Unlock the power of scale-across with Cisco converged silicon, systems, and optics
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
Generative AI training clusters have outgrown single data centers due to power constraints, forcing GPU workloads to span multiple geographically distributed sites. This creates a new networking category called 'scale-across' — distinct from traditional DCI — that must sustain lossless, low-latency, high-bandwidth synchronous traffic between tens of thousands of GPU ports across distances of 10+ km. Scale-across demands co-design of silicon, systems, and coherent optics to address deep buffering, proactive congestion control, power efficiency, and reliability. Cisco's response includes the Silicon One P200 chip (51.2 Tbps), 400G/800G ZR/ZR+ coherent pluggables (750K+ ports shipped), and new open line systems like the Open Transport 3000 Series, with two hyperscaler design wins announced for fiscal Q3 2026 and AI order forecasts exceeding $3.6B in Q4 2026.
