SoftBank plans to collaborate with Nvidia on AI supercomputers, old love is still the most beautiful

Bloomberg reports that SoftBank's telecommunications division is planning to build Japan's most powerful AI supercomputer. This computer will be built on Nvidia's DGX B200 series products, combining computer processors with AI accelerator chips, making it the first supercomputer to use Nvidia's new Blackwell chip.

Bloomberg reported that SoftBank's telecommunications division plans to build Japan's most powerful AI supercomputer, which will be based on Nvidia's DGX B200 series incorporating computer processors with AI accelerator chips, making it the first supercomputer to use Nvidia's new Blackwell chip.

Old love is the most beautiful. Initially, Masayoshi Son abandoned Nvidia, but today SoftBank has returned to the embrace of Jensen Huang, promising to buy more chips. During the Nvidia AI Summit this Wednesday, SoftBank's founder Masayoshi Son and Jensen Huang took the stage together in Tokyo. Reflecting onSoftBank selling off 4.9% of Nvidia's stake for only $4 billion at the time, which is now valued at nearly $178 billion, Masayoshi Son made a gesture of crying in Jensen Huang's arms, creating a theatrical effect.

Jensen Huang, wearing his favorite black leather jacket, told Masayoshi Son that he would have been Nvidia's largest shareholder, but later sold off most of his shares. While Son pretended to cry, Huang also choked up and said, "Let's cry together," although both successful individuals joked about shedding tears.

Nvidia CEO Jensen Huang earlier announced the new Blackwell product lineup, but various geopolitical and economic obstacles have slowed down the launch. Huang stated that once production increases, supply will be sufficient, and global customers and investors are eagerly anticipating the launch.

In addition to the supercomputer project, SoftBank's telecommunications division will also use Nvidia's AI equipment to provide artificial intelligence services through traditional Cellular Networks, and the new AI Grid network will be deployed throughout Japan.

The Japanese government has allocated 4 trillion yen, approximately $260 billion, to enhance its domestic chip capacity. Rapidus Corp. will build a state-of-the-art foundry from scratch to challenge Taiwan Semiconductor Manufacturing Company in producing Nvidia chips. Prime Minister Shigeru Ishiba also pledged over $65 billion in new support for Japan's semiconductor and artificial intelligence industries.

The new artificial intelligence network AI – RAN, also known as AI Radio, will be more suitable for remote robots, autonomous driving cars, and other services, requiring less power. SoftBank's telecommunications division will begin testing the network with Fujitsu Limited and IBM's Red Hat.

Nvidia once again secures a victory, as news of Japan beginning to build an AI chip foundry suggests that chip manufacturing orders may once again shift from Taiwan to elsewhere. In the global chip resource battle, another superpower has joined the fray.