AnalysisDevelopersSeptember 24, 2026

H3C pitches token efficiency over more GPUs for agentic AI

At the 2026 Apsara Conference, H3C argued that adding GPUs no longer yields proportional gains as trillion-token agent workloads emerge. Its UniPoD S80000 SuperPod scales from 32 to 1,024 GPUs, and up to 16,384, supporting CPUs, GPUs, NPUs and DPUs.

1 source

More stories today

Open the live feed