AT&T & Ericsson: AI-Powered Cloud RAN Boosts Network Performance

AT&T and Ericsson have demonstrated a significant advancement in cloud-based radio access network (RAN) technology, successfully testing Ericsson’s AI-native Link Adaptation software on a Cloud RAN stack powered by Intel’s Xeon 6 System-on-Chip (SoC). The demonstration, conducted over AT&T’s licensed spectrum, marks the first instance of production-grade Ericsson RAN software operating on Intel’s new silicon, according to a statement released by Ericsson.

The AI-native Link Adaptation dynamically adjusts to fluctuating signal quality and interference, enhancing RAN performance on both purpose-built and cloud-based infrastructure. The test results showed a potential throughput increase of up to 20% compared to traditional, rule-based link adaptation methods, as highlighted in an Intel fact sheet released during Mobile World Congress 2026.

This milestone is particularly notable because it decouples RAN software from specific hardware vendors. According to a report from techblog.comsoc.org, this separation allows AT&T to integrate artificial intelligence capabilities and update network functions more rapidly, circumventing the limitations imposed by vendor lock-in. The portability of the Ericsson RAN software to the Intel Xeon 6 SoC represents an industry first.

Intel’s Xeon 6 SoC is designed for high-performance, cloud-native processing of RAN workloads and incorporates integrated Intel Advanced Vector Extensions (AVX) and Intel Advanced Matrix Extension (AMX) technologies. These features eliminate the need for separate accelerator cards, optimizing capacity, efficiency, and total cost of ownership, according to Intel.

The collaboration between AT&T, Ericsson, and Intel aims to build a more open, programmable, and reliable RAN network. The demonstration highlights the potential of Cloud RAN to deliver improved network speed, flexibility, and performance, leveraging the power of AI and advanced silicon technology.