NVIDIA's CEO Picked the Next Trillion Dollar Stock

Key Concepts

• Custom Silicon (XPU): Tailor-made AI accelerator chips designed for specific hyperscaler workloads rather than off-the-shelf hardware.
• Optical Interconnects: Technology using light waves instead of copper wires to move data at high speeds within and between data centers.
• Agentic AI: An advanced phase of AI where systems perform multi-step reasoning, use tools, and access memory, significantly increasing data traffic.
• Silicon Photonics: Integrating optical components onto silicon chips to enable high-speed data movement inside server boxes.
• Hyperscalers: Large-scale cloud providers (e.g., Amazon, Microsoft, Meta, Alphabet) that drive demand for custom infrastructure.
• CXL (Compute Express Link): An open standard for high-speed CPU-to-device and CPU-to-memory interconnects.

---

1. Marvell’s Strategic Transformation

Marvell Technology has successfully pivoted from a legacy provider of enterprise networking and storage to an AI infrastructure powerhouse.

• Financial Growth: In Q1 FY2027, Marvell reported record revenue of $2.418 billion (up 28% YoY).
• Segment Shift: Data center infrastructure now accounts for over 76% of total business ($1.83 billion).
• Projections: Management expects total revenue to grow 40% to $11.5 billion in FY2027, accelerating to $16.5 billion in FY2028 (a 45% growth rate).

2. The Nvidia Alliance

Nvidia CEO Jensen Huang publicly endorsed Marvell at Computex 2026, labeling it a potential "trillion-dollar company." This is backed by a $2 billion investment from Nvidia via Series A convertible preferred stock.

• Technical Integration: The partnership focuses on Silicon Photonics and the integration of Nvidia NVLink Fusion.
• Market Positioning: Marvell acts as a "bridge," allowing cloud providers to mix custom silicon with Nvidia’s merchant hardware, effectively positioning Marvell to collect a "toll" on data transmission across these hybrid ecosystems.

3. Growth Pillars: The "Plumbing" of AI

Marvell’s strategy is built on three distinct scaling frameworks:

• Scale Out: Standard networking between servers using PAM4 DSPs (800G and 1.6T products). This segment is growing at over 70% YoY.
• Scale Across: Connecting multiple data centers to function as one. Marvell is launching a 1.6T data center interconnect module using 2nm manufacturing, targeting a $1 billion annualized run rate by FY2028.
• Scale Up: Increasing chip density within a single server rack. Through the acquisition of Celestial AI, Marvell is securing its position in silicon photonics, targeting $1 billion per quarter in this segment by FY2029.

4. Custom Silicon and XPU Attach Rates

Marvell is becoming the primary engineering house for hyperscalers (e.g., Amazon’s Tranium, Microsoft) who require proprietary chips.

• The "Attach" Strategy: When a client buys a custom XPU, they must also purchase Marvell’s secondary components, such as CXL memory controllers, security accelerators, and PCIe switches.
• Revenue Outlook: Custom silicon revenue is expected to double to over $4 billion by FY2028, with a long-term goal of capturing 20% of the $55 billion total addressable market (TAM).

5. Risks and Market Considerations

Despite the bullish outlook, the presenters highlighted significant risks:

• Customer Concentration: Heavy reliance on a small group of tier-one hyperscalers. If these giants choose to insource engineering or delay hardware refreshes, Marvell’s revenue could suffer.
• Valuation: The stock trades at a premium, with a trailing P/E of ~90 and a forward P/E of ~65. Much of the future growth is already "baked into" the current price, making it a "priced for perfection" scenario.

Synthesis and Conclusion

Marvell Technology has positioned itself as the essential "plumbing" layer for the next generation of AI. By dominating the communication layer—moving data via light and providing the custom silicon required for Agentic AI—the company has moved beyond being a single-product vendor. While the valuation is high and the company faces risks related to customer concentration, its deep integration into the hardware stacks of the world's largest cloud providers suggests a strong, sticky, and scalable long-term growth trajectory.