Supermicro unveiled a new line of servers for operators building out AI-augmented radio access networks (AI-RAN).
Showcased at Mobile World Congress in Barcelona, the trio of servers – ARS-111L-FR, ARS-221GL-NR, and ARS-111GL-NHR – are touted as being optimized for telecom networks and distributed AI workloads.
The vendor’s AI-RAN range is designed in line with Nvidia’s Aerial RAN Computer (ARC) architecture, which several operators and vendors going all in on AI-RAN are looking to support, like Nokia and SoftBank.
The flagship ARS-111GL-NHR features Nvidia’s GH200 Grace Hopper Superchip system, providing telecom operators with a hardware underlay to power their AI-RAN efforts. It also features a 900 Gb/s NVLink interconnect to provide high-speed chip-to-chip bandwidth for AI inference at the network edge.
The mid-tier ARS-221GL-NR also features a Grace Superchip system capable of supporting up to two double-width Nvidia RTX Pro 6000 Blackwell server edition GPUs, making it capable of handling AI workloads for telecom operators.
Rounding out the trio is the ARS-111L-FR, which is tailored for distributed RAN workloads. Loaded with an Nvidia Grace CPU C1, it’s short-depth makes it ideal for placement at the edge. It also features potential space for two low-profile GPU cards.
Among the industry players working with Supermicro at MWC are operators Telenor, SK Telecom, and Viettel High Tech.
Nokia, meanwhile, validated the ability for operators to employ its AnyRAN software suite on Supermicro’s ARS-111GL-NHR. The Finnish firm is also set to introduce validated infrastructure design for hybrid AI-machine learning training and inference based on Supermicro’s AS -8125GS-TNMR2 servers.
“Delivering AI to the RAN at scale requires infrastructure optimized for telecom networks,” Supermicro CEO Charles Liang explained. “As operators embed intelligence across their networks and advance sovereign AI strategies, Supermicro’s flexible Data Center Building Block Solutions (DCBBS) and deep ecosystem collaborations enable rapid deployment of high-performance, energy-efficient solutions that help ensure data sovereignty and long-term scalability.”
Comments