Nvidia has taken a new step in artificial intelligence and data center solutions with NVLink Fusion, which it announced at the Computex 2025 event. NVLink, the high-speed connection technology that the company has limited to its own hardware until now, is now compatible with third-party processors and special AI accelerators.

Nvidia announces NVLink Fusion technology

NVLink significantly increases efficiency in artificial intelligence servers by providing high bandwidth and low latency, compared to the PCIe standard between the CPU and GPU, up to 14 times higher data transfer speed.

However, this technology has long been exclusive to Nvidia. With NVLink Fusion, this limitation is eliminated, and manufacturers such as Qualcomm and Fujitsu are working on infrastructures that can directly integrate their own CPU solutions with Nvidia systems.

Qualcomm, in particular, is developing a new generation server processor in collaboration with Nvidia. Fujitsu is preparing to benefit from this new platform with its Monaka processors with 144 cores and 3D stacked architecture.

It is stated that Nvidia will establish partnerships not only with CPU manufacturers but also with companies that develop special-purpose artificial intelligence accelerators as part of the program. Companies such as MediaTek, Marvell and Alchip have started working to integrate their own special AI chips into Nvidia’s systems.

The infrastructure is also being completed on the software side. Synopsys and Cadence will provide the necessary IP and design tools for the development of NVLink Fusion-compatible chips, while Astera Labs will provide special connectivity solutions for NVLink.

However, it was noteworthy that major players such as AMD, Intel and Broadcom were left out of this structure. So what do you think about this? You can share your views with us in the comments section below.