CompetitorsNarrow moat

Broadcom (AVGO) — moat facet

Broadcom rarely sells against anyone in an open market — its rivals are the ones positioned to take the next design cycle, and each is exploiting a different feature of the model.

Broadcom competes in a way most chip companies do not: it rarely sells against anyone in an open market. Its silicon is designed into a customer's product years ahead, and by the time a rival could bid, the socket is gone. So the interesting question is not who makes similar parts but who is positioned to take the next generation — and the four answers here are four different kinds of opponent.

Four rivals, four ways of competingNvidiabundles networking with the GPUs — 4% to 21.5% shareMarvellleads optical DSPs ~60-70% vs Broadcom ~30%Ciscosells Silicon One G300 merchant — 102.4 Tb/sAristabuys Tomahawk; competes with nobody hereNutanix / Proxmoxsell the exit from VMwareNone is winning on silicon merit — and none has yet forced Broadcom to cut price.
Each rival exploits a different feature of Broadcom's model rather than out-engineering it.

Nvidia is the strangest of them. Broadcom's fastest-growing business exists precisely because hyperscalers want an alternative to buying Nvidia GPUs, which makes Broadcom the armourer rather than the combatant. Nvidia has struck back on Broadcom's own ground, taking Ethernet switch share from under 4% to about 21.5% in a single year by bundling networking with the accelerators everyone already needs1. Marvell is the conventional rival — genuinely second in custom silicon, and genuinely first in the optical DSPs Broadcom would like to own2. Cisco and Arista occupy the same category and behave oppositely: one buys Broadcom's silicon and builds a business on it, the other builds its own and now sells it to everybody else3. And in software, the VMware customers who found the new licensing intolerable had somewhere to go4.

What unites them is that none is winning on price or performance alone. Each is exploiting a structural feature of Broadcom's model: that arming customers against a supplier invites retaliation, that a portfolio built by acquisition has gaps, that merchant silicon creates the very competence customers need to leave, and that captive pricing eventually creates a market for the alternative.

The number to watch across all four is Broadcom's ~40% operating margin. It is the clearest evidence that these contests have not yet reached price, and the first place they will show up when they do.

Moat trajectory: Holding steady

Broadcom is not losing any of these contests, and in the largest of them — custom accelerators — it is extending. What has changed is that each rival has found a way to compete that does not require beating Broadcom's silicon: Nvidia bundles, Cisco sells its own chips merchant, Nutanix sells the exit. None has yet reached price, which is why the ~40% operating margin still holds. Stable, and worth re-reading the moment that margin moves.

The number that tests this moat
Third-party estimate
Custom-accelerator design share, Broadcom against Marvell
~70% against ~20-25%

Broadcom's rivals win by bundling, by selling merchant silicon or by selling the exit, not by beating its custom silicon. This share is the test: Marvell or an Asian design house closing the gap would mean the silicon contest itself is turning.

Source: Forbes, design-share estimates (Dec 2025) ↗
Dig deeper
DIG DEEPER 01
Nvidia: The Rival Broadcom Arms Its Customers Against
Broadcom is paid by Nvidia's customers to reduce their dependence on Nvidia — so Nvidia attacked the one market where it could bundle, and took a…
DIG DEEPER 02
Marvell: The Second Source That Leads One Market
Broadcom beats Marvell roughly three to one in custom silicon and loses optical DSPs by about the same margin — proof its advantages are specific, not…
DIG DEEPER 03
Cisco and Arista: One Customer Defected, One Didn't
Merchant silicon built Arista's business and taught Cisco to build its own; the same model produced a loyal customer and a new competitor.
DIG DEEPER 04
Nutanix, Proxmox and Where the VMware Customers Went
Captive pricing works until the captives find an exit — and the hard part of leaving VMware is now Broadcom's real defence, not the switching costs it…
References
  1. Third-party estimateNVIDIA's Ethernet switch share rose from under 4% to about 21.5% in a single year.
    Third-party AI-networking market analysis — NVIDIA's Ethernet switch share rose from under 4% to approximately 21.5% in a single year, won by bundling networking with the GPUs customers already required; Broadcom retains a near-monopoly position in hyperscaler Ethernet switch silicon with its Tomahawk and Trident families — 2026 · publ. 2026 · source ↗
  2. Third-party estimateMarvell leads high-speed PAM4 optical DSPs with roughly 60-70% share against Broadcom's ~30%.
    Third-party analysis of the high-speed optical DSP market — Marvell and Broadcom together hold more than 90% of the PAM4 DSP market; within it Marvell holds roughly 60-70% against Broadcom's approximately 30%; Marvell was first to 200G-per-lane 1.6T silicon in 5nm (Nova, 2023) followed by the 3nm Ara platform, and Broadcom's Sian3 targets the same 3nm 200G-per-lane ground — 2026 · publ. 2026 · source ↗
  3. ReportedCisco's Silicon One G300 is sold as merchant silicon to other vendors, not reserved for Cisco's own systems.
    The Register — Cisco unveils the 102.4 Tbps Silicon One G300 switch chip, aimed at Broadcom's Tomahawk 6 and NVIDIA's Spectrum-X, and sold as merchant silicon to other vendors in addition to powering Cisco's own networking appliances — February 2026 · publ. February 10, 2026 · source ↗
  4. Third-party estimateNutanix reports thousands of VMware migrations, its strongest new-customer growth in nearly a decade.
    Trade-press reporting of Nutanix management claims — roughly 30,000 customers migrated from VMware according to chief executive Rajiv Ramaswami, with 640 VMware customers added in the first quarter of fiscal 2026 on top of more than 2,700 across fiscal 2025, described as the company's strongest new-customer growth in nearly a decade; drivers cited are price increases, forced bundling and the end of perpetual licences; Red Hat is characterised as playing primarily where containerisation rather than virtualisation is the point — FY2025-FY2026 · publ. April 2026 · source ↗
Sources
Generated September 22, 2026