Nvidia Commits $2 Billion to Marvell Technology in NVLink Fusion Partnership

Nvidia has invested $2 billion in Marvell Technology and formalized a partnership to integrate Marvell's silicon into Nvidia's NVLink Fusion ecosystem, the companies announced Tuesday. The deal ties Marvell more closely to Nvidia's AI factory and AI-RAN infrastructure, allowing data center operators to combine Marvell's custom application-specific integrated circuits with Nvidia's NVLink interconnect fabric. NVLink Fusion is designed to let third-party chips connect at high bandwidth to Nvidia GPUs, potentially displacing the PCIe connections that currently limit performance in heterogeneous AI clusters. The partnership is notable because Marvell is simultaneously a competitor and a supplier to Nvidia. The company designs custom AI chips for hyperscalers including Amazon Web Services and Google, which are among Nvidia's biggest customers but also represent a potential threat to GPU dominance if custom silicon proves cost-effective at scale. Nvidia's $2 billion investment signals confidence that NVLink Fusion can become an industry standard for AI data center interconnects, extending the company's architectural influence beyond its own GPU products. By anchoring Marvell's roadmap to NVLink, Nvidia effectively creates a gravitational pull that could shape how third-party chipmakers design their next-generation AI accelerators. Marvell shares rose on the news. The company has been expanding its custom silicon business aggressively, with revenue from AI-related products growing substantially over the past two years as hyperscalers seek alternatives to merchant silicon for at-scale inference workloads. Financial terms beyond the $2 billion investment figure were not disclosed.