AI-RAN on NPUs: Baseband Processing Without Baseband Chips

2026-07-05

AI Summary

This work investigates the feasibility of replacing dedicated baseband processors with neural processing units (NPUs) for wireless physical-layer processing. By uncovering the intrinsic structural similarity between baseband algorithms and the matrix/vector computation engines of NPUs, the authors propose a communication algorithm restructuring methodology tailored to NPU architectures. Specifically, the OFDM transceiver pipeline is reformulated into AI-native computational primitives, prioritizing engine utilization over mere arithmetic reduction. Using an Ascend 310B1 edge NPU integrated with a USRP X300 platform, the system demonstrates end-to-end over-the-air transmission in the 3.0 GHz band, marking the first complete implementation of an OFDM transceiver on a general-purpose NPU and validating the viability of wireless communication without specialized baseband hardware.

Abstract

AI-RAN aims to unify artificial intelligence and radio access network workloads on a shared compute substrate. While this paradigm has so far been demonstrated primarily on Graphics Processing Units (GPUs), it remains unclear whether Neural Processing Units (NPUs), which are AI accelerators optimized for inference, can also support wireless baseband processing. Here, we provide the first affirmative answer by resolving the fundamental mismatch between baseband workloads and NPU architecture. A computational isomorphism exists: matrix and vector engines NPUs dedicate to inference inherently cover physical-layer operations. Yet NPU architectures are natively shaped for dense-tensor AI inference, not baseband. This architectural mismatch surfaces as opposing optimization objectives: traditional baseband minimizes arithmetic operations, whereas NPU performance demands maximizing engine utilization. We close this gap by reconstructing communication algorithms onto AI compute primitives, prioritizing engine utilization over arithmetic count. We validate this with a complete OFDM transceiver on an Ascend 310B1 edge NPU, demonstrating end-to-end over-the-air transmission via USRP X300 at 3.0 GHz.

Problem

• AI-RAN
• NPU
• baseband processing
• architectural mismatch
• wireless communication

Innovation

• AI-RAN
• NPU
• baseband processing
• computational isomorphism
• OFDM transceiver