Nvidia, the global leader in AI computing, has announced a major leap in inter-chip communication with the launch of NVLink Fusion, a new version of its high-speed connection technology. The announcement was made by CEO Jensen Huang during the Computex AI exhibition held at the Taipei Music Center, which runs from May 20 to 23.
NVLink Fusion is designed to tie multiple chips together, dramatically improving the speed and efficiency of data transfer between them—a crucial factor in developing and deploying advanced artificial intelligence (AI) systems. The technology builds on Nvidia’s original NVLink, which has powered data-intensive connections between GPUs in high-performance computing environments for years.
Unlike previous iterations, NVLink Fusion will be available for use by third-party chip designers, marking a strategic shift in Nvidia’s business model. This move allows other companies to harness Nvidia’s connectivity tech to create custom AI hardware solutions that require seamless integration of multiple chips. Notably, Marvell Technology and MediaTek have already expressed plans to incorporate NVLink Fusion into their custom AI chip development efforts.
One standout example of NVLink’s capability is its use in Nvidia’s powerful GB200 chip, which connects two Blackwell GPUs with a Grace CPU. This chip trio exemplifies the scale and speed at which Nvidia aims to facilitate AI operations, enabling efficient communication across massive datasets.
By opening NVLink Fusion to the broader semiconductor industry, Nvidia is positioning itself as not just a leader in AI chips, but also as an enabler of the entire AI hardware ecosystem. This strategy could lead to faster innovation in AI infrastructure, benefiting sectors ranging from cloud computing to autonomous vehicles.
In a further sign of Nvidia’s commitment to the region, Huang also announced the company’s plan to establish a new Taiwan headquarters in the northern suburbs of Taipei, reinforcing its presence in one of the world’s most critical semiconductor hubs.
With NVLink Fusion, Nvidia is not just enhancing chip performance it is building the connective tissue for the next generation of AI
