NVIDIA and Marvell Technology announced a strategic partnership to integrate Marvell’s custom silicon and networking capabilities into NVIDIA’s NVLink Fusion ecosystem, extending the reach of NVIDIA’s AI factory architecture. The collaboration also includes a $2 billion investment by NVIDIA in Marvell, signaling deeper alignment around next-generation AI infrastructure spanning compute, networking, and optical interconnects.
The partnership centers on NVLink Fusion, a rack-scale architecture designed to support semi-custom AI systems. Marvell will contribute custom XPUs and scale-up networking compatible with NVLink Fusion, while NVIDIA will provide its full-stack infrastructure including the Vera CPU, ConnectX NICs, BlueField DPUs, NVLink interconnect, and Spectrum-X switches. The combined platform enables heterogeneous AI systems that integrate GPUs, custom accelerators, and networking into a unified architecture.
Beyond data center infrastructure, the companies plan to collaborate on silicon photonics and optical interconnect technologies, as well as AI-native telecom architectures using NVIDIA’s Aerial AI-RAN for 5G and 6G networks. The effort aims to extend AI infrastructure principles into carrier networks, positioning telco environments as distributed AI compute platforms.
- NVIDIA invests $2 billion in Marvell to deepen ecosystem alignment
- Marvell to deliver custom XPUs and NVLink Fusion-compatible scale-up networking
- NVIDIA contributes full-stack AI infrastructure including CPUs, GPUs, DPUs, NICs, and switching
- NVLink Fusion enables heterogeneous AI systems with tight integration across compute and networking
- Joint development in silicon photonics and advanced optical interconnects
- Expansion into AI-RAN architectures for 5G/6G networks
“The inference inflection has arrived. Token generation demand is surging, and the world is racing to build AI factories,” said Jensen Huang, founder and CEO of NVIDIA. “Together with Marvell, we are enabling customers to leverage NVIDIA’s AI infrastructure ecosystem and scale to build specialized AI compute.”
🌐 Analysis: NVIDIA continues to extend NVLink from a GPU interconnect into a broader rack-scale system architecture, now incorporating third-party silicon such as Marvell’s XPUs. This approach mirrors broader industry trends toward heterogeneous compute and disaggregated infrastructure, where custom accelerators coexist with GPUs under unified interconnect frameworks.
| NVIDIA Strategic Investments Across AI Infrastructure Stack (2024–2026) | ||
|---|---|---|
| Company / Date | Investment Type / AI Layer | Strategic Description |
Marvell Technology March 2026 | Equity investment Silicon / Interconnect | Approximately $2 billion. Adds custom XPUs, optical DSP, and silicon photonics to the NVLink Fusion ecosystem, extending NVIDIA’s architecture into heterogeneous scale-up systems and next-generation AI interconnect fabrics. |
Coherent Corp. March 2026 | Strategic investment + supply alignment Optics / Interconnect | Reported at roughly $300 million or more. Strengthens access to lasers, optical engines, and related materials needed to scale 800G and 1.6T optical interconnects for AI clusters. |
Lumentum March 2026 | Strategic investment + supply agreement Optics / Interconnect | Reported at roughly $200 million to $300 million. Expands NVIDIA’s access to EML lasers and other photonic components used in high-speed pluggable optics and scale-out AI fabrics. |
Crusoe Energy Systems August 2025 | Strategic investment Power / Data Center | Undisclosed. Supports AI data center development tied to stranded or low-cost energy sources, addressing one of the central bottlenecks in AI factory buildouts: large-scale power availability. |
Nebius June 2025 | Equity investment Cloud / Compute | Part of an approximately $700 million financing round. Expands sovereign AI cloud infrastructure in Europe using NVIDIA GPUs, broadening regional deployment capacity for AI training and inference. |
xAI April 2025 | Strategic participation AI Models | Participated in a funding round reported at about $6 billion. Reinforces demand for very large GPU clusters by backing a developer of frontier-scale foundation models. |
Figure AI March 2025 | Equity participation AI Applications | Participated in a round of about $675 million. Extends NVIDIA’s reach into embodied AI, where humanoid robotics requires simulation, training, inference, and edge deployment workflows. |
RunPod January 2025 | Strategic investment Cloud / Compute | Undisclosed. Backs a distributed GPU compute marketplace that offers elastic access to AI infrastructure outside the traditional hyperscale model. |
Lambda November 2024 | Equity investment Cloud / Platform | Participated in a funding round reported at about $320 million. Supports an AI cloud platform that combines GPU infrastructure with developer-facing software and services. |
Recogni June 2024 | Strategic investment Edge / Inference | Undisclosed. Targets ultra-efficient inference ASICs for automotive and edge AI, complementing centralized training infrastructure with distributed inference capability. |
CoreWeave March 2024 | Equity investment Cloud / Compute | More than $100 million in follow-on participation. Helped reinforce a GPU-native cloud provider that became an important deployment channel for NVIDIA-based AI clusters. |
🌐 At the same time, Marvell strengthens its position in AI infrastructure by combining its optical DSP, custom silicon, and emerging silicon photonics capabilities with NVIDIA’s dominant AI platform. The partnership aligns with ongoing moves by hyperscalers to co-design silicon and interconnects, similar to efforts seen with other ecosystem players building scale-up and scale-out fabrics for AI workloads.
🌐 We’re tracking the latest developments in networking silicon. Follow our ongoing coverage at: https://convergedigest.com/category/semiconductors/
