Optical interconnects seen entering AI racks in 2028

The analyst's statement reflects a broader industry recognition that copper-based interconnects are reaching their performance limits as AI data centers scale up. Large language models and other compute-intensive AI workloads generate enormous volumes of data movement between processors and memory—a load that traditional electrical signaling over copper cannot efficiently handle. Optical fiber, by contrast, offers dramatically higher bandwidth and lower latency, making it the natural next step as AI clusters grow larger and more demanding. The 2028 projection suggests this is not imminent but represents a near-term inflection point when the economic and technical case for optical deployment becomes compelling enough to justify the infrastructure transition.