The Quiet Gross Margin Reality Check That Points to a Dangerous Valuation Gap Between Marvell and Broadcom

Quick Read

MRVL trades at a richer 68x forward P/E than AVGO’s 33x, despite posting gross margins of 52% versus Broadcom’s 67%.

Broadcom’s 46% FCF margin and VMware’s sticky software annuity make it a reliable compounder Marvell’s hyperscaler-dependent model can’t yet match.

Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks — and Broadcom didn’t make the cut. Grab the names FREE today.

Marvell Technology (NASDAQ: MRVL) and Broadcom (NASDAQ: AVGO) both just delivered AI-fueled earnings beats, yet the businesses look nothing alike under the hood. Marvell posted 231.37% YTD gains chasing custom XPU wins. Broadcom quietly compounded with elite margins and a software stack. The contrast deserves a closer look before anyone pays up.

Custom Silicon Lifts Marvell. Networking Plus Software Lifts Broadcom.

Marvell delivered Q1 FY2027 revenue of $2.42B, up 27.6% YoY, with Data Center contributing $1.83B, or 76% of revenue. CEO Matt Murphy guided Q2 to $2.70B, calling out “exceptional AI-related bookings” across 800G/1.6T optics and custom XPU programs. Encouraging, but the GAAP gross margin sat at 52.1%. That is the structural reality of bespoke silicon: hyperscalers hold the leverage.

Broadcom’s Q2 FY2026 told a different story. Revenue hit $22.19B, up 47.9% YoY, with AI semiconductor revenue of $10.8B, up 143%. Hock Tan guided Q3 AI revenue to $16.0B, a 200%-plus jump. Tomahawk and Jericho switches anchor the networking layer, and VMware adds $7.18B of sticky software revenue. That mix is why margins look the way they do.

The Gross Margin Reality Check Nobody Wants to Talk About

Broadcom’s AI segment alone generates more quarterly revenue than Marvell’s entire data center business. Yet Marvell trades at a richer forward multiple. That is a gap worth sitting with. Broadcom’s pricing power to absorb wafer fabrication costs comes from owning a networking standard rather than depending on individual hyperscaler relationships.

Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks — and Broadcom didn’t make the cut. Grab the names FREE today.

The Next Test Is Whether Marvell Can Defend Margins

I will be watching Marvell’s guided 58.25-59.25% non-GAAP gross margin band. Mix shift toward custom XPU work could pressure that, especially as Celestial AI and XConn integrations consume cash. For Broadcom, the question is whether AI revenue scales to Hock Tan’s $100B by 2027 ambition without VMware decelerating from its 9% YoY pace. Trade restrictions and hyperscaler concentration remain real risks for both.

Why I Lean Toward Broadcom on the Numbers I Can See

If I had to allocate fresh capital today, I would lean toward Broadcom. The 46% FCF margin, the dominant networking franchise, and the VMware annuity give me a compounding engine I can underwrite. Marvell intrigues me as a growth bet, and the 271.43% one-year run reflects real bookings momentum. Still, paying a richer multiple for a structurally lower-margin model looks like an unfavorable risk/reward setup. If Marvell’s margin band drifts above 60% on mix, my view changes. Until then, Broadcom looks like the safer compounder, and Marvell looks like the more exciting story stock.

Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks — and Broadcom didn’t make the cut. Grab the names FREE today.