Key Takeaways:
Marvell Technology Inc. secured a $2 billion investment from Nvidia Corp. in a strategic partnership that directly challenges Broadcom Ltd.’s dominance in data center interconnects and positions Marvell as a key player in the AI infrastructure build-out.
“By connecting Marvell’s leadership in high-performance analog, optical DSP, silicon photonics and custom silicon to NVIDIA’s expanding AI ecosystem through NVLink Fusion, we are enabling customers to build scalable, efficient AI infrastructure,” said Matt Murphy, chairman and CEO of Marvell.
The deal integrates Marvell’s custom XPU and networking technology with Nvidia’s NVLink Fusion platform, a rack-scale architecture for building semi-custom AI systems. Marvell’s interconnect product range, which already accounts for a significant portion of its $6.1 billion data center revenue, is now projected to grow by more than 50 percent in fiscal 2027.
For investors, the partnership provides a massive validation for Marvell, which trades at just 26 times forward earnings compared to Broadcom’s far richer valuation. The deal could unlock a significant re-rating for Marvell’s $85 billion market cap as it becomes an essential supplier for nearly every AI workload running on Nvidia’s ubiquitous platform.
As artificial intelligence models grow in complexity, the physical constraints of data centers are becoming the primary bottleneck. The sheer volume of data moving between thousands of GPUs generates excessive heat and signal degradation when using traditional electrical wiring. By 2027, industry experts predict electricity will become too inefficient for high-speed AI data transmission, creating what some call an "interconnect tax" on performance.
This physical barrier is driving the industry’s shift to optical connections, which use light to move massive datasets with greater speed and efficiency. Marvell has already established a strong foothold in this transition, supplying the optical digital signal processors (DSPs) used in the pluggable modules that connect servers. The company commands a leading share in high-speed 800G modules, a critical component for modern AI factories.
The Nvidia partnership accelerates this shift. Beyond today’s pluggable modules, the collaboration will focus on next-generation silicon photonics, a technology that integrates optical components directly onto silicon chips. This allows optics to be placed closer to the computation, reducing the need for power-hungry signal processing and enabling faster access to memory. According to analysis from Counterpoint, this could reduce the need for scarce and expensive HBM4 memory by as much as 30 percent.
The collaboration also serves as a strategic counterweight to Broadcom, the current market leader in interconnect and custom silicon. Hyperscale data center operators have grown wary of vendor concentration risk, a concern amplified by Broadcom's aggressive pricing strategies following its acquisition of VMware.
Nvidia’s $2 billion investment and deep integration of Marvell into its ecosystem offers customers a credible, high-performance alternative. Marvell can now position itself as a more neutral partner for custom silicon projects, leveraging the full power of Nvidia’s software stack, supply chain, and existing platforms like the Vera CPU, ConnectX NICs, and Spectrum-X switches.
This creates a significant opening for Marvell to capture share in the custom ASIC market. For fiscal year 2026, Marvell's data center business generated $6.1 billion, representing 74.4% of its total revenue. The company has already doubled its target for data center switch revenue to over $600 million, and the Nvidia deal is set to turbocharge its growth in the broader interconnect segment.
While Marvell's current market capitalization of approximately $85 billion is a fraction of Broadcom's $1.4 trillion, the direct endorsement from the leader in AI computing suggests a substantial re-rating may be on the horizon. With revenue projected to grow over 30 percent annually for the next two years, Marvell’s valuation at 18 times fiscal 2028 consensus earnings appears increasingly conservative as it becomes embedded in the world’s expanding network of AI factories.
This article is for informational purposes only and does not constitute investment advice.
