Nvidia made a $2 billion equity investment in Marvell Technology on March 31, 2026, as part of a strategic partnership that opens Nvidia’s previously closed interconnect ecosystem to third-party chip manufacturers for the first time.
The deal centres on a platform called NVLink Fusion — a rack-scale architecture that lets custom AI accelerators from other companies connect directly to Nvidia’s infrastructure at speeds of up to 1.8 terabytes per second. Previously, NVLink was a proprietary protocol used exclusively to link Nvidia GPUs together. Under this new framework, Marvell’s custom silicon can plug into the same ecosystem.
Jensen Huang, Nvidia’s founder and CEO, put the rationale plainly: “Token generation demand is surging, and the world is racing to build AI factories. Together with Marvell, we are enabling customers to leverage NVIDIA’s AI infrastructure ecosystem and scale to build specialized AI compute.”
Marvell’s stock surged roughly 11% on the announcement.
What NVLink Fusion Actually Does
Traditional AI infrastructure has been built around stacks of Nvidia GPUs connected via high-speed interconnects. Nvidia’s NVLink protocol — which enables much faster data transfer than PCIe, the standard alternative — was a key competitive advantage, but it only worked within Nvidia’s own hardware.
NVLink Fusion changes that. It creates a platform where custom XPUs — accelerators designed by companies like Marvell for specific AI workloads — can communicate with Nvidia GPUs, CPUs, and networking hardware at the same high bandwidth speeds. The result is a more modular, semi-custom AI infrastructure stack where customers are not forced to buy everything from Nvidia, but still operate within an Nvidia-anchored ecosystem.
For Marvell specifically, the partnership means its custom chips can serve major cloud customers — including Amazon Web Services, which Marvell has a long-standing chip design relationship with — while connecting seamlessly to Nvidia infrastructure that those same cloud providers also rely on heavily.
Marvell CEO Matt Murphy described it this way: “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.”
The Silicon Photonics Dimension
Beyond the chip interconnect partnership, Nvidia and Marvell also agreed to collaborate on silicon photonics — a technology that uses light rather than copper wire to transmit data.
This matters because moving data between chips is increasingly becoming the bottleneck in AI infrastructure. GPUs are fast. Memory is improving. But the connections between components — the interconnects — are struggling to keep up with the amount of data that modern AI training and inference workloads require. Silicon photonics offers dramatically higher bandwidth and lower power consumption than electrical interconnects.
Marvell acquired Celestial AI, a silicon photonics startup, for $3.25 billion in December 2025. That acquisition is now a key part of what Nvidia is buying into with this investment.
The two companies will also work together on AI-native telecommunications infrastructure, specifically for 5G and 6G networks using Nvidia’s Aerial AI-RAN platform.
Nvidia’s Strategic Shift
The Marvell investment is not an isolated move. Since the start of 2026, Nvidia has made 19 strategic investments across the AI ecosystem, including a reported $100 billion commitment to OpenAI and smaller but significant equity stakes in infrastructure, networking, and silicon companies.
Nvidia currently holds approximately $63 billion in cash and is expected to generate as much as $400 billion in free cash flow over the next two years. That level of cash generation gives it flexibility to simultaneously invest in partners, fund research, and maintain its hardware roadmap.
The broader strategic picture is this: major cloud providers — Amazon, Google, Microsoft, Meta — have all invested heavily in designing their own custom AI chips, in part to reduce their dependence on Nvidia GPU purchases. NVLink Fusion is Nvidia’s response. Instead of fighting that trend, Nvidia is embracing it — bringing custom silicon designs into its ecosystem rather than competing against them. Customers get more flexibility. Nvidia maintains its central role in the infrastructure stack.
It is a significant strategic evolution from a company that built its dominance on proprietary hardware.
What This Means for Business
For most businesses, the Nvidia-Marvell deal is not something to act on directly. But it shapes the AI landscape in ways that matter.
AI infrastructure is getting more modular. The era of every workload needing standard Nvidia GPUs is giving way to a world of specialised chips for specific tasks. That will eventually translate into better price-performance for the AI services businesses use — cheaper inference, faster responses, more capable products at lower cost.
The AI infrastructure arms race is accelerating. Global spending on AI infrastructure is projected to exceed $630 billion in 2026. This level of investment means the underlying capabilities of AI systems will continue improving rapidly. Businesses that understand this trajectory can plan their own AI strategies around where the technology is heading, not just where it is today.
Vendor ecosystems are consolidating around a few major platforms. Nvidia is positioning itself as the connective tissue of AI infrastructure — the platform that other platforms connect to. That kind of consolidation has happened in every previous technology wave, and it shapes which tools become available, how they are priced, and what compliance and security standards matter. Businesses building serious AI capabilities should understand which ecosystem their tools sit within.
For businesses at an earlier stage of the AI journey — figuring out where to start, what to prioritise, or how to evaluate vendors — these infrastructure decisions are made by the platforms you choose. The more important decision is whether your business is building the internal capacity to take advantage of what those platforms offer.
For a deeper walkthrough of tools like this and how they fit together, the free Working With Claude field guide covers the ecosystem end to end. Get the guide.
Source
NVIDIA Newsroom