At OFC 2026 last month， Cisco's chief architect Rakesh Chopra presented on scale-across networking architectures and key deployment trends driving strong demand for traditional DCI equipment. "Traditional" DCI connects CPUs across the frontend network while scale-across connects GPUs over the back-end to enable loss-intolerant， synchronous data flows. In scale-across networking， hyperscalers manage oversubscription of intra-datacenter bandwidth relative to inter-datacenter bandwidth using deep switch buffers and proactive congestion control. The bandwidth needs of scale-across is approximately 14x the bandwidth needs of traditional DCI. As such， significant buildout of scale-across infrastructure at various hyperscalers is expected to result in multi-billion dollar opportunities for 800G coherent pluggables， deep buffered switches and the like. SemiAnalysis's AI Networking Model will initiate estimates of scale across networking equipment spend at various hyperscalers， coming soon.