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

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

Rig A

NVIDIA · accelerator

Change

Rig B

AWS · cloud-model

Change

0 of 30 fields comparable

Spec	NVIDIA GB200	AWS Trainium2
Bf16 Tflops PER Chip	—	667
Blackwell Gpus	2	—
Chips PER Trn2 Instance	—	16
Chips PER Ultraserver	—	64
CPU Memory Bandwidth GB S MAX	512	—
CPU Memory Lpddr5x GB MAX	480	—
FP8 Dense Petaflops PER Chip	—	1.3
FP8 Pflops	20	—
FP8 Sparse Petaflops PER Chip	—	5.2
Grace Cpus	1	—
HBM Bandwidth TB S	16	—
HBM Bandwidth TB S PER Chip	—	2.9
HBM Capacity GB PER Chip	—	96
HBM Type	—	HBM3
Neuroncores PER Chip	—	8
Nvfp4 Pflops Dense	20	—
Nvfp4 Pflops Sparse	40	—
Nvlink Bandwidth TB S	3.6	—
Nvlink C2C Bandwidth GB S	900	—
Total Hbm3e GB	372	—
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