Nvidia bolsters AI infrastructure pitch with $2bn stake in Marvell

• Not for the first time, Nvidia has teamed up with and invested in a company deemed to be a semiconductor sector rival
• It is collaborating with, and has invested in, data infrastructure chip giant Marvell Technology
• One of the aims of the partnership is to encourage greater investments in AI factory and AI-RAN infrastructure

Nvidia has continued to spread its sphere of influence in the telecom tech sector by striking a strategic partnership with, and taking a $2bn stake in, data infrastructure chipmaker Marvell Technology in a move that aims to give network operators a broader range of options as they build AI factories and AI-RAN architectures. 

Nvidia is increasingly keen to be seen as a company that offers technology options to its customers while still playing a dominant role in technology supply relationships. In this instance it is specifically aiming to “connect Marvell to the Nvidia AI factory and AI-RAN ecosystem through Nvidia NVLink Fusion”, an AI infrastructure design that enables customers to integrate processors built by companies other than Nvidia into an Nvidia platform. 

This might seem strange, as Marvell is, on paper, a rival to Nvidia. But Nvidia has long been considering future market dynamics – it is already a dominant supplier of the underlying technology for AI infrastructure but it knows that customers will be wary of being reliant on tech developed and supplied by just one company and will want to avoid ‘lock in’. By teaming up with, and sometimes also investing in, companies regarded as rivals, it is not only broadening its business opportunities and keeping control of the AI tech ‘narrative’ but also sending an important message about ecosystem collaboration to customers though of course, ultimately, Nvidia will always seek to be in the supply chain driving seat. 

So how is the Nvidia-Marvell relationship set to play out? 

Nvidia noted that “Marvell will provide custom XPUs” – specialised, application-specific processors – “and NVLink Fusion-compatible scale-up networking, while Nvidia will provide the supporting technologies, including Vera CPU, ConnectX NICs [network interface cards/adapters], BlueField DPUs [data processing units], NVLink interconnect and Spectrum-X switches, and the rack-scale AI compute.”

It added: “For customers developing custom XPUs, NVLink Fusion enables a heterogeneous AI infrastructure fully compatible with Nvidia systems, allowing seamless integration with Nvidia GPU, LPU [language processing unit], networking and storage platforms while leveraging Nvidia’s rich technology stack global supply chain ecosystem.” 

In addition, Nvidia and Marvell will collaborate on the development of silicon photonics technology, which enables optical rather than electrical data processing and transmission and is regarded as key to the development of next-generation datacentres. 

Those developments will play a role in the plans that Nvidia has for telco-run AI factories, as Nvidia noted that it will work with Marvell to “transform” telecom networks into AI infrastructure through the provision of “advanced optical interconnect solutions and silicon photonics technology” for AI factories and datacentres as well as, for radio access network (RAN) infrastructure, the provision of Nvidia’s Aerial AI-RAN platform for 5G and 6G deployments that would see mobile operators deploy AI infrastructure at or near their cell sites. 

Matt Murphy, chairman and CEO of Marvell, stated: “Our expanded partnership with Nvidia reflects the growing importance of high-speed connectivity, optical interconnect and accelerated infrastructure in scaling AI. By connecting Marvell’s leadership in high-performance analogue, 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.”

Nvidia’s CEO, Jensen Huang, claimed: “The inference inflection has arrived. Token-generation demand is surging, and the world is racing to build AI factories.” He added, “Together with Marvell, we are enabling customers to leverage Nvidia’s AI infrastructure ecosystem and scale to build specialised AI compute.”

Nvidia, it seems, is very keen on taking $2bn stakes in networking component partners – just last month it pumped $2bn each into optical tech firms Lumentum and Coherent. Also this year it invested $2bn in Amsterdam-based Nebius, which builds and runs AI infrastructure and offers a range of AI cloud services to enterprise users, and the same amount in New Jersey-based AI factory builder Coreweave.

And, of course, it can afford it – for its fiscal year that ended on 25 January 2026, Nvidia reported revenues of $215.9bn (up by 65%) and an operating profit of $130.4bn (up 60%).  

The AI tech firm isn’t limited to the $2bn investment figure, though – last year it pumped $5bn into chip giant Intel and took a $1bn stake in Nokia, another move designed to further the appeal of AI-RAN architectures to the mobile operator community.     

- Ray Le Maistre, Editorial Director, TelecomTV