Key Highlights

  • Marvell (Nasdaq: MRVL) surged over 20% pre-market on June 2, reaching US$262.47.
  • Nvidia CEO Jensen Huang called Marvell the next trillion-dollar company at Computex in Taipei.
  • Huang's remark implies a roughly fivefold rise from Marvell's current market cap of approximately US$192 billion.
  • Marvell posted record Q1 Revenue of US$2.42 billion, up 27.6% year-over-year, with Q2 guidance of up to US$2.84 billion.
  • Nvidia holds a US$2 billion Equity stake in Marvell as part of a semi-custom AI infrastructure collaboration.

An Endorsement That Reframes the Valuation Conversation

Marvell Technology, Inc. (NASDAQ:MRVL), a Wilmington-based fabless semiconductor company that designs high-performance data infrastructure solutions spanning data centre networking, custom AI accelerators and optical interconnect technologies, entered pre-market trading on June 2 with shares surging over 20%, the result of a single, consequential public endorsement. Nvidia CEO Jensen Huang, speaking alongside Marvell CEO Matt Murphy at the Computex conference in Taipei, told Murphy directly: "That's why you're going to be the next trillion-dollar company."

Huang's remark was not an offhand compliment. It was a strategic signal delivered from the most influential platform in the global semiconductor industry, and markets responded accordingly. Marvell's Market Capitalisation stood at approximately US$192 billion at last close. Reaching a trillion-dollar valuation would require the stock to rise roughly fivefold from current levels, precisely the trajectory Huang appeared to suggest.

What the Teralynx T100 Means for AI Infrastructure

Monday's session had already seen Marvell shares gain 8% following the launch of the Teralynx T100, the industry's first 102.4 Tbps switch silicon designed exclusively for AI workloads. Unlike legacy switching platforms built for traditional enterprise and cloud environments, the T100 was architected from the ground up for AI Training and inference, eliminating the legacy elements that add power and complexity to competing solutions.

At under 1,000 watts of typical power consumption, the T100 delivers up to 25% lower power than competitive solutions, a specification that carries material implications as GPU rack densities approach 120KW. At that power level, switching and networking components account for approximately 15 to 25% of total rack power, making low-power switch silicon a strategic requirement rather than a preference. The T100's support for 512-port scale-out radix and advanced scale-up fabric protocols, including the Ethernet Scale-Up Networking standard and Ultra Ethernet Consortium requirements, positions it for deployment across the largest AI clusters currently being built by hyperscalers and cloud operators.

The product runs on a monolithic 102.4 Tbps device manufactured on 3nm process technology, a process node that underscores Marvell's commitment to silicon efficiency at the frontier of AI infrastructure.

Custom Silicon, Optical Networking and the Revenue Trajectory

The Nvidia endorsement amplifies a financial narrative that was already strengthening. Marvell reported record Q1 fiscal 2026 revenue of US$2.42 billion, up 27.6% year-over-year, with Q2 guidance raised to as much as US$2.84 billion. Analysts raised price targets following the Earnings beat, with some analyst projecting that Marvell's optical-networking revenue could grow by as much as 90% in each of the next two fiscal years.

The company's custom chips Business, which designs application-specific AI accelerators for hyperscale cloud customers, is forecast to surpass US$10 billion in revenue by fiscal 2029. That projection, issued by Marvell's own management, establishes a medium-term growth anchor that gives institutional investors a quantifiable framework for valuation modelling at the trillion-dollar level.

Marvell has also expanded its addressable market through Acquisition. The company acquired optical networking firm Celestial AI for US$3.25 billion and interconnect technology company XConn for US$540 million in recent months, deepening its capabilities across the full AI data centre stack.

Nvidia's own US$2 billion equity Investment in Marvell, structured as part of a collaboration enabling customers to integrate components from both companies in semi-custom AI infrastructure, adds a further layer of strategic interdependence. It also signals that Nvidia views Marvell's networking and interconnect capabilities as foundational to the continued scaling of its own GPU ecosystem, a relationship that reduces Demand risk for Marvell's most strategically critical product lines.

Semiconductor Sector Context

The Huang endorsement arrives at a moment when the broader semiconductor sector is navigating a bifurcated demand environment. Conventional logic and memory markets remain under cyclical pressure, while AI-oriented infrastructure spending continues to accelerate. Marvell's repositioning as a pure-play data infrastructure supplier, with its Data Centre segment accounting for approximately 72% of fiscal 2026 revenue, places it squarely in the high-growth segment of that bifurcation.

Peers including Broadcom Inc. (NASDAQ:AVGO) compete in portions of the custom AI silicon and data centre switching markets, but Marvell's combination of optical DSP expertise, high-radix switch silicon and Nvidia alignment gives it a differentiated position that is difficult to replicate at equivalent scale and speed.

Conclusion

Jensen Huang's trillion-dollar declaration at Computex is the kind of public endorsement that reconfigures how institutional Capital thinks about a company's long-term valuation ceiling. For Marvell, it arrives at a moment of genuine operational strength: record revenue, rising guidance, a purpose-built AI switch silicon in sampling, a US$2 billion Nvidia investment, and a custom chip revenue trajectory approaching double digits by fiscal 2029. The pre-market surge reflects the market repricing that combination. Whether the stock sustains these gains will depend on execution against a demanding Growth Curve, but the structural case for Marvell as a central node in AI infrastructure has rarely looked more credible.