Nvidia seems to be going all in on artificial intelligence. Nvidia CEO Jensen Huang made a slew of announcements and revealed new products on Monday that are aimed at keeping the company at the center of artificial intelligence development and computing.
The chipmaker had one particular product that it seemed to highlight, its new “NVLink Fusion” program. “NV link fusion is so that you can build semi-custom AI infrastructure, not just semi-custom chips,” Huang said at the Computex 2025 in Taiwan, Asia’s biggest electronics conference.
READ: Nvidia and Anthropic clash over US AI chip policy (May 2, 2025)
The NVLink Fusion seems to be built upon the NVLink, which is a high-speed interconnect technology that enables fast communication between GPUs, surpassing the bandwidth limitations of PCIe. Designed for data centers, AI, and HPC workloads, NVLink allows multiple GPUs to share data and memory efficiently. It offers significantly higher bandwidth—up to 600 GB/s in GPUs like the A100 using multiple NVLink connections. Unlike PCIe, NVLink supports unified memory access, enabling GPUs to access each other’s memory directly.
In large systems, NVLink works with NVSwitch to provide all-to-all GPU communication with consistent low latency. While NVLink was briefly available in high-end consumer GPUs like the RTX 3090, Nvidia has since shifted its focus to enterprise and AI platforms. NVLink enhances scalability and performance in multi-GPU setups, making it critical for training large AI models and scientific simulations. It is currently implemented in Nvidia’s Hopper, Ampere, and Volta-based systems such as the DGX series and other HPC solutions.
READ: All about Saudi crown prince’s new AI company Humain (May 14, 2025)
The program will reportedly allow customers and partners to use non-Nvidia central processing units and graphics processing units together with Nvidia’s products and its NVLink.
According to Ray Wang, a Washington-based semiconductor and technology analyst, if it is widely adopted, NVLink Fusion could broaden Nvidia’s industry footprint by fostering deeper collaboration with custom CPU developers and ASIC designers in building the AI infrastructure of the future