“`html
Nvidia Introduces NVLink Fusion Technology at Computex
Nvidia has announced NVLink Fusion, a groundbreaking technology aimed at improving chip-to-chip communication, which is vital for developing artificial intelligence tools, according to reports from Reuters.
The company intends to extend NVLink Fusion to various chip manufacturers, allowing them to create bespoke AI systems by interconnecting multiple chips. This announcement was delivered by Nvidia CEO Jensen Huang during his keynote presentation at the Taipei Music Center, the venue for the Computex AI exhibition that took place from 20 to 23 May.
As reported by Reuters, companies such as Marvell Technology and MediaTek are anticipated to integrate NVLink Fusion into their forthcoming chip designs. Nvidia also identified Alchip, Fujitsu, and Qualcomm as its partners in this venture.
Currently, NVLink technology is employed to enable rapid data transfer between chips within Nvidia’s systems, including the GB200 platform, which merges two Blackwell GPUs with a Grace CPU.
Nonetheless, Nvidia encounters challenges in the Chinese market due to US government-imposed restrictions on exporting Nvidia’s H20 chips. As mentioned by Reuters, Nvidia indicated last month that it would face $5.5 billion in costs as a consequence of these export controls.
In addition to this new product, Nvidia disclosed plans to establish a new headquarters in the northern suburbs of Taipei, Taiwan.
During his keynote, Huang also reflected on the company’s growth, emphasising Nvidia’s evolution from a gaming-centric GPU manufacturer to a leading player in the AI surge that commenced with ChatGPT’s launch in 2022. He mentioned that the company is now working on CPUs capable of running Microsoft’s Windows OS, utilising technology acquired from Arm Holdings.
“`
