Ericsson on June 15 unveiled 'AI in RAN', a software solution that applies carrier-grade artificial intelligence models to baseband and radio equipment at base stations to raise 5G network performance and operating efficiency.

Ericsson explained that AI in RAN can improve performance, capacity and energy efficiency of existing 5G networks without adding separate hardware. It is a strategy to accelerate a shift to an AI-native RAN by applying AI directly to the RAN so the network can assess the communications environment in real time and optimise it.

AI in RAN is based on carrier-grade AI models that run in real time in the RAN environment of base stations. It applies a software architecture that continuously learns using scalable high-quality data and uses agentic AI to raise network automation and operating efficiency.

Ericsson's carrier-grade AI models support ultra-low-latency inference at the microsecond level. They are designed so AI can analyse data in real time and adjust network resources even when radio conditions and user traffic change quickly.

The solution links with 5G Advanced on Ericsson's dedicated hardware-based platform and its Cloud RAN platform. Using Ericsson's own silicon, it processes AI inference on radios and the latest RAN compute equipment, and on Cloud RAN it can flexibly deploy AI functions on partner platforms as well.

The first function will be applied in phases from the second quarter of this year. Key functions include an AI-native scheduler that adjusts transmission methods according to radio link conditions, AI-based macro positioning, AI-based beamforming, AI-based multi-carrier management, and performance management event data and granular statistical indicators.

Ericsson explained that it confirmed performance improvements through more than 15 commercial network deployments and trial applications worldwide. Download throughput rose by up to 20 percent and spectrum efficiency increased by up to 10 percent. In environments with concentrated traffic, it can accommodate up to twice as many users as existing systems. Coverage prediction accuracy reached 90 to 95 percent. User positioning precision improved by up to five times.

In South Korea, SK Telecom is working with Ericsson to advance AI-RAN technology. The two companies are conducting research and development and commercial-network demonstrations to raise network performance and energy efficiency and build an intelligent and automated network operating environment.

Ryu Tak-gi (류탁기), head of network technology at SK Telecom, said, "We are advancing AI-RAN technology in cooperation with Ericsson and maximising network performance and energy efficiency." He said, "Through research and development, real network demonstrations and software innovation, we will lay the foundation for the AI-native 6G era."

Morten Lerner, head of network strategy and product management at Ericsson, said, "Ericsson is redefining what is possible for mobile networks by applying powerful AI capabilities to telecommunications operator networks."