AI RAN Orchestration:

A practical step toward shared infrastructure

Download now

Architecture, trade-offs, and considerations

AI-driven applications, ranging from immersive experiences to real-time decision systems place new demands on latency, uplink capacity, and orchestration across radio, transport, and compute. This paper examines how RAN and AI can coexist within a shared architectural framework, focusing on automation, orchestration, and lifecycle management using a concrete orchestration use case developed with SoftBank and implemented on Ericsson technology.

Report

#rApps #Orchestration #AI-RAN

Dynamic allocation of compute resources between RAN-related workloads and other edge AI applications, enabling more efficient hardware utilization and new revenue opportunities

The ability for rApps AI models and non-RAN AI workloads to share edge infrastructure in a controlled manner

SoftBank’s AI orchestrator integrated with Ericsson’s SMO can leverage R1 interfaces within the O-RAN architecture to coexist and collaborate with other workload orchestrators through loose coupling

The concept of AI-and-RAN involves using a shared information technology infrastructure to host both AI workloads and RAN workloads, creating opportunities to improve infrastructure utilization and flexibility. As AI workloads and RAN workloads increasingly intersect at the infrastructure level, several considerations remain open, spanning both business and technology dimensions. A key question is under what conditions shared infrastructure delivers sufficient value to justify additional complexity, including understanding when improvements in utilization, cost efficiency, or service enablement translate into tangible benefits, and when simpler deployment models remain preferable.

Download the report

AI RAN Orchestration: A practical step toward shared infrastructure

Download now

AI RAN Orchestration:

Architecture, trade-offs, and considerations

Key achievements:

Why this matters