---
title: "Nvidia's Open Wireless Tools Pass 1M Downloads, Lift Uplink 60%"
date: 2026-11-09
url: https://modevc.com/charts/nvidia-open-5g-6g-tools-million-downloads/
keywords: ["nvidia sionna ai aerial", "open source 5g 6g tools", "neural receiver uplink benchmark"]
publisher: Mode VC
---

# Nvidia's Open Wireless Tools Pass 1M Downloads, Lift Uplink 60%

Nvidia's open-source AI Aerial and Sionna tools passed 1 million downloads across 8,000 institutions, with partner trials reporting a 60% uplink gain and a 45.6% RAN power drop.

## Chart: Partner trials on Nvidia's open wireless stack report up to 60% uplink gain, 52.6% faster robot tasks, and 45.6% lower RAN power

Reported peak percentage gains across telecom and robotics trials using Nvidia AI Aerial and Sionna, November 2026

| Trial benchmark | Reported improvement |
|---|---|
| SoftBank neural receiver uplink throughput gain | 60% |
| King's College London robot task time reduction | 52.6% |
| University of Leeds RAN power consumption drop | 45.6% |

Note: Figures reported by Nvidia on 6 Nov 2026. SoftBank measured maximum uplink throughput gain across 15 interference settings; King's College London measured robot task completion time reduction; University of Leeds measured RAN power consumption drop on a DGX Spark testbed.

## The news

- 2026-11-07: [Nvidia says open-source 5G and 6G tools passed 1 million downloads](https://www.siliconreport.com/nvidia-says-open-source-5g-and-6g-tools-passed-1-million-downloads) (siliconreport.com)

## What happened

On 7 Nov 2026, Nvidia reported in a blog post dated 6 Nov 2026 that its open-source wireless research platforms, AI Aerial and Sionna, have surpassed 1 million downloads. More than 8,000 research and academic institutions globally now use the tools, and developer adoption expanded tenfold over the past year.

The release highlighted three measured engineering benchmarks built on the open software stack. SoftBank integrated a neural receiver into Aerial CUDA-Accelerated RAN and recorded a maximum uplink throughput gain of roughly 60% over conventional receivers across 15 interference test settings.

In academic trials, King's College London reduced robot task completion times by up to 52.6% in fault-tested manipulation experiments by transmitting compact object descriptions instead of periodic images. Separately, University of Leeds researchers demonstrated a 45.6% drop in radio access network power consumption using an AI demand-prediction setup on a DGX Spark testbed, while the Centre Tecnològic de Telecomunicacions de Catalunya cut simulation runtimes from about three hours down to 10 to 20 minutes.

## What the chart shows

The chart plots the three percentage improvements reported across the telecom and robotics trials: a 60% maximum uplink throughput gain at SoftBank, a 52.6% maximum reduction in robot task completion time at King's College London, and a 45.6% reduction in radio access network power consumption at the University of Leeds.

Each figure measures a different physical metric in a separate testbed, so the bars illustrate the breadth of workloads running on AI Aerial and Sionna rather than a head-to-head benchmark on a single dataset. SoftBank's 60% figure is a maximum uplink gain across 15 interference settings, and King's College London's 52.6% figure is an upper bound in manipulation trials.

All figures are reported by Nvidia from partner and university testbeds rather than audited by an independent standards body. The underlying software packages remain publicly available on GitHub, while the hardware execution layer relies on CUDA-accelerated RAN and DGX Spark systems.

## What it means for founders

Nvidia is running the classic complement commoditization playbook in telecom. By giving away the simulation and ray-tracing layer for free on GitHub to over 8,000 academic and research institutions, Nvidia ensures that the next generation of wireless algorithms is written natively against CUDA-Accelerated RAN and DGX Spark hardware.

When graduate students and telecom engineers cut simulation runtimes down to 10 to 20 minutes on Sionna, their workflows, test harnesses, and neural receiver weights become coupled to Nvidia's compute primitives. By the time a carrier like SoftBank moves a neural receiver from lab simulation into commercial radio access networks, the hardware vendor selection is already locked in upstream.

Founders building vertical engineering or simulation tools should study this wedge. Giving an open-source simulation library to university labs is not charity; it is a zero-CAC enterprise distribution channel that plants your runtime inside commercial buyers years before procurement RFPs are written.

## What would change our mind

We would change our mind if telecom carriers deploy the neural receivers and demand-prediction models trained in Sionna onto non-CUDA baseband ASICs from competing silicon vendors without performance loss or porting friction.

If hardware-agnostic compilers allow carriers to capture the 60% uplink gain and 45.6% power reduction on commodity radio hardware, then open-sourcing the simulation layer will have subsidized the broader industry without locking in hardware demand. Until open compiler toolchains close that gap, owning the university simulation standard remains one of the highest-leverage infrastructure moats in deep tech.

> Free open-source simulation across 8,000 institutions is Nvidia's upstream distribution wedge to lock telecom workloads onto CUDA hardware.

**What we would back:** Developer and simulation toolmakers that seed open-source standards in academic labs and monetize the production deployment runtime.

## Sources

- [siliconreport.com: Nvidia says open-source 5G and 6G tools passed 1 million downloads](https://www.siliconreport.com/nvidia-says-open-source-5g-and-6g-tools-passed-1-million-downloads)