In a grand plan to turn its cell phone base stations into a nationwide AI grid in Japan, Softbank Group is using Nvidia’s Blackwell processors to create the first artificial intelligence ( AI ) supercomputer.  ,
Blackwell, the most effective Artificial computing system yet devised, was announced by Nvidia next March. In an outdoor test in Kanagawa district, west of Tokyo, Softbank has now demonstrated continuous 5G device service for AI inference workloads.
Japan’s third-largest smart telecom carrier is then ready to create the world’s second combined Artificial and 5G telecoms network.
This new concept, known as Artificial Intelligence plus Radio Access Network, or AI-RAN, is enabled by Nvidia AI Aerial, which runs 5G radio on Compute Unified Device Architecture ( CUDA ), Nvidia’s parallel computing software for Graphics Processing Unit ( GPU) -accelerated applications.
Nvidia’s original design for computer games is an integrated circuit ( IC ) called GPUs. According to Nvidia, AI Aerial will enable mobile telecom firms to” Deploy 5G and 6G telecoms networks that can handle words, files, videos, AI, and generative AI loads on one common system”.
That is, your smart telecom service provider will henceforth be able to supplement voice, data and video traffic with low-latency, high-quality AI inferencing, drawing conclusions from information provided by phones, digital cameras, computers, robots and autonomous cars.
Rapid response is, of course, necessary for automatic driving. In cities like Tokyo, which has both a challenging road map and a high concentration of 5G base stations, Softbank’s AI grid may offer what Nvidia general Jensen Huang calls “air traffic control … for autonomous cars”.
At the Nvidia AI Summit Japan event in Tokyo on November 12 and 13, this and other exciting new initiatives were revealed.
Nvidia CEO Jensen Huang invited Masayoshi Son to speak at his keynote address, where they discussed the potential for collaboration between their businesses and the future of AI.
” The world has nothing like this”, said Huang about Softbank’s planned AI grid. ” Japan will be the world’s first”.
Son predicted that “every other telco will have to follow this new wave.” ” …this intelligence network … becomes one big neural brain … For the infrastructure of intelligence for Japan”.
To be sure, the scale of the project is unprecedented. ” Together”, said Huang,” we’re going to build Japan’s largest AI factory … When it’s built, it will have 25 AI exaflops. Just remember, the largest supercomputer in the world just recently was 1 exaflop… to produce the AI”. ( An exaflop is one quintillion floating-point operations per second. )
This, he continued,” will be distributed across Softbank’s 200, 000 sites]mobile telecom base stations ] here in Japan, serve 55 million customers … We’re also going to put on top of it a new type of store, an AI store so that the AIs that were created by Softbank and the AIs created by third parties could be provided to the 55 million customers”.
From mechatronics to AI robotics
In addition, Huang sees great potential in combining robotics and AI. ” The era of physical AI is here”, he told the audience. Japan is the only nation that I can think of that would be a better place to take the lead in the AI-robots revolution. And the reason for that is, as you know, this country loves robots … In fact, here in Japan, 50 % of the world’s manufacturing robots are built”.
” I hope that Japan will take advantage of the latest breakthroughs in artificial intelligence”, he continued,” and combine that with your expertise in mechatronics. Japan is the only nation with the highest mechatronics proficiency. You must take advantage of this extraordinary opportunity, the Nvidia CEO said.
While talking with Softbank’s Son, Huang noted,” …artificial intelligence is very different than software. You must have domain expertise and data in order to use artificial intelligence. And in robotics, Japan has world-leading domain expertise.
Leading Japanese industrial robotics firms Fanuc, Yaskawa, Kawasaki Heavy Industries, and Mitsubishi Electric are supplied by Nvidia. Additionally, it provides Rapyuta, a relatively new business that was founded by two Indian graduates from the Tokyo Institute of Technology and specializes in warehouse automation and multi-robot coordination.
Nvidia provides its high-tech wares to a wide range of Japanese industries spanning automotive, telecoms, internet service, medical equipment, pharmaceuticals, advertising and convenience stores, among many others.
Its customers are among Japan’s most prominent corporate players, including Honda, Sony, Hitachi, Denso, NTT, KDDI, Rakuten, Dentsu and Lawson.  ,
It also works closely with Japan’s academia and government, including the Institute of Science Tokyo, the University of Tokyo, Japan’s national scientific research institute Riken, Nagoya University, the National Institute of Advanced Industrial Science and Technology, the National Institute of Information and Communications Technology and the ministries of communications, education, and economy, trade and industry.
These lists show what Huang meant when he said,” …we’re here to partner with the Japan ecosystem”.
Nvidia, which was saved from bankruptcy in 1996 by an investment from computer game developer SEGA, has been in Japan for decades.
Its GPUs were used in the Sony PlayStation and Nintendo Switch in addition to SEGA Arcades. The first supercomputer to use CUDA was created by the Tokyo Institute of Technology.
For all, for one and all.
Looking ahead, Huang asked Son what he finds most exciting about Japan’s AI future.
Son replied that he is passionate about AI robotics, medical solutions and AI “agents” that can improve the Japanese lifestyle, both enterprise AI agents and personal AI agents. He thinks that “each of us ought to have its own personal agents.”
Then, Huang posed the question,” Could you picture an AI agent who has lived your entire life?” Son responded that each of us will have a personal agent from the age of one who grows up with us and is familiar with everything: a personal digital twin, a surrogate parent who can identify you when you are ill, a tutor who can recall everything it has ever taught you.
Huang and Son are not, to be sure, making any warnings about AI’s risks. They do, however, recognize the necessity for each nation to create its own AI based on the data it owns as a resource.
” It makes no sense to outsource that to somebody else”, said Huang. Every country, every company, will produce its own intelligence, its own AI.
How is it possible for a business to not develop artificial intelligence? asked Huang. That, said Son, would be like “giving your brain away to someone else”.
Follow this writer on , X: @ScottFo83517667