Nvidia (NASDAQ: NVDA) announced a significant move to expand its influence in AI hardware on Monday, revealing plans to sell its NVLink Fusion technology to other chipmakers, enabling faster communication between artificial intelligence (AI) chips.
The announcement came during CEO Jensen Huang’s keynote at the Computex AI exhibition in Taipei.
The new NVLink Fusion technology is designed to boost chip-to-chip data transfer, a critical need for developers building multi-chip AI systems. The offering will make Nvidia’s interconnect technology—once proprietary—available to external chip designers, allowing them to create custom high-performance AI infrastructure.
MediaTek and Marvell Technology have already committed to adopting NVLink Fusion, underscoring Nvidia's growing ecosystem and its dominance in the rapidly evolving AI chip sector.
“Now, we are a full-stack computing company,” Huang told the audience at Taipei Music Center, noting the company’s evolution from graphics processors to AI platforms. “Our mission is to build the engine of the next industrial revolution.”
From Gaming Roots to AI Supremacy
NVLink was originally developed by Nvidia to move large volumes of data between GPUs in high-performance computing. In Nvidia's GB200 architecture, for instance, two Blackwell GPUs are paired with a Grace CPU, interconnected by NVLink for seamless data exchange.
The new Fusion iteration opens that capability to third parties, expanding Nvidia’s role from chip provider to AI infrastructure enabler.
At Computex, Huang also highlighted Nvidia’s roadmap:
- The Blackwell Ultra AI chip is due out later this year.
- Future processors, the Rubin series, followed by Feynman chips, are scheduled through 2028.
- The newly launched DGX Spark desktop AI system for researchers is now in full production and expected to be available in a few weeks.
Strategic Expansion in Asia
Alongside the product announcements, Huang also confirmed that Nvidia will build a new Taiwan headquarters in the northern suburbs of Taipei, reinforcing the company’s ties with one of the world’s most critical semiconductor hubs.
Nvidia’s appearance at Computex marks the first major industry gathering in Asia since U.S. President Donald Trump proposed sweeping tariffs aimed at reshoring chip production to the U.S. Despite these pressures, Nvidia’s expanding presence in Taiwan signals its commitment to global manufacturing and innovation partnerships.
Last year, Huang’s popularity at Computex sparked what locals dubbed “Jensanity”, and this year’s announcements are expected to generate similar buzz among industry leaders and investors.
As the AI race accelerates, Nvidia continues to set the pace—not just through new chip architectures, but by offering the underlying connective technology that may power the next generation of intelligent machines.
Read next
11:00
Nvidia
Nvidia will take a $5.5 billion charge after the U.S. restricted exports of its H20 AI chip to China, citing security concerns. The move targets China's access to advanced tech, impacting Nvidia's key market amid growing global AI competition.
21:00
Nvidia
Nvidia CEO Jensen Huang forecasted a rapid arrival of humanoid robots in manufacturing, stating that widespread use in factory settings is likely within the next few years.
22:15
Nvidia
The Nvidia’s annual software developer conference this week, CEO Jensen Huang is set to defend the company’s leading position in the AI chip market amid growing competition and cost pressures on its largest customers.
22:00
Nvidia
Chip designers Nvidia and Broadcom are conducting manufacturing tests using Intel’s advanced 18A process, according to sources familiar with the matter. The tests are part of an effort by both companies to evaluate whether Intel’s technology can meet the demanding requirements for advanced artificia
17:00
Nvidia
During the latest earnings call, Nvidia CEO Jensen Huang reiterated his bullish outlook for the company, dismissing concerns that DeepSeek’s R1 reasoning model might undercut Nvidia’s sales.
What is your opinion on this topic?
Leave the first comment