Skip to content
Watchingmedium conviction · 1 yearActive · reviewed today$MRVL$196.40Fri 30 Oct

Marvell: the custom-chip challenger, buying the laser maker

By MenFem Editorial·Semiconductors — Custom silicon·5 November 2026·Methodology·
ai-infrastructuresemiconductorscustom-siliconoptical
The next review, the day it is written.
Share

Price since the call

$196.40+24.4% since opened

Entered at $157.87, the close on 5 Nov; last close 30 Oct. The grade comes from the falsifier, never this number.

What would prove this wrong · verbatim, as published

A watching call is a claim about the WAITING, so it fails in two directions. It stops being true if the case proves itself while we wait: data-centre revenue growing at least 40% year on year for two consecutive quarters and both shareholder votes on the Lumentum deal passing would mean the hesitation cost a move. It stops being true in the other direction if the custom-silicon case breaks: data-centre revenue falling year on year in any quarter, or the Google agreement ended, would leave nothing to wait for.
  • leg 1 · The case proves itself: data-centre revenue +40% YoY two quarters running and both Lumentum votes pass · not read yet
  • leg 2 · The case breaks: data-centre revenue down YoY in any quarter, or the Google agreement ended · not read yet

Scrub the call’s life

week ending 30 Oct · close 196.40 · +24.4% since openedNo review yet by this week.

The argument, taken apart

Turn it, hover a ring, a moon or a weight
One call, taken apart
Rings are the falsifier’s legs. Moons are the catalysts it named. Weights are its risks, longer for higher severity.
The claim

Watching: real custom-silicon growth and a Google deal, but 18 months of merger, dilution and a customer base that is still forming.

What would make it wrong · 2 legs
  1. Leg 1 · Not read yet
    The case proves itself: data-centre revenue +40% YoY two quarters running and both Lumentum votes pass
    Not read at a review yet.
  2. Leg 2 · Not read yet
    The case breaks: data-centre revenue down YoY in any quarter, or the Google agreement ended
    Not read at a review yet.
The window
5 Nov 2026 to 5 Nov 2027
0% gone · 365 days left
next print 26 Nov 2026
The reviews
Not reviewed since it was published
  • Leg 1, not read yet: The case proves itself: data-centre revenue +40% YoY two quarters running and both Lumentum votes pass
  • Leg 2, not read yet: The case breaks: data-centre revenue down YoY in any quarter, or the Google agreement ended
  • Q3 FY27 results (late Nov / early Dec 2026)
  • Lumentum merger approvals
  • high severity: Merger
  • high severity: Customer base
  • medium severity: China exposure

The timeline · 1 entry · newest firstreviews · notes · moves · the book

  1. 5 Nov 2026
    Opened

    Marvell: the custom-chip challenger, buying the laser maker

Key Points

  • Q2 FY27 revenue $2.74B (+37%); data centre 79%; non-GAAP gross margin 58.9%.
  • Google agreement across five product lines; FY28 target raised to $20B.
  • Buying Lumentum all-stock (Marvell holders ~67%); close expected 1H 2028.
  • Risks: ~86M shares of prior dilution, 44% via one distributor, 42% shipped to China, Trainium reportedly lost.

Marvell designs custom AI chips for the hyperscalers that want an alternative to buying everything from NVIDIA, and the optical and networking chips that connect them. Data centre was 79% of revenue in the quarter to August 2026; revenue rose 37% to $2.74B with a non-GAAP gross margin of 58.9%, and it guided the next quarter to $3.15B. In July it signed a Google agreement across five product lines, and at its October investor day it raised its fiscal 2028 revenue target to $20B. On 2 November it agreed to buy Lumentum, the indium-phosphide laser maker, in an all-stock deal that leaves Marvell holders with about 67% of the combined company and closes in the first half of 2028. That puts the laser chip, the part AI datacentres are short of, inside Marvell. MenFem withdrew its own Lumentum call the same day because the deal made it untestable; the laser-chip view now lives here. We are WATCHING, not buying. For the next eighteen months the shares will trade on the merger: shareholder votes, regulators, and the issuance of about a third of the company. Before that issuance, about 86M shares of dilution were already disclosed through the Google warrant, NVIDIA's convertible preferred and older warrants. Google's purchases are discretionary under its agreement, the Amazon Trainium work is reported lost, 44% of revenue runs through one distributor and 42% of shipments go to China. The entry price is the live price on 4 November, after the deal, not the pre-deal close. WATCHING, conviction MEDIUM, one year. This view was formed by MenFem's research process at Connor's request and adopted by him.

Bull Case

Custom-silicon revenue ramps with Google and others toward the $12B+ FY29 target, the Lumentum deal closes, and Marvell owns both the custom chip and the laser that connects it.

Bear Case

What I'd have to believe for the other side: waiting is the expensive choice. The shares fell sharply on the deal, the business is growing near 40% with a 59% gross margin, and once the merger noise clears the market will price a company that owns custom compute and the laser bottleneck together.

Catalysts

Q3 FY27 results (late Nov / early Dec 2026)Earnings

Custom-silicon share, gross margin, any Trainium or Maia comment, the Lumentum financing.

Lumentum merger approvalsRegulatory

Both shareholder votes and regulatory clearances before a 1H 2028 close.

Risk factors

MergerHigh

Eighteen months of votes, regulators and integration, with about a third of the company issued.

Customer baseHigh

Google's purchases are discretionary; the Amazon programme is reportedly lost; one distributor carries 44% of revenue.

China exposureMedium

42% of shipments go to China.

Key Metrics

Q2 FY27 revenue
$2.74B
Q3 FY27 revenue guide
$3.15B
Lumentum exchange ratio
4.500 MRVL per LITE

Weekly closes

entry 157.87
See all Semiconductors — Custom silicon calls