Ericsson, Nokia, NVIDIA, T-Mobile US establish the AI RAN Innovation Center
As T-Mobile US CEO Mike Sievert pointed out during the carrier’s recent Capital Markets Day, delivering cellular connectivity is “harder than it looks.” As such, T-Mo is looking to leverage artificial intelligence (AI) to optimize its radio access network (RAN) in partnership with legacy telecoms kit providers Ericsson and Nokia, and accelerated computing powerhouse NVIDIA. In addition to bringing AI to bear on the RAN, Sievert and NVIDIA CEO Jensen Huang also discussed the opportunity for fallow accelerated computing resources deployed in a cellular network to create a new revenue stream.
Huang, in a fireside chat with Sievert, said NVIDIA has already shipped the new ARC-1 supercomputer to Bellevue, Washington, where it will be used in the AI RAN Innovation Center, jointly founded by Ericsson, Nokia, NVIDIA and T-Mobile US. “We have the ability to do two things,” Huang said. First, the distributed AI can be used for multi-dimensional network optimizations; second, AI-equipped base stations can host third-party workloads, similar in concept to the GPU-as-a-Service models being explored by Singtel and SK Telecom.
Sievert said the big goal of the AI RAN collaboration is to augment and improve existing RAN transformations around hardware/software disaggregation and virtualization. “If we can be co-authors of this transformation, we can have our engineers in there deep…working with the world’s leaders like NVIDIA, like Ericsson, like Nokia, to create the future.” And, he said, “T-Mobile customers will benefit disproportionately.”
NVIDIA AI Aerial hardware/software for “designing, simulating, training and deploying” AI RAN
On Sept. 18, NVIDIA announced its “full suite” AI Aerial platform for “high-performance, software-defined RAN along with training, simulation and inference so that telecom operators can participate at any stage of development to deployment for next-generation wireless networks.” This includes RAN-specific software libraries on NVIDIA’s CUDA platform, PyTorch and TensorFlow libraries with AI RAN frameworks to support “spectral efficiency” and signal processing for 5G and 6G. NVIDIA Sionna, “a link-level simulator that provides development and training of neural network-based 5G and 6G radio algorithms, is also part of the package. And NVIDIA Aerial Omniverse Digital Twin provides scalable digital twin capabilities to simulate real world network and user scenarios.
This solution will be used in the AI RAN Innovation Center. Nokia’s President of Mobile Networks Tommi Uitto said in a statement, “By bringing together leading companies in the telecom and AI industries, we can unlock the full potential of AI in our networks, improving performance, reducing costs and creating new opportunities for customers.”
Ericsson EVP and Head of Business Area Networks Fredrik Jejdling said, “Ericsson has invested in our AI RAN solution, allowing [operators] to deploy portable RAN software running across multiple platforms. We are now evaluating the performance and cost of NVIDIA accelerated computing in this context.”
