The Data-Infrastructure PortfolioThin moat

Marvell Technology (MRVL) — moat facet

Ballast rather than engine: useful for credibility and for absorbing IP costs, but not the reason anyone owns the shares.

Beyond custom silicon and optics, Marvell sells a range of data-infrastructure chips: Ethernet switching silicon, storage controllers, and the networking parts that go into enterprise equipment and carrier infrastructure. In fiscal 2026 these amounted to a few hundred million dollars a quarter — enterprise networking, carrier infrastructure and consumer together — against a data-center segment several times their size1.

Communications and other share of revenue (%)60%FY202428%FY202526%FY202621%Q2 FY2027Marvell Form 10-K FY2026 and Q2 FY2027 release
Everything outside the data center went from three-fifths of Marvell to a fifth in under three years.

The strategic argument for keeping them is that a customer building a data center needs all of it, and a supplier who can provide compute, switching, storage and optical connectivity with interfaces known to work together is worth more than four suppliers who cannot. That systems-level credibility is part of what gets Marvell into custom-silicon conversations in the first place.

The financial argument is more mixed. These businesses are cyclical, lower-growth and lower-margin than the data center, and they compete against entrenched incumbents — Broadcom in switching most of all. Carrier infrastructure in particular has spent years in a telecom spending trough before recovering sharply.

The honest read is that these lines are ballast rather than engine: useful for credibility and for absorbing IP development costs, but not the reason anyone owns the shares. Marvell has already shown, with the automotive sale, that it will part with what does not fit.

Moat trajectory: Narrowing

Not because these businesses are deteriorating but because they keep shrinking as a share of the whole — from roughly 39% of revenue to 26% in three years — while facing the same opponent that leads in custom silicon. The portfolio argument gets harder to make each year.

The number that tests this moat
Moat Explorer calc
Communications and other revenue, latest quarter
$567.8M in Q2 fiscal 2027, up 10%

The non-data-center businesses together; growth here is the cyclical recovery in enterprise and carrier markets.

How it's calculated: 567.8 / 515.6.
Source: Marvell Q2 fiscal 2027 results release (Exhibit 99.1, 27 August 2026) ↗
Aspects of the moat
DIG DEEPER 01
Switching Silicon and the Innovium Bet
Switching isn't only a silicon problem — it's software, diagnostics and confidence a fabric will hold under load.
DIG DEEPER 02
Storage: The Business Marvell Started In
The 1995 business, given a second life by AI training sets that have to be fed from somewhere.
DIG DEEPER 03
Carrier and Enterprise: The Cyclical Half
Growth rates here say more about the comparison base than the business — long troughs, sharp recoveries, neither chosen.
DIG DEEPER 04
The Systems Argument
Customers don't buy chips, they build systems — but Broadcom makes the same argument holding a stronger hand in most categories.
⚠ Threats to the moat
⚠ THREAT 01
The Legacy Lines Keep Shrinking
The diversification that breadth was supposed to provide has been overwhelmed by the growth of a single segment.
⚠ THREAT 02
Broadcom Owns the Switching Market Too
The same opponent, in the adjacent business, with a decade of field-proven deployment behind it.
References
  1. Third-party estimateEnterprise networking, carrier infrastructure and consumer together are a few hundred million dollars a quarter against a far larger data-center segment.
    Marvell quarterly segment disclosures — data center revenue rose to roughly three-quarters of total revenue, from about 61% two years earlier; carrier infrastructure revenue nearly doubled year on year in Q3 FY2026 off a depressed base — FY2024-FY2026 · publ. 2026 · source ↗
Sources
Generated September 23, 2026