NVIDIA's open 5G/6G tools pass 1 million downloads as researchers test 6G ideas on real hardware

NVIDIA published a status report on November 6, roughly a year after it open sourced its Aerial wireless software. Adoption of NVIDIA AI Aerial and NVIDIA Sionna has grown 10x over the past year, its open source 5G/6G tools have passed 1 million downloads, and more than 8,000 research and academic institutions now use the platforms.

The pitch behind the toolchain is continuity. Researchers can move between simulation, AI training and real over-the-air experiments on one foundation instead of stitching together separate stacks. NVIDIA frames the goal as AI-native 6G, in which connectivity, computing and sensing run on a single software-defined platform serving people, machines and agents.

King's College London supplied the robotics example. Researchers attached a camera-equipped robot arm to an NVIDIA Jetson AGX Orin running Sionna, which handles both the software-defined 5G link and edge AI processing. Rather than sending full images, the robot transmits compact descriptions of objects, their positions and their relationships, an approach called goal-oriented semantic communications.

In fruit-packing and cube-stacking trials with deliberately injected faults, that cut task completion time by up to 52.6% and improved success rates versus sending periodic images. The setup also detected dropped objects, recovered from failed actions and, with a digital twin, checked planned moves for collisions before execution.

On the receiver side, SoftBank integrated a neural receiver into NVIDIA Aerial CUDA-Accelerated RAN and tested it with a commercial 5G smartphone. Under interference it delivered higher uplink throughput than a conventional receiver at all 15 tested settings, with a maximum gain of roughly 60%. ETH Zurich took a different route using the Aerial Testbed, tuning neural receivers on local measurements to cut block error rates by more than a third against selected pretrained versions.

Energy is the third theme. A University of Leeds team predicts network demand 10 minutes ahead and uses the forecast to decide when base stations can sleep, checking first that neighbouring active stations can still serve affected devices. Tested on a DGX Spark-powered portable 5G network-in-a-box alongside a digital twin, the approach cut RAN power consumption by 45.6% against a full-power baseline while keeping devices connected.

Simulation speed improved too. CTTC brought GPU-accelerated radio channel models running on ACAR into its open source 5G-LENA simulator, reducing runtime from about three hours to 10 to 20 minutes in an early evaluation while preserving expected statistical behaviour and network performance.

The projects are still research results, well short of a commercial 6G standard, but they show the open toolchain moving ideas from papers into reproducible over-the-air experiments. NVIDIA said it will host 6G developer sessions at GTC in Washington, D.C.