According to both companies, SoftBank’s telecommunications division aims to create Japan’s most powerful AI supercomputer to enhance various local services. This supercomputer will be built on Nvidia’s DGX B200 platform, which integrates traditional processors with specialized AI accelerator chips. A subsequent initiative will incorporate the more sophisticated Grace Blackwell.
On Wednesday, shares of SoftBank Corp., Japan’s third-largest wireless provider, rebounded, increasing by 1%. Meanwhile, SoftBank Group, which includes the conglomerate’s investment divisions, saw a decline of 1.7%.
Nvidia’s chips have become highly sought after by leading technology firms, which rely on these components to develop and execute AI models. This process necessitates extensive data input, a task that accelerator chips excel at managing.
This announcement suggests that SoftBank, which previously held a 4.9% stake in Nvidia until early 2019, has positioned itself advantageously for acquiring these chips. Founder Masayoshi Son has expressed his intent to make significant investments in AI. On Tuesday, the company reported a return to profitability, driven by increasing technology valuations.
Earlier this year, Nvidia CEO Jensen Huang unveiled the new Blackwell series, although production delays hindered its launch. Huang has assured that supply will be abundant once manufacturing scales up, but demand for the new chips remains high among customers.
During Nvidia’s AI Summit in Tokyo on Wednesday, Son joined Huang on stage. When reminded of SoftBank’s former Nvidia stake, which would now be valued at approximately $178 billion, Son reacted with a laugh and a cringe, before embracing Huang warmly.
Nvidia has been actively traveling around the world to organize events that promote what it refers to as the new industrial revolution. Recent gatherings in India and Japan are focused on expanding the implementation of AI systems within national initiatives, while also reducing Nvidia’s dependence on a limited number of major US clients.
Alongside the introduction of a new computer and plans for a second model, SoftBank’s telecommunications division will utilize Nvidia technology to deliver AI services via cellular networks. Conventional hardware, which relies on custom chips optimized for mobile data traffic, is not suitable for the emerging AI services.
Huang stated, “This will lead to the creation of an AI grid throughout Japan,” emphasizing that it will transform the existing communications infrastructure into an AI-driven network.
The new AI radio access networks (AI-RANs) will be more effective for applications such as remote robotics, support for autonomous vehicles, and other services, according to Huang. Additionally, they will consume less power.
The telecom division is set to begin network testing in collaboration with partners Fujitsu Ltd. and Red Hat, a subsidiary of International Business Machines Corp.
Son expressed to Huang, “We are planning to purchase a significant amount of your chips.”
Japan stands on the brink of transformation, according to the Japanese billionaire, who has frequently criticized the government and corporations for their slow adoption of new technologies. He noted that this time, the Japanese government is facilitating advancements in AI and robotics.
The Japanese government has committed ¥4 trillion ($26 billion) to enhance its domestic semiconductor production capabilities. This includes an ambitious initiative involving Rapidus Corp. to establish a cutting-edge foundry from the ground up to compete with Taiwan Semiconductor Manufacturing Co., which produces Nvidia chips. Prime Minister Shigeru Ishiba has also promised over $65 billion in new investments for the semiconductor and AI industries.
“However, they should do more,” Son advised Huang. “As you mentioned, this is a pivotal moment for a reset and a chance to catch up in this revolution. We cannot afford to miss this opportunity.”