A token-cost estimate says 15T daily tokens could fit in under two SuperPods
teortaxesTex · x · 2026-08-04
A token-throughput back-of-the-envelope suggests astronomical traffic, but the metric may be misleading
The post estimates that if usage really reached 15T strictly “tokens out” per day, and if everyone had the author’s harness discipline, the implied scale would be enormous: about 3.37K trillion tokens per day, or roughly 100 quadrillion tokens per month.
The author adds two important caveats:
- The real number is probably somewhere in between.
- If the reported figure is actually “tokens processed” instead of output tokens, the economics look far less impressive.
A reply argues that in a reasonable agentic harness, 200M Flash tokens including cache hits costs about $1, implying roughly $5 per 1B and $7,500 for 15T processed tokens—“too cheap to meter” unless the metric excludes cache hits or counts only output plus cache misses.
The concluding claim is that such scale could fit in less than two SuperPods.
Related event: Debate Over Cost and Compute of 15 Trillion Tokens(2 posts)→
More from Infra
- RunPod Test: Generating 30s Video with MiniMax H3 on A100 Costs Just $0.72 — shadowtheimpure · 2026-08-04
- RTX 5090 Inference Test: Capping Power at 480W Costs Less Than 3% Performance — WonderfulEagle7096 · 2026-08-04
- DeepSeek V4 Hits 1,328 tok/s Prefill on Single RTX 6000 via Krasis — mrstoatey · 2026-08-04
- Mach-1 Additive: 35B Model Runs at 120 t/s on Laptops Using 1.7-bit Weights — pbaylies · 2026-08-04
- Surgery on open-weights models: Optimizing inference with hand-rolled Rust implementations — doodlestein · 2026-08-04
- Anthropic Locks in $10B Compute Deal with Volta, a Cloud Startup Just Months Old — The Decoder · 2026-08-04