Nvidia is allowing you to bring your own CPU and custom processors to connect to its GPUs and AI infrastructure via NVLink Fusion. The upshot is that Nvidia is happy to open up on AI factories so it can develop its ecosystem.
Speaking at his Computex 2025 keynote, Nvidia CEO Jensen Huang unveiled NVLink Fusion. NVLink Fusion allows cloud providers and presumably sovereign AI efforts and ultimately private infrastructure to use any ASIC or CPU to scale out Nvidia GPUs. For cloud providers like AWS, Google Cloud and Microsoft, NVLink Fusion gives them the option to couple their custom CPUs with Nvidia.
Initially, MediaTek, Marvell, Alchip Technologies, Astera Labs, Synopsys and Cadence are the early adopters of NVLink Fusion for its custom silicon. Qualcomm also announced its data center efforts and moves to integrate its CPUs into Nvidia infrastructure. Fujitsu and Qualcomm CPUs can also be integrated into Nvidia GPUs via NVLink Fusion.
The move makes sense on many fronts, but Huang summarized the strategic importance of NVLink succinctly. Huang said:
"Nothing gives me more joy than when you buy everything from Nvidia. I just want you guys to know that, but it gives me tremendous joy when you buy anything from Nvidia."
Bottom line: Nvidia will offer its fully integrated AI stack, but will also disaggregate it since in the long run the GPUs, platform and ecosystem plays are more important.
Nvidia NeMo Microservices generally available, aims for AI agent data flywheel | Nvidia GTC 2025: Six lingering questions | Nvidia launches Blackwell Ultra, Dynamo. outlines roadmap through 2027
Constellation Research analyst Holger Mueller said:
"Nvidia once again acknowledges the importance of the network for AI. The speed and efficiency how data is served to the precious and inexpensive CPUs is what matters. With allowing partners to work with the NVLink network Nvidia prioritizes the network over its own inhouse designs - which is true to its DNA as component vendor."
Two major players missing from the NVLink Fusion announcement are Broadcom and AMD. Qualcomm CEO Cristiano Amon said partnering with Nvidia advances its efforts into the data center.
NVLink Fusion can connect custom CPUs and ASICs to GPUs via Nvidia ConnectX-8 SuperNICs, Nvidia Spectrum-X Ethernet and Nvidia Quantum-X800 InfiniBand switches, with co-packaged optics in the near future.
Other Nvidia announcements from Computex include:
- Nvidia launched its first DGX PCs with a bevy of partners including Acer, ASUS, Dell, HP, Lenovo and others. Nvidia launches DGX Spark, DGX Station personal AI supercomputers
- Foxconn, Nvidia, Taiwan and TSMC are collaborating to build a Nvidia Blackwell supercomputer. The move gives Taiwan a supercomputer for R&D.
- Cisco, Dell Technologies, Hewlett Packard Enterprise and Lenovo will offer integrated AI systems powered by Nvidia RTX Pro servers.
- Japan’s National Institute of Advanced Industrial Science and Technology (AIST) opened the ABCI-Q, the world's largest supercomputer dedicated to quantum computing research. Here’s what Nvidia CEO Jensen Huang said about quantum computing, Project Digits and robotics
More:
