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

Jensen Huang said Marvell could become a $1 trillion company, implying a near 5x gain from its current roughly $220 billion market value, but the article argues that outcome is unlikely anytime soon. The key strategic development is a new Marvell-Nvidia partnership to ensure networking and compute compatibility in future AI data centers, even as the companies compete in adjacent ASIC and GPU-related markets. Wall Street currently expects Marvell to earn about $4.05 per share in FY2027, far below the $38 EPS needed to justify a $1 trillion valuation at 30x earnings.

Analysis

The real incremental signal is not the headline partnership, but the admission that AI infrastructure is bifurcating into a compute stack and a connectivity stack. That is structurally positive for MRVL because it reduces the odds of a winner-take-all GPU outcome and increases the value of “plumbing” that sits in every data center configuration; however, it also means MRVL’s addressable upside depends more on design-win concentration than on broad AI capex alone. In practice, that usually translates into slower but more durable multiple support if execution stays clean.

The near-term second-order effect is that NVDA’s endorsement likely improves MRVL’s credibility with other hyperscalers and OEMs, which can compress sales-cycle friction over the next 2-4 quarters. The risk is that partnership optics can outrun fundamentals: if MRVL’s networking attach rate does not translate into faster EPS inflection by FY27, the stock may de-rate as the market realizes the “trillion-dollar path” is a long-duration story, not a next-12-month rerating catalyst. Amazon is the key shadow competitor here because custom silicon plus in-house networking optimization can keep pricing pressure on the entire merchant ecosystem.

Consensus may be underestimating how much of MRVL’s upside is already a function of its role as a strategic enabler rather than a direct AI share winner. That makes it attractive as a relative-value long versus other AI hardware names with more cyclical end-demand exposure, but less compelling as a standalone momentum chase after a positive founder endorsement. The contrarian read is that NVDA’s cooperation may actually be defensive: it helps preserve ecosystem compatibility, but it does not necessarily imply MRVL has become economically indispensable.

Key Decisions for Investors

• Long MRVL / short a basket of higher-beta AI hardware names for 3-6 months: MRVL has a cleaner ‘picks-and-shovels’ claim and should hold up better if AI capex broadens but multiples compress; target 15-20% relative outperformance, stop if hyperscaler networking insourcing headlines accelerate.
• Sell out-of-the-money covered calls on MRVL into any post-announcement strength for 30-60 days: the market is likely to overpay for the narrative before fundamentals catch up, and upside is constrained until the next earnings/design-win update.
• Maintain NVDA long exposure but hedge with a small MRVL long rather than adding outright NVDA here: the partnership reduces ecosystem risk for NVDA, but MRVL offers more convexity if connectivity becomes the next bottleneck; use a 6-12 month horizon.
• Watch AMZN custom-silicon signals closely; if AWS networking/ASIC integration deepens, trim MRVL on any relative weakness because AMZN is the most credible source of longer-term margin pressure over the next 12-24 months.
• For high-conviction accounts, initiate a pair trade: long MRVL / short INTC over 6-12 months, based on AI infrastructure relevance versus legacy turnaround optionality; reward is cleaner thematic exposure with lower fundamental execution risk.