats Street consensus (~$2.44B) by 1%+ (i.e., ≥$2.465B).

60%P2
✓

Marvell Q2 FY27 revenue guidance is ≥ $2.55B (midpoint at or above consensus).

55%P3

If <$2.50B → multi-quarter de-rate likely; expect stock to test 50DMA $128 within 5 trading days.

✗

Non-GAAP gross margin prints at or above 60.0%.

70%P4
✗

Management discloses Microsoft Maia by name as a revenue-contributing program in Q1 FY27 (vs. 'ramping in CY26' generic language).

40%P5

If named → reinforces diversification thesis; offsets in-line revenue with positive mix signal.

✗

Management explicitly addresses Trainium 3 award status with new specific detail (units, revenue, program scope) beyond Q4 FY26 'working on Trainium 3/4' language.

45%P6

If 'sole-source' explicitly stated → bull catalyst; if 'partial' or 'multi-vendor' explicitly stated → bear catalyst.

✗

Marvell announces a 4th hyperscaler custom XPU design win during the Q1 FY27 call.

15%P7

If announced → expect 5-8% additional upside on top of any base print reaction.

✗

Marvell announces a commercial customer win for its own CPO (co-packaged optics) architecture during the Q1 FY27 call.

12%P8

If announced → directly de-risks CPO displacement bear case; LITE/COHR mildly bearish read-through.

✗

Stock closes May 28, 2026 (T+1 to print) BELOW $165.00 (i.e., setup risk dominates upside on stretched valuation).

55%P9

Embeds same-week NVDA print risk — if both MRVL and NVDA underwhelm, AI-infra sympathy de-rate likely.

✓

Stock closes May 28, 2026 (T+1 to print) ABOVE $185.00 (i.e., strong beat-raise breakout).

15%P10
✓

FY27 full-year revenue guidance is reaffirmed at >$11B (vs. cut to $10.0-10.5B).

80%P11

If cut → bear-case activation; expect 25-35% downside over 1-2 weeks.

✗

Management explicitly addresses NVIDIA Spectrum-6 Photonics CPO competition on the earnings call with offensive framing (e.g., 'our CPO wins are...') rather than defensive framing.

20%P12

Key Questions

  1. Trainium 3 status: does Murphy give NUMBERS (units, revenue, program scope) or stays at 'we are working with AWS on Trainium 3/4' — the most important single read on the print.
  2. Custom silicon revenue disaggregation: how much of DC revenue is Trainium (AWS legacy) vs Maia (MSFT ramp) vs Arke (META future)? Mix delta vs Q4 FY26 is the diversification signal.
  3. Q2 FY27 guidance vs $2.55-2.65B consensus: any number ≤$2.50B is a meaningful negative even on Q1 in-line.
  4. Non-GAAP gross margin trajectory: holding 60% = thesis intact; 58-59% = custom silicon margin compression activates; <58% = bear trigger.
  5. CPO commentary tone: defensive ('we participate') vs offensive ('our CPO wins are...') signals whether MRVL has actual customer traction or is reacting to NVDA Spectrum-6.
  6. 1.6T optical DSP revenue inflection: explicit Q2/H2 FY27 1.6T DSP revenue mention (Ara in mass volume per OFC 2026)?
  7. Meta MTIA Arke timeline: any change to 'early 2027 mass deployment' messaging? Post AMD $60B deal, deferral is the under-priced risk.
  8. Microsoft Maia production update: first quarter of Maia revenue contribution disclosed? Or still 'ramping in CY26'?
  9. New design wins: does the 18-socket / 50-opportunity / $75B pipeline number grow? A 4th hyperscaler attribution would be the bullish surprise.
  10. Capital allocation: any commentary on Celestial AI integration milestones, additional M&A, or buyback acceleration?

Monitoring Checklist

0/14 checked

Post-Event Results

Brier Score0.215
correct4/12
Trade Actions3

Actual Outcomes

  • • Q1 FY27 revenue $2,417.8M (+28% YoY, +9% QoQ) — $18M above the $2.40B guide midpoint but ~1% below Street consensus ~$2.44B. [source: Marvell IR press release, May 27 2026]
  • • Non-GAAP gross margin 58.9% (GAAP 52.1%) — BELOW the 60% thesis-intact threshold; the custom-silicon-mix margin-compression bear mechanism activated (though it proved immaterial to the tape).
  • • Non-GAAP diluted EPS $0.80 vs Street ~$0.75 (beat).
  • • Q2 FY27 guidance $2,700M ±5% (non-GAAP EPS $0.93 ±$0.05) — well above ~$2.55B consensus and above the Scenario-1 ≥$2.65B threshold.
  • • FY27 full-year outlook raised to ~$11.5B (~40% YoY); FY28 target raised to $16.5B (~45% YoY). [earnings call]
  • • Data center revenue $1,832.7M (76% of total, +27% YoY); management guided FY27 DC revenue +~50% YoY and interconnect +>70%.
  • • Call specifics: Amazon Trainium 3 NOT addressed with new specific detail; Microsoft Maia NOT named; NO 4th-hyperscaler XPU win announced ('won several new designs… no additional details at this time'); NO commercial win for Marvell's own CPO disclosed. [earnings call transcript]
  • • CPO/NPO framing stayed DEFENSIVE / multi-standard ('we will pivot at the right time… follow our customers… follow the market') — no offensive Spectrum-6 rebuttal. Marvell announced the Polariton acquisition (plasmonic modulator, >1 THz bandwidth) to bolster its silicon-photonics roadmap.
  • • Stock closed $204.83 on May 28 (T+1, +3.1% on the day). Decisively, the shares had ALREADY re-rated from ~$169 (May 19) to ~$199 INTO the print, so the beat-and-raise held/extended an already-extended setup rather than triggering the predicted sell-the-news fade. Raymond James lifted PT to $235 (from $105). A second, exogenous leg followed ~June 1: Jensen Huang called Marvell a 'next trillion-dollar company' at Computex 2026.

Market Reaction

MRVL+3.1%

Trade Recommendations

→ Positioning call was directionally WRONG: the thesis bet on 'stretched multiple → sell-the-news at $168' (P9 conf 0.55 sub-$165; P10 only 0.15 above-$185). Stock closed $204.83. Lesson: a stretched multiple is not a catalyst — a strong beat-raise on an accelerating AI-infra name can extend an already-extended setup.

→ Fundamental forecasts were sound: Q1 rev ≥$2.40B (✓), Q2 guide ≥$2.55B (✓), FY27 >$11B (✓). The one fundamental miss — non-GAAP GM <60% (58.9%) — was the correct bear mechanism (custom-silicon mix dilution) but immaterial to the stock.

→ No actionable new trade from this scoring: MRVL thesis already current (Hold, lastUpdated 2026-06-04). Event scored 9 days late; read-throughs already absorbed by the pipeline.