Nvidia, telecom firms partner to develop AI-powered 6G networks

Nvidia said it has formed an alliance with telecom firms including Nokia Oyj, SoftBank Group, and T‑Mobile US to promote 6G networks that support AI.

The announcement was timed for the opening of a telecom industry conference in Barcelona.

Nvidia said the members will commit to building 6G networks using software-controlled radios running on general-purpose computers that use AI to help direct radio traffic safely and efficiently.

Nvidia said 5G was designed to connect people via voice and data and isn’t capable of supporting the widespread use of AI.

Ronnie Vasishta, who heads Nvidia’s telecommunications business and strategy, said networks will need far greater efficiency and AI can help route traffic safely and efficiently.

He added that telecom networks would require “hundreds of thousands of times” more efficiency because there isn’t enough radio spectrum to support new uses.

Nvidia already offers versions of its chips, computers, and software for networks and said the alliance could help expand AI demand beyond data centers.

Nvidia said the target 6G architecture is open radios controlled by software on general-purpose computers to allow new providers to enter the market.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Field trials put Nvidia’s AI-RAN plan to work

The program centers on AI-RAN, backed by its NVIDIA AI Aerial platform 1.
Field deployments are underway, beyond lab setups 2.
T-Mobile US completed an over-the-air call using the system. The setup ran radio access network (RAN) Layer 1 workloads with graphics processing unit (GPU) acceleration alongside AI applications on the NVIDIA Grace Hopper platform 3.
SoftBank Corp. ran a live field trial that reached a milestone. It used fully software-defined 5G on NVIDIA’s AI-RAN platform 2.
The project targets networks that can support ’physical AI’ such as autonomous cars and robots. Nvidia says this shift raises expectations for security and trust in future networks 4.

An open 6G push could still depend on Nvidia

The company’s open-network message lines up with OCUDU (Open CU DU), a US government-linked open-source effort for 6G core RAN components that seeks to put open-source code into the center of 6G networks 5.
A software-defined route can undercut traditional telecom equipment vendors, while adding a fresh dependency 5.
Nvidia’s Aerial platform includes an open-source reference platform for RAN, yet most work still sits inside CUDA (Compute Unified Device Architecture). Parts also cannot run on general-purpose central processing units (CPUs), which can lock teams into its hardware and software stack 5.
The approach could recast telecom operators as distributed AI cloud providers. Nokia described a demo where unused GPU capacity in a distributed AI-RAN network could be sold as AI compute to external customers 3.