In a groundbreaking advancement for telecommunications, NVIDIA and SoftBank have introduced the world’s first network to integrate AI and 5G capabilities, marking a significant step forward for the industry. This innovative system, known as AI Radio Access Network (AI-RAN), enables telecom networks to simultaneously run AI and 5G workloads, opening new revenue streams and optimizing network efficiency. This technology represents a potentially multi-billion-dollar opportunity for telecommunications providers, enabling them to offer services in emerging areas like autonomous vehicles, robotics, and real-time AI inference.
Setting the Stage: The First AI and 5G Network Trial
NVIDIA and SoftBank’s AI-RAN system was recently tested in Kanagawa Prefecture, Japan, where it delivered high-grade 5G performance while processing AI inference workloads concurrently. This trial showcased the network’s ability to handle extensive demands efficiently, supporting applications that require instant AI-powered decision-making at the edge, including robotics, autonomous vehicles, and multimodal generation.
SoftBank Group CEO Masayoshi Son emphasized the importance of this advancement, noting its transformative potential. He said, “Every other telco will have to follow this new wave,” hinting at a forthcoming shift in the telecommunications industry toward AI-driven, multifunctional networks.
Monetizing Excess Network Capacity with AI-RAN
Traditional telecom infrastructure is often underutilized due to capacity limitations, typically operating at only about one-third of its peak potential. AI-RAN transforms this limitation into a valuable asset. By utilizing NVIDIA’s advanced AI technology, telcos can repurpose this unused capacity for AI inference workloads, generating substantial revenue from previously idle infrastructure. According to projections from NVIDIA and SoftBank, telco operators stand to earn around $5 in AI revenue for every $1 invested in AI-RAN. SoftBank’s estimates suggest that each AI-RAN server could yield up to a 219% return on investment, making this innovation not only groundbreaking in functionality but also financially compelling.
Japan’s Most Powerful AI Supercomputer: A Step Toward AI Dominance
To further Japan’s AI capabilities, SoftBank is constructing Japan’s most powerful supercomputer, the NVIDIA Blackwell-powered DGX SuperPOD. This supercomputer, based on NVIDIA’s Blackwell platform, will serve as the foundation for SoftBank’s sovereign AI initiatives. Beyond bolstering Japan’s AI infrastructure, it positions SoftBank to cater to local demand for secure AI compute and empowers other institutions in Japan’s research, education, and enterprise sectors.
NVIDIA CEO Jensen Huang highlighted the significance of Japan’s investments in AI, noting that “Japan is leaping into the AI industrial revolution to become a global leader.” With the AI-RAN network and the supercomputer in development, SoftBank is positioning itself as a central player in Japan’s AI and technology landscape, with wide-reaching benefits for the telecommunications, healthcare, and robotics sectors.
AI-RAN’s Strategic Ecosystem and Technology Partners
The success of AI-RAN lies not only in NVIDIA’s technology but also in the strategic partnerships driving the initiative. SoftBank collaborated with major tech companies such as Fujitsu and Red Hat to bring this innovative network to life. The AI-RAN architecture relies on NVIDIA’s Aerial™ CUDA®-accelerated RAN libraries, which optimize 5G radio stacks for AI applications, while NVIDIA AI Enterprise software provides the underlying AI infrastructure.
One of AI-RAN’s standout features is its energy efficiency. SoftBank plans to incorporate NVIDIA Aerial RAN Computer-1 systems, which are estimated to consume 40% less power than traditional telecom infrastructures, further reducing operational costs while supporting higher workloads.
Real-World Applications and Economic Benefits
The implications of AI-RAN extend beyond telecommunications, offering transformative potential across various industries. During field trials, SoftBank deployed applications that require real-time, localized AI processing, such as autonomous vehicle support, robotics control, and multimodal retrieval. AI-RAN’s ability to optimize resources based on demand ensures seamless, low-latency processing, making it an ideal solution for edge AI applications.
The financial impact of this technology is equally impressive. By converting networks from single-purpose to multipurpose systems, NVIDIA estimates that AI-RAN can increase telecom revenue by five times for every dollar of capital expenditure (capex) invested. This growth potential makes AI-RAN an attractive investment for telecom operators worldwide, incentivizing further development in the space.
SoftBank’s ‘AITRAS’: A Pioneer in AI-RAN Technology
The AI-RAN system’s orchestrator, AITRAS, is a product of a five-year collaboration between NVIDIA and SoftBank. This orchestrator coordinates AI and RAN workloads, enabling efficient handling of dense data across large areas using NVIDIA’s accelerated GPU servers. As Ryuji Wakikawa, head of SoftBank’s Research Institute of Advanced Technology, remarked, AITRAS is expected to be a key driver of business model transformation within telecommunications, helping mobile operators leverage AI to meet new consumer and industry demands.
By shifting to a multi-functional AI-RAN architecture, SoftBank aims to redefine telecom’s role, from simply providing connectivity to enabling advanced AI-driven services. The deployment of AITRAS is anticipated to pave the way for telecom providers worldwide to adopt similar AI-powered infrastructures.
Toward an AI-Driven Future for Telecommunications
NVIDIA and SoftBank’s AI-RAN collaboration represents a seismic shift in telecommunications, as it provides a platform for operators to tap into the expanding market for AI services without compromising on core 5G performance. As AI and 5G technologies continue to advance, innovations like AI-RAN will likely reshape the telecom industry, allowing operators to capitalize on emerging AI applications and enhance network efficiency.
With SoftBank poised to receive the first NVIDIA DGX™ B200 systems and plans for future supercomputing platforms based on NVIDIA’s Grace Blackwell platform, Japan is positioned to strengthen its AI infrastructure significantly. This project marks a critical milestone for the country, aligning with the global push to harness AI for economic growth and societal transformation.
Through AI-RAN and its ongoing commitment to AI infrastructure, NVIDIA and SoftBank are not only redefining telecommunications but are also contributing to a broader AI ecosystem that spans robotics, transportation, healthcare, and beyond. This collaboration exemplifies the role of AI in redefining industry boundaries and shaping future technologies.
