NVLink Fusion: Nvidia To Sell Hybrid Systems Using AI Chips
Nvidia’s NVLink Fusion breaks the mold by enabling customers to build AI systems using a mix of Nvidia and non-Nvidia chips—giving the company more reach in the trillion-dollar data center market.
Summary
- NVLink Fusion allows non-Nvidia CPUs and AI chips to work with Nvidia’s GPUs.
- The initiative targets AI data centers and custom infrastructure systems.
- Partners include Qualcomm, Fujitsu, Marvell, MediaTek, Synopsys, and Cadence.
Nvidia Opens Up Its Ecosystem: NVLink Fusion
During his keynote at Computex 2025, Nvidia CEO Jensen Huang introduced NVLink Fusion, a new initiative that breaks from Nvidia’s long-standing approach of only supporting its own chips. This update allows companies to build semi-custom AI systems using CPUs and AI chips from other vendors, combined with Nvidia’s high-performance GPUs.
In simple terms, NVLink Fusion is a flexible high-speed interconnect technology that lets different processors work together. It’s a move aimed at staying central to the future of AI infrastructure—even when competitors’ chips are part of the mix.
Why NVLink Fusion Is a Big Deal
Nvidia’s dominance in AI is well known, especially in the GPU space. But cloud giants like Amazon, Google, and Microsoft are building their own custom chips. NVLink Fusion gives Nvidia a way to stay in the game by letting those same companies use Nvidia GPUs alongside their in-house processors.
“It’s a strategic move to be part of future AI system roadmaps, even if the systems aren’t fully Nvidia-based,” said Rosenblatt analyst Kevin Cassidy. Melius Research’s Ben Reitzes agrees, calling it “a big deal” and saying it will help Nvidia target more of the $1 trillion in expected data center spending.
Who’s Involved?
Nvidia has already partnered with companies like Qualcomm and Fujitsu for CPUs, while Marvell, MediaTek, and Alchip are contributing custom AI chips. Design software companies Cadence and Synopsys are also on board, making this a robust ecosystem from the start.
Notably, major competitors like Intel, AMD, and Broadcom are not part of the initiative—yet.
Pros
- Increases Nvidia’s market reach beyond its own chip lineup
- Allows customers to build custom AI infrastructure more flexibly
- Strengthens Nvidia’s role in hybrid data center setups
Cons
- Could reduce demand for Nvidia’s own CPUs like Grace
- Opens ecosystem to competitors, potentially reducing exclusivity
- Requires managing compatibility and support across third-party chips
Conclusion
NVLink Fusion represents a major pivot in Nvidia’s strategy. By embracing non-Nvidia chips, it secures Nvidia’s GPU dominance while allowing more flexibility for customers building custom AI systems. With big names already onboard and more expected to follow, NVLink Fusion could shape the next phase of AI infrastructure worldwide.
FAQs
Q: What is NVLink Fusion?
A: NVLink Fusion is Nvidia’s new initiative allowing third-party CPUs and AI chips to work with Nvidia GPUs in custom AI systems using its NVLink interconnect.
Q: Who are Nvidia’s partners in NVLink Fusion?
A: CPU partners include Fujitsu and Qualcomm. AI chip partners include Marvell, MediaTek, Astera Labs, and Alchip.
Q: Why is this shift important?
A: It helps Nvidia stay central to the AI ecosystem, even when customers use alternative chips in their data center infrastructure.
