SemiAnalysis goodput-expense model, Large LLM Pretrain scenario (5,184 GB300 GPUs, $4/GPU-hr): Gold-tier, hyperscaler and Silver-tier providers
6.1% / 10.5% / 20.9% goodput expensemodeled
| Value kind | modeled — Interpret this value according to its displayed kind, scope, source, and as-of date. |
|---|---|
| Scope | Calculator output for one large-pretraining scenario with stated fault-tolerance and repair-time assumptions, expressed as goodput-expense uplift. The same model gives 0.23–0.96% for multimodal RL research and about 0.5% of TCO for inference endpoints. Not a measured provider loss rate. |
| As of | 2026-04-20 |
| Source | SemiAnalysis, How Much Do GPU Clusters Really Cost? (ClusterMAX 2.1 update), 20 April 2026. |
| Review | checking…review by 2027-01-20 · standard cadence |
| Recorded changes | last 2026-10-16 · 2 revisions tracked |
| Claim id | reliability-tax-paid-gpu-hours-lost-to-failures |
Where the guide uses it
← Full numbers register — every date-stamped figure in the guide, with revision history.