Nvidia CEO Jensen Huang Believes That Marvell Technology Could Become a $1 Trillion Company. Here's Why He Is Right

Marvell Technology is being positioned as a major AI infrastructure beneficiary, with management highlighting accelerating networking demand and a 70% jump in DCI business this year versus a prior 50% estimate. The article also cites $600 million in switching revenue for fiscal 2027, rising to $1 billion in fiscal 2028, and custom silicon growth potentially exceeding 100% in fiscal 2028. The piece argues Marvell could eventually reach a trillion-dollar valuation, reinforcing a bullish long-term outlook.

Analysis

The market is starting to re-rate Marvell not as a cyclical semiconductor supplier but as a toll collector on AI cluster scaling. The key second-order effect is that once compute and memory are partially solved, networking becomes the next hard constraint, which shifts budget share toward interconnect, switching, and optical layers with better visibility and stickier design wins. That is structurally favorable for MRVL, but also creates a broader read-through for the optical supply chain: component shortages should sustain pricing power and delay normalization well beyond the initial AI buildout wave.

What matters for investors is that this is still an early-cycle bottleneck, so the setup is likely to be stair-stepped rather than linear. The fastest upside is in the next 4-8 quarters as hyperscaler capex converts into revenue and the market extrapolates forward guidance; the bigger risk over 2-3 years is competitive substitution, especially if customers internalize more networking silicon in-house or if larger platform vendors bundle adjacent functionality. That means the stock can compound sharply, but only if Marvell keeps expanding share in the most constrained parts of the stack.

The consensus may be underestimating the operating leverage in the mix shift. If optical and switching are genuinely supply-constrained, incremental revenue should carry above-average gross margin and disproportionately boost EPS, which makes the multiple expansion case more credible than a simple top-line story. The contrarian risk is that enthusiasm is front-running the cycle: if supply catches up faster than expected, the market could de-rate the whole AI networking basket before the long-duration earnings stream is fully visible.

Key Decisions for Investors

• Long MRVL on pullbacks over the next 1-3 weeks; target a 6-12 month hold. Best risk/reward is a momentum-plus-revision trade, but size modestly because the stock has already re-rated on the thesis.
• Pair trade: long MRVL / short a broad semiconductor ETF over 3-6 months. The idea is to isolate AI networking scarcity while hedging against an industry-wide multiple reset if AI capex sentiment cools.
• Add MU on weakness as a secondary beneficiary of the same AI infrastructure bottleneck over 6-12 months. Memory and networking should both remain capacity-constrained, but MRVL has more direct upside to networking scarcity while MU offers more obvious cash-flow support.
• Use call spreads rather than outright equity in MRVL for 3-9 months. A capped-upside structure fits a thesis where the fundamental path is strong but valuation already reflects part of the good news.
• Avoid chasing NVDA on this headline; the relative winner here is the networking layer, not compute. If anything, this favors a rotation into suppliers exposed to interconnect, optics, and custom silicon versus the core GPU complex.