Nvidia plans to sell tech to speed AI chip communication
Nvidia announced NVLink Fusion, a new chip-to-chip communication technology aimed at accelerating AI development, during Computex 2025 in Taipei, with partners like Marvell, MediaTek, Alchip, Fujitsu, and Qualcomm adopting it. The company also revealed plans to expand its Taiwan headquarters and disclosed $5.5 billion in charges due to U.S. restrictions limiting its China sales, including $15 billion in lost revenue from the H20 chip ban.
Nvidia introduced NVLink Fusion, a technology designed to improve chip-to-chip communication for AI systems, during the Computex AI exhibition in Taipei from May 20 to 23. The company plans to sell NVLink Fusion to chip designers, including Marvell Technology, MediaTek, Alchip, Fujitsu, and Qualcomm, to help build high-performance custom AI systems. CEO Jensen Huang announced the technology at the Taipei Music Center, highlighting its role in linking multiple chips for faster data exchange, such as in Nvidia’s GB200, which combines two Blackwell GPUs with a Grace processor. Huang also disclosed that Nvidia has taken a $5.5 billion charge due to U.S. export restrictions, which forced the company to forgo $15 billion in potential sales in China, including the H20 chip. Nvidia’s expansion into AI hardware began with its graphics chips but has since dominated the AI boom following ChatGPT’s 2022 launch. The company is now developing new AI processors, including the Blackwell Ultra (available later this year) and Feynman chips (set for 2028), alongside a desktop AI system, DGX Spark, which Huang said would be in production within weeks. Huang also announced plans to establish a Taiwan headquarters in Taipei’s northern suburbs, reinforcing Nvidia’s commitment to the region. The Computex event, featuring 1,400 exhibitors, marks the first major Asia-focused tech gathering since U.S. President Donald Trump threatened tariffs to encourage domestic production. Nvidia’s strategy now focuses on transitioning from AI model development to application deployment, as outlined in its March developer conference. The company continues to innovate beyond graphics, including Windows-compatible CPUs using Arm Holdings technology, while maintaining its leadership in AI infrastructure.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.