Lightning Cloud H100 Throughput Significantly Beats GCP
LightningAI · x · 2026-07-16
Lightning AI compared the performance differences across clouds using the same benchmark, scripts, and H100 SKU:
- Lightning: 48–49 GB/s, with all 8 GPUs participating
- GCP: 26.4 GB/s, with all 8 GPUs participating
They attribute this gap primarily to topology. The article further explains the complete tech stack of Lightning Cloud, including:
- provisioning gate
- 4 fixes upstreamed to Cloud Hypervisor
- a scheduler capable of scheduling 8-node tasks onto machines sharing the same switch
This highlights how cloud infrastructure and cluster topology significantly impact training/communication performance.
More from Infra
- Oracle Q1: Triple-digit cloud infrastructure growth as AI demand booms — DavidLinthicum · 2026-09-11
- SF Compute founder: buying compute is 'an absolutely awful experience' right now — IgorCarron · 2026-09-11
- SmolVM open-sources persistent computer infrastructure for agents that outlive chat sessions — aniketmaurya · 2026-09-11
- PyTorch Day Korea 2026 launches first offline conf, CFP closes Sept 13 — PyTorch · 2026-09-11
- Local LLM server dilemma: 4x CMP-170HX (price up 53% in 20 days) vs Mac Studio M5 Ultra — rumboll · 2026-09-11
- llama.cpp lands Flash Attention tuning for RDNA4, big prefill gains on AMD — pmttyji · 2026-09-11