Nvidia is building an IP licensing empire on the back of NVLink

Nvidia is expanding its IP licensing portfolio by inviting competitors to license its networking technology, NVLink Fusion. This move comes as many companies, including MediaTek, Amazon, Fujitsu, Qualcomm, Arm, and Marvell, are developing custom AI ASICs to compete with Nvidia's GPUs. MediaTek is the latest to join the list by incorporating NVLink Fusion into its datacenter XPU offering, in exchange for a $3.5 billion investment from Nvidia.

Nvidia's move is aimed at keeping competitors reliant on its products, while also allowing them to build competitive accelerators without worrying about networking technology. Nvidia offers a range of networking technologies and topologies within its NVLink Fusion platform, including its MGX rack designs, which are part of the Open Compute Project.

This allows competitors to seamlessly integrate their compute blades into Nvidia's existing rack systems, reducing the time and effort required to go from silicon to AI racks. Nvidia's licensing ambitions extend beyond networking technology, as demonstrated by its recent unveiling of NVHBM, a custom base with an integrated memory controller designed to boost performance and efficiency for AI ASICs.

While Amazon and other companies have already adopted NVHBM for their AI accelerator families, Nvidia's licensing strategy is a testament to its focus on leveraging its engineering expertise to generate revenue from various aspects of the AI ecosystem.