Taipei, May 20 — Nvidia announced on Monday that it will open its cutting-edge chip interconnect technology, NVLink Fusion, to other companies, signaling a significant step in expanding the AI hardware ecosystem. The announcement was made during CEO Jensen Huang’s keynote at the Computex 2025 trade show at the Taipei Music Center.
NVLink Fusion, an evolution of Nvidia’s existing NVLink technology, is designed to supercharge chip-to-chip communication, enabling more efficient collaboration among multiple chips in complex AI systems. This is a critical component in powering advanced AI applications, from large language models to autonomous systems.
Industry Adoption Begins
Semiconductor companies such as Marvell Technology and MediaTek have already committed to integrating NVLink Fusion into their custom chip development projects, pointing to broad industry adoption of Nvidia’s interconnect standard. The technology is currently a key feature in Nvidia’s GB200 system, which pairs two Blackwell GPUs with a Grace CPU for ultra-fast data exchange.
“NVLink Fusion will redefine custom AI infrastructure,” Huang said during his presentation. “We’re enabling a new wave of chip innovation that brings even greater performance and efficiency.”
Nvidia’s Expanding Footprint in Taiwan
During the keynote, Huang also unveiled plans for a new Nvidia headquarters in Taiwan, set to be built in the northern suburbs of Taipei. This move further underscores the company’s commitment to the region, which plays a critical role in global semiconductor production.
From Graphics to AI Leadership
Reflecting on Nvidia’s transformation, Huang remarked, “There was a time when 90% of my presentations focused on graphics chips.” Today, however, Nvidia has emerged as a leader in AI chip development, propelled by soaring demand for generative AI tools such as ChatGPT.
Nvidia’s influence continues to expand, with reports confirming the company is working on Arm-based CPUs compatible with Microsoft Windows — an effort to diversify beyond GPUs and into general-purpose computing.
Product Roadmap and AI Vision
Huang reiterated Nvidia’s commitment to AI innovation, highlighting developments announced earlier this year at the company’s developer conference, including the upcoming Blackwell Ultra chips and a long-term roadmap featuring the Rubin series and future Feynman processors, expected in 2028.
In addition, Nvidia revealed that its new desktop AI system, DGX Spark, is now in full production and will be available to researchers within weeks.
Global Tech Community Gathers at Computex
This year’s Computex hosts more than 1,400 exhibitors and marks the first major tech industry gathering in Asia since the U.S. announced potential tariffs aimed at boosting domestic chip manufacturing. The event underscores Asia’s continued importance in the global semiconductor supply chain.
As Huang took the stage to an enthusiastic crowd, local media once again dubbed the excitement “Jensanity” — a nod to his rising prominence as a central figure in the global tech industry.
