← Sazabi Research

Marvell Technology (MRVL)

The rack around the accelerator — custom AI silicon and 800G/1.6T connectivity — Quick Take

📅 October 7, 2026 TechnologySemiconductorsAI InfrastructureNASDAQ
Market Cap
~$258B
Revenue (TTM)
$9.45B
P/E (Fwd)
~51x
Rev Growth
+31%

Why This Company Now

Technology was the dominant sector in the Wed 7 Oct US VolumeSurge (7 of 32 passing names, ahead of Financial Services 6 and Industrials 5), and MRVL was the #1 name by turnover at $14.7B — roughly 2.8x the next largest passing name (NBIS at $5.3B) — closing +5.81% at $287.01 on 51.2M shares. The company-specific catalyst is the Oct 6 Investor Day: management raised its FY2028 revenue target to approximately $20B (from a prior ~$18B outlook and above the ~$18.2B Street consensus), implying about 67% year-over-year growth, and for the first time put a FY2031 revenue target of $70B–$90B on the wall against last year's ~$8.2B. FY2027 is still framed around ~$12B. The shares spiked roughly 8–9.5% intraday (adding ~$23B of market value in a single session) before settling at the $287 close. The HK VolumeSurge passed only 1 of 2,817 names (0178.HK, $8M turnover — far below the $15B floor), so no HK signal applied.

Financial Snapshot

Fiscal YearRevenueRev GrowthOp MarginNet Income
FY2022 (Jan '22)$4,462M+50.3%-1.3%-$421M
FY2023 (Jan '23)$5,920M+32.7%6.7%-$164M
FY2024 (Feb '24)$5,508M-7.0%-7.9%-$933M
FY2025 (Feb '25)$5,767M+4.7%-0.2%-$885M
FY2026 (Jan '26)$8,195M+42.1%16.3%$2,670M
TTM (Aug '26)$9,450M+30.6%16.8%$2,640M

Fiscal year ends late January/early February. FY2026 was Marvell's first meaningfully GAAP-profitable year (net income $2.67B, +$3.56B swing from FY2025) after a brutal FY2024–25 trough driven by the storage/legacy-business reset. TTM gross margin 52.2%, operating margin 16.8%, free cash flow $1.73B (18.3% of revenue); cash & investments $3.93B against total debt $5.29B — net debt down to ~$1.35B from ~$4.1B at FY2022, a marked deleveraging inside the AI buildout.

Q2 FY2027 (quarter ended Aug 1 2026, reported 27 Aug): record revenue $2.739B, +37% YoY ($39M above the guidance midpoint) with Data Center revenue growth accelerating to +46% YoY; GAAP gross margin 53.1% / non-GAAP 58.9%; GAAP net income $308.0M ($0.33 diluted EPS) and non-GAAP net income $865.9M ($0.94 diluted EPS, vs ~$0.87 consensus); operating cash flow $605.5M. Q3 FY2027 guidance: revenue $3.150B ±5%, non-GAAP gross margin 57.5–58.5%, non-GAAP diluted EPS $1.10 ±$0.05 — i.e., sequential revenue growth of ~15% and gross margin guided down ~40bps at the midpoint. Quarterly dividend $0.06 (ex-date Oct 9 2026).

Bull vs Bear

✅ Bull Case

  • The ceiling keeps moving. The FY2028 target has been lifted ~50% in roughly ten months (~$13B → ~$15B → $16.5B → $18B → ~$20B), and the newly minted FY2031 band of $70–90B is the first multi-year number management has been willing to defend publicly. Data Center is now roughly four-fifths of revenue and grew 46% YoY in Q2.
  • It sells the toll road, not the destination. Custom ASICs for hyperscalers plus 800G/1.6T PAM4 DSPs, electro-optics and co-packaged optics — including industry-first 2nm optical-interconnect demos for AI data centres (Sept 22). Marvell earns on the rack and the fabric regardless of whose accelerator wins, and Nvidia's earlier strategic investment and AI partnership remains a credibility stamp.
  • Real profitability, real deleveraging. FY2026 delivered $2.67B GAAP net income and FY2025's losses are behind it; TTM FCF $1.73B (18.3% margin) and net debt has fallen to ~$1.35B from ~$4.1B in FY2022 — a balance sheet that can fund the scale-up without the dilution that worries most AI-adjacent names.

🔴 Bear Case

  • The multiple is priced for the story. At $287.01 the stock trades at ~26x TTM sales, ~96x TTM GAAP earnings, ~51x forward and ~145x free cash flow, after a +214% year-to-date run, and is still ~9% below its June 2026 high of $316.43. Consensus targets cluster almost exactly at spot (~$266–$305 depending on the source, ~$304.99 per 45 analysts) — the market has already paid for the new guidance.
  • $70–90B by FY2031 is a slide, not a backlog. Custom-silicon programs are notoriously program-gated and can slip or be in-sourced; hyperscaler capex can pause after any digestion scare. Broadcom dominates the custom-ASIC socket today and every large customer runs its own silicon team as the ultimate competitor.
  • Mono-line growth, and margin eyes are already flat. With Data Center at ~80% of the mix, the entire thesis is a leveraged bet on one capex cycle; non-GAAP gross margin is guided down sequentially (58.9% → 57.5–58.5%) even as revenue inflects, and the stock has had 62 single-day moves greater than 5% in the past year.

Verdict

Marvell is the cleanest listed expression of the part of the AI buildout that isn't Nvidia: the custom silicon and optical/interconnect plumbing that every hyperscaler needs and almost none of them make themselves — and the Investor Day converted a narrative into explicit, escalating numbers ($20B FY28, $70–90B FY31) rather than a vague faith in "AI demand." The risk isn't that the story is fake; it's that the price already assumes the ambitious end lands on time, with consensus targets sitting at spot, gross margin flat-to-down, and the whole equity now levered to a single customer cohort's capex. Interesting to understand and worth a place on an AI-infrastructure watchlist, but at ~26x sales after a 214% year the margin for error on any program slip is thinner than the 2031 chart implies.

Disclaimer: AI-generated quick take for informational purposes only. Not investment advice. Data from public filings and financial data providers.