Nvidia CEO Jensen Huang Says This Will Be the Next $1 Trillion Company

Jensen Huang publicly backed Marvell as a potential future $1 trillion company and confirmed a partnership to ensure Marvell's networking technology is compatible with Nvidia's data center products. The article highlights Marvell's exposure to AI infrastructure, but notes it would need roughly $1,140 per share and about $38 in EPS to justify a $1 trillion valuation at 30x earnings versus FY2027 EPS estimates of $4.05. The tone is supportive but tempered by the long timeline and valuation gap.

Analysis

The partnership is less about near-term revenue and more about Marvell de-risking its role in the AI stack. By becoming the connectivity layer that sits between multiple compute platforms, MRVL can monetize the “picks-and-shovels” layer while reducing customer concentration risk versus a pure custom-ASIC story. That matters because networking attach rates typically scale with capex cycles more predictably than one-off chip wins, which should support multiple expansion if execution stays clean.

The second-order winner is NVDA, not because it needs Marvell for sales, but because interoperability lowers friction for enterprise and hyperscale deployments. Any reduction in switching costs accelerates adoption of Nvidia-oriented architectures and can extend the useful life of its ecosystem even when some customers hedge with custom silicon. The main loser set is smaller networking and interconnect vendors that compete on incremental performance rather than platform compatibility; once a dominant ecosystem standardizes around a few reference designs, pricing power tends to shift upstream.

The market may be underpricing how long it takes for Marvell’s earnings power to catch the valuation narrative. The “next trillion” framing is a years-long optionality story, not a 12-month catalyst, and the gap between current earnings and the level implied by that valuation means sentiment can outrun fundamentals for a while before reverting. The key reversal risk is that AI capex slows or hyperscalers internalize more of the stack, which would pressure both gross margin mix and the multiple.

Contrarian angle: the consensus is likely overreacting to the headline partnership as if it were a full endorsement of MRVL’s standalone economics. In reality, the partnership primarily validates product quality and compatibility, but it does not guarantee share gains in custom silicon or sustainable margin expansion. For investors, the setup is better viewed as a quality confirmation than a fundamental inflection.

Key Decisions for Investors

• Stay tactically long MRVL on pullbacks, but only as a 6-12 month trade; target upside is multiple expansion from ecosystem validation, while downside is capped by slower-than-expected EPS leverage if AI spend moderates.
• Add NVDA on any 3-5% weakness as a relative winner from interoperability; the thesis is ecosystem lock-in, not incremental direct revenue, so risk/reward is better than chasing MRVL after headline-driven strength.
• Pair trade: long NVDA / short a basket of smaller AI networking and interconnect names if available, betting platform standardization compresses weaker competitors’ pricing power over the next 2-4 quarters.
• Avoid paying up for MRVL if it rerates aggressively on the partnership news; consider selling out-of-the-money calls against a core long to monetize implied volatility while the “trillion-dollar” narrative remains the primary driver.
• For longer-dated exposure, use call spreads in MRVL rather than outright stock to express 12-18 month upside while limiting damage if hyperscaler capex growth decelerates or ASIC demand disappoints.