Nvidia Unveils New Chip Interconnect Technology to Boost AI Development

Nvidia announced on May 20 that it will sell a new technology that links chips together to speed up the exchange of data between chips - a key factor in building and deploying Artificial Intelligence (AI) tools.

At the Taipei Music Center in Taiwan, where Computex AI is taking place from May 20-23, Nvidia President and CEO Jensen Huang introduced a new version of NVLink interconnect technology, called NVLink Fusion, which Nvidia will provide to other chip designers to help build powerful custom AI systems with multiple interconnected chips.

According to Nvidia, Marvell Technology and MediaTek will use NVLink Fusion for their custom chip projects. Other partners include Alchip, Fujitsu and Qualcomm.

However, Nvidia's business prospects in China are "shadowed." In a recent interview, Mr. Huang said the group had "forgone $15 billion in revenue" in China after the US government restricted exports of the H20 chip line to this market.

In addition to unveiling the new technology, Mr. Huang also announced plans to build Nvidia’s new headquarters in the northern suburbs of Taipei. His Computex speech highlighted Nvidia’s history of developing AI chips, systems, and software.

He said that presentations used to focus on graphics chips, but now Nvidia has gone beyond its roots as a gaming graphics chip maker to become a leader in AI chips, which have exploded since ChatGPT launched in 2022.

Nvidia is also developing a central processing unit (CPU) that runs Microsoft's Windows operating system, using technology from Arm Holdings.

At its annual developer conference in March 2025, Huang outlined Nvidia’s plans to meet the shift in computing needs from building large AI models to running AI-based applications. The company announced several new generations of AI chips, including the Blackwell Ultra, which is expected to launch later this year.

After the Rubin line, Nvidia plans to launch its Feynman processor in 2028. The company also launched a desktop version of its AI chip, called the DGX Spark, aimed at AI researchers. Huang said the machine is now in full production and will be ready in “a few weeks.”

Computex 2025 is expected to gather about 1,400 participating businesses and is the first major event of the Asian computer and semiconductor industry since US President Donald Trump threatened to impose tariffs to encourage companies to increase production in the US./.

(Vietnam+)

Source: https://www.vietnamplus.vn/nvidia-ra-mat-cong-nghe-ket-noi-chip-moi-de-thuc-day-phat-trien-ai-post1039599.vnp