At the recent Mobile World Congress in Barcelona, Qualcomm proved that AI RAN is no longer a future-facing concept tied to AI-native 6G. There are production-ready solutions operators can deploy now on commercial infrastructure to improve network performance, increase efficiency and reduce operating costs, while also establishing the architectural foundation for the next generation of mobile networks.
The use cases Qualcomm is highlighting are practical and measurable. One is AI-driven uplink link adaptation, designed to improve uplink throughput and robustness in dynamic radio conditions. Another is AI-driven downlink beamforming channel prediction, which improves reliability and user experience, especially at the cell edge. As Qualcomm’s Nitin Agarwal, director of product management, told RCR Wireless News, beamforming predictions “will help with the cell tower edge, with the capacity, and even it’s going to improve the user experience.” Qualcomm is also applying AI to massive MIMO factory calibration, using machine learning predictions to improve manufacturing efficiency and reduce site commissioning time and cost.
Open RAN in action with Rakuten Mobile and Viettel
Agarwal emphasized field-proven capabilities of the Open RAN platforms these new RAN AI features are being introduced for. “This isn’t research, it’s in production,” he said. He pointed to Viettel’s massive MIMO deployment where Qualcomm is supporting Viettel’s 32T/32R massive MIMO infrastructure at approximately 2,000 sites. In high-traffic conditions, Agarwal said Viettel is tracking a 24% reduction in power consumption, which has a straight line to operational expense. “The operators are looking at [opex] reduction, so power is one of those things definitely where your electricity bill is going to be reduced,” Agarwal said.
In Japan, Rakuten Mobile is working to deploy around 3,000 massive MIMO radios in its ongoing 5G Open RAN build. Rakuten Mobile is using 1Finity, a Fujitsu company, equipment powered by Qualcomm’s Dragonwing QRU100 Platform.
Qualcomm believes that Open RAN’s recent commercial momentum and technology validation support the opportunity to deploy RAN AI on these O-RAN platforms.
Building an on-ramp to AI-native 6G
Qualcomm is positioning its current RAN AI work as the on-ramp to AI-native 6G, including telco compute for both centralized and distributed RAN architectures. That includes purpose-built telco server platforms for CU and DU workloads, heterogeneous compute spanning its Oryon CPUs, Hexagon NPUs and accelerators and AI engines which in combination, deliver an edge-oriented infrastructure designed to support AI-native network operations.
Looking more broadly at Qualcomm’s vision for RAN AI now and in the run up to 6G, Qualcomm EVP and GM of Technology Planning, Edge Solutions and Data Center, Durga Malladi, said in a statement: “By partnering with Qualcomm Technologies, operators gain far more than advanced technologies — they gain a trusted leader with decades of wireless innovation and proven, commercial-scale platforms that are already reshaping how networks are built and run. Our…expanding RAN AI portfolio give[s] operators the ability to unlock measurable performance, efficiency and cost improvements in the near term, while laying the critical foundation for enabling fully autonomous, AI-native networks on the path to 6G.”
