Moonshot inference may require GB200-class clusters
zephyr_z9 · x · 2026-07-20
A post arguing that Moonshot must be running GB200s to serve the model efficiently at inference time. The quote says the model would not fit on a single Hopper node, and that multi-node serving on non-NVL72 systems would be inefficient.
The attached image shows a large-scale compute rack setup labeled “Atlas 950 SuperPod” with visible specs including 256TB unified memory, 1024 GPUs, and 3-split RT. The overall takeaway is that the discussion is about the compute stack required for efficient inference at very large scale.
More from Infra
- Meta approves nonvoting Class C shares as 2026 capex guidance hits $135B–$145B — thione · 2026-10-26
- Unsloth Ships MiniMax-M3.1-Flash Quants: 2-bit Retains 91.5% Accuracy, Runs in 150GB RAM — danielhanchen · 2026-10-26
- Engineer with thousands of fabricated devices pours cold water on photonic crystal modulator yields — jwt0625 · 2026-10-26
- Anthropic S-1 reveals 48 cents of compute cost per dollar of Q2 revenue — nmasc_ · 2026-10-26
- Persistent AI joins UR network to route private AI inference with encrypted, IP-hidden traffic — 0xSammy · 2026-10-26
- Anthropic's disclosed compute commitments: $114.5B Google, $161.2B Broadcom top the list — BenBajarin · 2026-10-26