May 19th, 2025– Nvidia (NVDA.O) announced a major technological breakthrough at Computex 2025 with the launch of NVLink Fusion, a cutting-edge technology designed to accelerate chip-to-chip communication for artificial intelligence (AI) applications. Nvidia CEO Jensen Huang introduced NVLink Fusion during his keynote speech at the Taipei Music Center, where the annual technology exhibition is being held from May 20 to 23.
The new technology aims to enhance the connectivity of AI chips, enabling faster data exchange and significantly improving the performance of multi-chip AI systems. Marvell Technology (MRVL.O) and MediaTek are among the first major companies planning to integrate NVLink Fusion into their custom chip projects, marking a notable expansion of Nvidia’s influence beyond its own hardware ecosystem.
A New Era for AI Systems
NVLink Fusion represents an evolution of Nvidia’s NVLink technology, which was originally developed to facilitate high-bandwidth communication between GPUs. The latest version is engineered to support not only Nvidia’s own processors but also chips from other manufacturers, creating a universal solution for building robust AI infrastructure.
According to Huang, NVLink Fusion will be made available to other chip designers to foster the development of custom AI systems that require massive parallel processing capabilities. “This is about empowering innovation. NVLink Fusion is the bridge that enables multi-chip AI platforms to operate seamlessly,” Huang stated during his address.
The announcement underscores Nvidia’s strategic shift from its traditional role as a leader in gaming graphics to dominating the AI hardware sector. Huang noted that presentations, which once focused 90% on gaming graphics, now predominantly highlight AI-driven innovations—a testament to Nvidia’s evolving market influence.
Global Partnerships and Taiwan Expansion
In tandem with the launch of NVLink Fusion, Huang also unveiled plans for Nvidia to build a new Taiwan headquarters in the northern suburbs of Taipei. The strategic location aims to strengthen Nvidia’s presence in Asia and foster closer collaboration with local and global technology partners.
The new headquarters announcement follows Nvidia’s rapid expansion in AI technologies, supported by partnerships with key players in the chip manufacturing space. Marvell and MediaTek’s decision to adopt NVLink Fusion is viewed as a significant endorsement of Nvidia’s open-architecture approach to AI hardware.
Advancements in AI Hardware
During his keynote, Huang also showcased Nvidia’s recent advancements in AI chips, including the Blackwell Ultra and DGX Spark. The Blackwell Ultra, set to launch later this year, represents Nvidia’s latest leap in AI processing power, while the DGX Spark—a desktop version of its AI chips aimed at researchers—will be available in the coming weeks. This release is poised to democratize access to high-performance AI computing, making it more accessible to developers and academic researchers globally.
Nvidia’s roadmap extends even further, with plans to roll out Rubin chips next, followed by the Feynman processors, anticipated to debut in 2028. These chips are designed to meet the growing computational demands of AI models, reinforcing Nvidia’s commitment to maintaining its edge in the rapidly evolving AI landscape.
Computex 2025: A Hub for Global Tech Leaders
Huang’s announcement comes as Computex 2025 gathers over 1,400 exhibitors, marking the first significant assembly of computer and chip executives in Asia since the United States’ renewed push for localized semiconductor manufacturing. At last year’s Computex, Huang famously sparked “Jensanity“—a media sensation that saw the CEO mobbed by enthusiastic attendees, highlighting his star status in the tech community.
This year, Nvidia’s announcements further solidify its role as a cornerstone of global AI innovation. By providing technologies like NVLink Fusion to external chip designers, Nvidia is not just building products—it’s laying the groundwork for the next generation of AI infrastructure.
Shaping the Future of AI
The introduction of NVLink Fusion positions Nvidia as a crucial enabler of interconnected AI systems, a necessity as applications scale in complexity and size. With industry giants like Marvell and MediaTek already on board, the technology is expected to become a standard in multi-chip AI solutions.
Moreover, Nvidia’s investment in Taiwan reflects a broader strategy to strengthen its foothold in Asia, a critical region for semiconductor manufacturing and technological innovation. As the global race for AI dominance accelerates, Nvidia’s strategic moves could redefine the competitive landscape.
NVLink Fusion is more than just a technological upgrade—it’s a statement of intent from Nvidia, signaling its ambition to lead the AI revolution not just through innovation, but also through strategic alliances and global expansion.
With Computex 2025 serving as the launchpad, Nvidia has once again set the stage for what’s next in AI computing—one link at a time.
