Dell’Oro Group’s Data Center Switch – AI Back-End Networks reports track the global market for high-performance switching infrastructure used to interconnect accelerated server clusters in AI data centers. The research focuses on scale-up and scale-out network architectures, including Ethernet, InfiniBand, UALink and NVLink, and provides detailed analysis of manufacturer revenue, port shipments, average selling prices, and vendor market share.
This research focuses on how AI back-end networks are evolving across Ethernet, InfiniBand, UALink and NVLink technologies, rapid speed transitions from 400G to 800G, 1.6T, 3.2 T and beyond, and changing deployment strategies across hyperscalers, neocloud providers, and enterprises. It also examines the role of switching silicon, SerDes advancements, co-packaged optics, and open networking platforms such as SONiC in shaping next-generation AI data center architectures.
The AI industry is undergoing a major inflection point, driven by large language models and generative AI applications that require massive clusters of GPUs, and other accelerators. These environments depend on specialized back-end networks built around scale-up and scale-out/scale-across architectures, which differ significantly from traditional front-end networks.
