news 
The AI Network Is The Computer, Says Nvidia
In a recent conversation with Bob Metcalfe, inventor of the Ethernet standard, Jensen Huang, Nvidia’s co-founder and CEO, expressed his vision for the future of computer applications. According to Huang, this future involves “part of the application runs in the data center, another part in a data center at the edge, and another part in an autonomous machine roaming around the world.”
Nvidia is taking significant steps towards realizing this vision by developing its own networking platform called Spectrum-X. This platform, which uses standard Ethernet protocols but operates under the hood to provide extreme high-performance connectivity, has been used to build the largest AI supercomputer in the world.
Deierling explains that due to the unique nature of data processing for AI, the capabilities of the network are critical. He highlights that traditional cloud computing is different from AI workloads. Cloud computing serves millions of users, each transferring a small amount of data, and this data is completely unsynchronized. On the other hand, AI—and Nvidia’s processing units or GPUs—do things in parallel.
Source: www.forbes.com
