SoftBank will be first to receive Nvidia’s Blackwell chips to build supercomputer

SoftBank’s (OTCPK:SFTBY) telecom unit will receive the world’s first AI system featuring Nvidia’s (NASDAQ:NVDA) Blackwell chips to build Japan’s most powerful supercomputer, and plans to use the Grace Blackwell platform for its next supercomputer.

The announcements were made by Nvidia’s (NVDA) CEO Jensen Huang during his keynote at the company’s AI Summit in Japan.

SoftBank Corp. (OTCPK:SOBKY) will use its DGX SuperPOD supercomputer – built using the Blackwell-powered DGX B200 system – for its own generative AI development and AI-related business, as well as that of universities, research institutions and businesses in Japan.

SoftBank (OTCPK:SOBKY) plans to build its next supercomputer – using Nvidia’s (NVDA) more advanced Grace Blackwell design – to run extremely compute-intensive workloads.

The Japanese telecom giant also successfully piloted the world’s first combined AI and 5G telecom network, called AI radio access network (AI-RAN), using the Nvidia AI Aerial computing platform. This breakthrough is expected to open AI revenue streams potentially worth billions of dollars to telecom operators.

Nvidia (NVDA) and SoftBank (OTCPK:SOBKY) estimate that telco operators can earn about $5 in AI inference revenue from every $1 of capex it invests in new AI-RAN infrastructure. SoftBank expects to achieve a return of up to 219% for every AI-RAN server it adds to its infrastructure.

SoftBank (OTCPK:SOBKY) also plans to create an AI marketplace using Nvidia AI Enterprise software to meet AI computing demand and maintain AI-RAN’s performance. The service – dubbed AITRAS – will support AI training and edge AI inference (running trained AI models directly on hardware).