TPU cost per million tokens beats NVIDIA Blackwell, giving Google an edge
cgarciae88 · x · 2026-10-04
Investor SouthernValue95 compared per-million-token inference costs of Google TPU versus NVIDIA Blackwell, concluding TPU has the advantage. With GCP doubling down on TPUs, the emerging third-party TPU ecosystem is worth watching.
More from Infra
- Local LLM benchmarks are mostly noise: c=1 tokens/s hides real concurrency performance — TheZachMueller · 2026-10-04
- Ahmad Osman on MTS Live: how local AI is becoming competitive, DGX Station in tow — basedjensen · 2026-10-04
- Building a Miro clone 5x on 3 local rigs: tokens/sec is useless, thinking variance hits 5x — julianharris · 2026-10-04
- RAM prices nearly quadruple in two years as users chase 256k contexts — lxfater · 2026-10-04
- Running a 100B+ Qwen3 model locally on 64GB RAM: good vibes, short 192K context — lxfater · 2026-10-04
- Strata calibrate nearly tripled decode speed: 256K context on a 16GB GPU — MoonsvnLyn · 2026-10-04