Skip to content
MCS · Rigs · Compare

Side-by-side, with the numbers that matter.

Two rigs, one table. Higher numbers win in cobalt; TDP and price flip.

Rig A

Google Cloud TPU v5p

Google · cloud-model

Change

Rig B

AWS Trainium2

AWS · cloud-model

Change

1 of 30 fields comparable

⇄ swap sides
Google Cloud TPU v5pAWS Trainium2
Bf16 Tflops PER Chip
459
667
SpecGoogle Cloud TPU v5pAWS Trainium2
Bf16 Tflops PER Chip459667
Chips PER POD8,960—
Chips PER Trn2 Instance—16
Chips PER Ultraserver—64
FP8 Dense Petaflops PER Chip—1.3
FP8 Sparse Petaflops PER Chip—5.2
FP8 Tflops PER Chip459—
HBM Bandwidth GB S2,765—
HBM Bandwidth TB S PER Chip—2.9
HBM Capacity GB PER Chip—96
HBM Capacity GIB95—
HBM Type—HBM3
ICI Bandwidth GB S1,200—
Neuroncores PER Chip—8
Sparse Cores PER Chip4—
Statusavailable (earlier generation)—
Successor Tpu7x Chips PER POD9,216—
Successor Tpu7x GA2026-03-31—
Tensor Cores PER Chip2—
Topology3D torus—
Trainium3 GA—2025-12-02
Trainium3 HBM Bandwidth TB S—4.9
Trainium3 Hbm3e GB PER Chip—144
Trn2 Instance FP8 Petaflops—20.8
Trn2 Instance HBM TB—1.5
Trn3 Ultraserver FP8 Pflops—362
Trn3 Ultraserver MAX Chips—144
Ultraserver FP8 Petaflops—83.2
Ultraserver Total HBM Bandwidth TB S—185
Ultraserver Total HBM TB—6