NVIDIA invests $2B in Marvell to deepen AI infra partnership
The deal focuses on custom AI chips, high-speed networking and optical connectivity, and reflects NVIDIA’s broader strategy to build an ecosystem around its AI computing platform.
Tech giant NVIDIA has invested $2 billion in Marvell Technology as part of a strategic partnership focused on building next-generation AI infrastructure.
The deal centres on combining Marvell’s custom chips and high-speed networking technology with NVIDIA’s AI computing platform, particularly a system called NVLink Fusion.
NVLink is a high-speed connection technology that allows chips such as CPUs and GPUs to communicate much faster than traditional connections. NVLink Fusion extends this idea by allowing companies to build semi-custom AI systems using their own chips while still connecting them into NVIDIA’s computing platform.
The partnership allows companies building AI data centres to mix and match components. Marvell will provide custom silicon, optical connectivity and networking technology, while NVIDIA will provide GPUs, CPUs, networking hardware and the software platform that ties everything together.
The companies are also expected to work together on silicon photonics, which uses light instead of electrical signals to move data faster and more efficiently between chips.
The partnership strengthens NVIDIA’s position at the centre of the AI infrastructure market. Even when companies use custom chips designed by firms such as Marvell, those chips will still connect through NVIDIA’s platform, which keeps NVIDIA’s technology at the core of AI systems.
Marvell gets access to NVIDIA’s large AI ecosystem and customer base. Marvell already designs custom AI chips for large cloud companies such as Amazon and Microsoft, and integrating with NVIDIA’s platform makes those chips more attractive and easier to deploy in large AI data centres.
The biggest challenge in AI is no longer just building powerful chips, but connecting thousands of chips together efficiently in data centres. This is why networking, optical connections and power efficiency are becoming critical parts of AI infrastructure. The NVIDIA Marvell partnership focuses specifically on these bottlenecks, especially bandwidth and energy efficiency.
The deal also shows that the AI industry is moving towards modular infrastructure, where different companies build specialised components that work together in a shared system rather than building everything themselves.
NVIDIA’s partnerships
The Marvell deal is part of NVIDIA’s broader strategy to build what it calls an AI ecosystem around its platform. NVIDIA is trying to become the central platform that connects chips, networking, software and entire data centres. NVLink Fusion is a key part of this strategy because it allows other companies’ chips to plug into NVIDIA systems while still relying on NVIDIA technology.
In the past six months, NVIDIA has invested about $2 billion each in several companies including Coherent, CoreWeave, Lumentum and Nebius, all of which are involved in AI infrastructure such as cloud computing or optical networking. These investments show that NVIDIA is betting not just in chip companies, but in the entire AI supply chain, including cloud providers, networking companies and optical technology firms.
NVIDIA has also announced partnerships with a wide range of companies across industries. These include telecom companies for AI-powered networks, industrial software companies for digital simulation, and cloud providers building AI data centres.