Huawei's 100K-card Super Cluster can train a 10T-param model on 100T tokens in 30 days
teortaxesTex · x · 2026-09-18
Per @tphuang, Huawei's 100,000-card Super Cluster—built from 25 Atlas-960 SuperNodes—can train a 10-trillion-parameter model on 100T tokens in 30 days.
- The same cluster can be partitioned into 352-card instances to serve inference for 10T-param models
- Expandable to 1M cards, with a large CPU SuperPoD and enterprise storage included
A significant datapoint in the race for hyperscale AI compute.
More from Infra
- Google Open-Sources Agent Substrate on GKE: 10x Density, 1,000+ Dormant Agents per Host — blaizedsouza · 2026-09-18
- Redditor crams six V100 GPUs into a standard full-tower case for local LLM inference — Odd_Caterpillar_2994 · 2026-09-18
- Crusoe raises $3.9B at $30.9B valuation to build data centers and modular AI factories — TechCrunch AI · 2026-09-18
- A 2.5-hour first-principles primer on the semiconductor supply chain worth your time — blaizedsouza · 2026-09-18
- YC F26's Dreamscale Labs Moves Robot AI Inference to the Cloud — ycombinator · 2026-09-18
- Community fine-tunes an MTP head for Bonsai 2 27B, ~1.25x inference speedup — cephaloform · 2026-09-18