Nvidia to Sell New AI Chip Tech

Nvidia to Sell AI Chip Tech That Boosts Inter-Chip Communication

TAIPEI, May 19 (Reuters) – Nvidia (NVDA.O) said it plans to sell technology that speeds up chip-to-chip communication, essential for developing and running artificial intelligence tools.

The company unveiled NVLink Fusion, a new version of its NVLink interconnect tech, which will be available for other chipmakers to use in building powerful custom AI systems.

Marvell Technology and MediaTek have expressed interest in using NVLink Fusion for their chip projects.

Nvidia first introduced NVLink years ago to handle massive data transfer between chips, such as in the GB200 system, which links two Blackwell GPUs with a Grace processor.

CEO Jensen Huang made the announcement at Taipei Music Center ahead of the Computex AI event, and also revealed plans to build a new Taiwan headquarters.

His keynote detailed Nvidia’s history in AI hardware and software. He noted that while 90% of past presentations focused on graphics chips, today it’s all about AI.

Nvidia has transformed from a gaming GPU maker into a leader in artificial intelligence technology.

According to Reuters, Nvidia is also developing CPUs for Microsoft’s Windows that use Arm Holdings’ technology.

At last year’s Computex, Huang caused a "Jensanity" wave in Taiwan, with fans and media following his every move.

During Nvidia’s March developer conference, Huang explained how the company is shifting from AI model building to deploying applications. He introduced new chips like the Blackwell Ultra, and upcoming Feynman processors due in 2028.

The company also launched DGX Spark, a desktop AI computer for researchers, which is now in full production and shipping soon.

With over 1,400 exhibitors, Computex will be Asia’s first major chip industry gathering since U.S. tariff threats.

Don't forget to follow Telsat News to stay up to date with technology.