Nvidia 6G Radio Chip Plan Moves AI-RAN Into Telecom Edge
Nvidia is working on a GPU-based chip for 6G radio units, extending AI-RAN into low-PHY radio processing while power, supplier integration and RAN spending remain the key tests.
Nvidia Moves AI-RAN Into The Radio Unit
Nvidia is working on a chip for 6G radio units, extending its AI-RAN strategy beyond central units and distributed units.
The proposal targets the radio unit, the antenna-side part of the RAN that has remained more dependent on custom silicon than server-based RAN compute.
The technical opening is massive MIMO.
In simpler 4G and 5G designs, much of the heavy Layer 1 processing sits in the distributed unit.
In advanced 5G radios, Layer 1 work is split between the DU and RU so beamforming and low-PHY functions can run near the antenna.
Nvidia wants a programmable GPU-based platform to take over RU tasks now handled by custom ASICs.
Nvidia says the workload is growing.
Basic radios have four transmitters and receivers.
5G Advanced and 6G designs could raise that count to 128, requiring 32 times more low-PHY processing power.
Ultra-MIMO designs in higher 6G spectrum bands could involve radio units with up to 1,024 transmitters and receivers.
Supplier Boundaries Are The Hard Part
The commercial test is whether vendors and operators will accept programmable compute in a network layer shaped by tightly linked hardware and software.
Nvidia says its CUDA platform, with about 6 million developers, can open more RAN work to software-defined development.
Its CUDA-based RAN architecture is branded Aerial.
The supplier map is unsettled.
Intel has confirmed that Granite Rapids, its latest virtual RAN product, has no radio-unit component and that it has no plan to design one.
Open RAN includes a 7.2x interface for DU-RU interoperability, but massive MIMO has rarely become commercially multivendor because software teams may need to expose closely held algorithms.
Power And RAN Spending Define The Window
Energy use remains the main objection.
Radio units can account for up to 90% of a mobile network energy consumption, and GPUs still carry a high-power reputation.
Nvidia says embedded systems in automotive and robotics can operate below 100W and at 100 degrees Celsius, making data-center GPUs the wrong comparison.
Marvell adds another signal.
Nvidia invested $2 billion in Marvell in March, while Marvell supplies RAN silicon to Samsung and Nokia.
Nokia received a $1 billion Nvidia investment and announced GPU-compatible RAN plans.
Operators worldwide spent $35 billion on RAN products last year, down from $45 billion in 2022.
The next test is whether programmable silicon can meet RU performance and power needs as custom RAN ASIC economics weaken.
Summary version for Medium
Nvidia 6G Radio Chip Plan Moves AI-RAN Into Telecom Edge
Why this hardware shift matters for AI and compute buyers
Quick Summary: Nvidia is working on a GPU-based chip for 6G radio units, extending AI-RAN into low-PHY radio processing while power, supplier integration and RAN spending remain the key tests.
Nvidia Moves AI-RAN Into The Radio Unit
Nvidia is working on a chip for 6G radio units, extending its AI-RAN strategy beyond central units and distributed units. The proposal targets the radio unit, the antenna-side part of the RAN that has remained more dependent on custom silicon than server-based RAN compute.
Supplier Boundaries Are The Hard Part
The commercial test is whether vendors and operators will accept programmable compute in a network layer shaped by tightly linked hardware and software.
- Point 1: The central topic is Nvidia, with the announcement framed around concrete operating detail.
- Point 2: Nvidia says its CUDA platform, with about 6 million developers, can open more RAN work to software-defined development.
AI deployments now depend on where models, data, approvals and operating teams meet.
Power And RAN Spending Define The Window
Energy use remains the main objection. Radio units can account for up to 90% of a mobile network energy consumption, and GPUs still carry a high-power reputation.
Read the Full Deep Dive
Want to explore the complete technical analysis, enterprise trade-offs, and detailed metrics?
Read the full article originally published at SendTech Times
