SoftBank Unveils AI Model To Streamline Telecom Operations

SoftBank Corp. SFTBF SFTBY has introduced a new generative AI model called the Large Telecom Model.

The Japanese investment firm designed LTM to improve the design, management, and functioning of cellular networks.

Built on extensive datasets from SoftBank's network and years of operational expertise, the LTM is set to optimize telecom operations. The company intends to continue its research and development to incorporate the model into its own operations.

These AI models can accurately predict configurations for real base stations with more than 90% accuracy.

Also Read: NIO And CATL To Build China's Largest Battery Swapping

This process reduces the time needed to make adjustments from days to minutes, while also minimizing human error. Furthermore, the LTM serves as the core of SoftBank's "AI for RAN" initiative, improving Radio Access Network (RAN) performance.

SoftBank’s partnership with Nvidia Corp NVDA was essential in enhancing the LTM’s performance. By utilizing Nvidia’s NIM for fine-tuning, SoftBank achieved substantial improvements in key performance metrics. That includes Time to First Token (TTFT) and Tokens Per Second (TPS).

The LTM incorporates SoftBank’s extensive mobile network data, providing valuable insights and expert knowledge to solve network issues effectively.

Through fine-tuning, LTM allows the development of AI models designed for specific use cases in mobile network operations, such as optimizing new base station deployments and adjusting configurations for special events.

SoftBank’s use of Nvidia’s DGX SuperPOD for distributed training and its continued collaboration on AI inference microservices will enhance LTM’s capabilities and its potential deployment across networks.

SoftBank aims to leverage the LTM to improve operational efficiency, deliver new services, and enhance mobile network experiences.

It is to be noted that the Trump Administration has revealed a new AI infrastructure initiative in the U.S. called “Stargate,” which includes OpenAI, SoftBank, Oracle Corp ORCL, and Abu Dhabi's MGX. The venture plans to invest $100 billion initially, with the potential to reach up to $500 billion over the next four years.

Price Action: SFTBF shares closed at $51.95 on Tuesday.

Read Next:

Hyperfine, Nvidia Partner To Make Brain MRIs More Affordable

Image: Shutterstock