NVIDIA Google TPU

TPU v6e (8 cores)

Eight-core v6e pods maximize efficiency gains for large-scale inference with minimal power use.

Launch year
2025
Memory
512 GB HBM3
Memory bandwidth
>1.0 TB/s
Peak FP16 / FP32
960 TFLOPS · 0 TFLOPS

Market snapshot

$6.56 /hr

Range $6.56 – $25.92

Catalog coverage

14 live offerings

Across 1 providers · 8 regions

  • Efficiency-first TPU design
  • Scalable pod fabric
  • Cloud-native automation
TPU v6e (8 cores) render

Last refreshed Oct 20, 2025, 2:01 AM

Performance snapshot

Normalized versus NVIDIA A100 (=1.0). Values use public reference benchmarks for training and inference workloads.

  • Projected TPU Throughput×1.00
  • Projected Efficiency×0.95
  • Latency×0.55

Provider availability

Price bands per provider pulled from the live catalog.

  • Google Cloud logoGoogle Cloud14 offers
    $6.56 – $25.92 /hr

Popular regions

  • europe-west4-a2 offers
  • us-east5-a2 offers
  • us-east5-c2 offers
  • us-east5-b2 offers
  • us-east1-d2 offers
  • asia-northeast1-b2 offers
  • us-south1-c1 offers
  • us-south1-a1 offers

Weighted average price $18.89 /hr · median $21.60