Nvidia has announced the launch of NVLink Fusion, an advanced technology designed to enhance chip-to-chip communication for the development and deployment of artificial intelligence tools. This cutting-edge tech, a new iteration of Nvidia's NVLink, will enable chip designers to create powerful custom AI systems by connecting multiple chips seamlessly.
Partnerships with industry giants like Marvell Technology, MediaTek, Alchip, Fujitsu, and Qualcomm have already been established for the adoption of NVLink Fusion in their respective custom chip endeavors. The innovation aims to streamline the exchange of massive amounts of data between different chips, facilitating the creation of complex systems such as Nvidia's GB200, which integrates multiple chips including graphics processing units and processors.
Nvidia's CEO, Jensen Huang, introduced NVLink Fusion during a keynote speech at the Taipei Music Center, coinciding with the Computex AI exhibition. Alongside this announcement, plans for the construction of a new Taiwan headquarters were also revealed, underlining Nvidia's commitment to advancing AI technology.
Huang emphasized Nvidia's evolution from a video game graphics chip manufacturer to a leader in AI chip development. Presenting a roadmap for future innovations, he highlighted the launch of new generations of AI chips like the Blackwell Ultra and upcoming Feynman processors. Additionally, Nvidia introduced the DGX Spark, a desktop version of its AI chips targeting researchers, set to be available in a few weeks.
The Computex event, hosting 1,400 exhibitors, marks a significant gathering for computer and chip industry luminaries in Asia. Against a backdrop of global economic shifts, Nvidia's latest technological advancements aim to redefine the landscape of AI systems and applications, driving the industry forward in the face of evolving computing needs.