AI Inference on the Rise... The Infrastructure to Watch Is 'AI-RAN'

"AI-RAN" Minimizes Latency
Kiwoom Securities Picks Solid and Nota as Top Stocks

Amid the growing trend of artificial intelligence (AI) inference, experts have highlighted the importance of "AI-RAN," an AI infrastructure designed to minimize latency for real-time inference.

Hakjun Kim, a researcher at Kiwoom Securities, stated in a report on April 29, "With the rising number of AI agents, the flow of utilized data is changing, and the increase in video inference and real-time inference is now a leading topic in AI." He added, "This trend is expected to intensify as the era of edge devices approaches, and AI infrastructure that minimizes latency for video and real-time inference will become increasingly prominent."

Kim identified "AI-RAN" as a key example of this type of AI infrastructure. AI-RAN refers to the concept of embedding AI capabilities directly at the base station, the lowest layer of the network. Kim explained, "By seamlessly integrating AI into both RAN hardware and software, AI-RAN not only enables new AI-based services and revenue opportunities, but also dramatically improves network utilization, spectrum efficiency, and overall performance."

He further explained that deploying AI-RAN at base stations can resolve processing delays. It allows AI servers to be hosted at data centers located at mobile carriers' base stations, which are closest to users. Since communication with the server no longer requires an internet connection, this delivers lower latency and enhanced security.

The expansion of AI infrastructure goes beyond this, extending into O-RAN, CPUs, and NPUs. Kim noted, "O-RAN is a general-purpose communication backbone necessary for AI-RAN to become mainstream, and interest is growing as AT&T has decided to increase its adoption." He added, "Moreover, as real-time inference increasingly takes place on edge devices, the share of CPUs and NPUs—which previously played only a supporting role—will inevitably rise in AI inference, naturally replacing the much more expensive GPUs."

Kim recommended several stocks that could benefit from the expansion of AI infrastructure, including telecommunications equipment companies RFHIC, Solid, and LIG Nex1 Q-Cuber, as well as Nota, a company specializing in edge device optimization. Regarding Solid, his top pick, Kim said, "Solid is likely to directly benefit from O-RAN investment through its hybrid distributed antenna systems (DAS)." For Nota, his top pick in the software sector, he assessed, "As inference becomes a major trend, the company's investment direction aligns with those of hyperscalers."

This content was produced with the assistance of AI translation services.