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 32 fields comparable

Spec	NVIDIA HGX B200	AWS Trainium2
Bf16 Tflops PER Chip	—	667
Chips PER Trn2 Instance	—	16
Chips PER Ultraserver	—	64
Fp16 Bf16 Pflops	36	—
FP32 TFLOPS	600	—
FP4 Pflops Dense	72	—
FP4 Pflops Sparse	144	—
FP64 TFLOPS	296	—
FP8 Dense Petaflops PER Chip	—	1.3
FP8 FP6 Pflops	72	—
FP8 Sparse Petaflops PER Chip	—	5.2
Gpus	8	—
HBM Bandwidth TB S PER Chip	—	2.9
HBM Capacity GB PER Chip	—	96
HBM Type	—	HBM3
Int8 Pops	72	—
Networking Bandwidth TB S	0.8	—
Neuroncores PER Chip	—	8
Nvlink Generation	5	—
Nvlink GPU TO GPU TB S	1.8	—
Total HBM GB	1,400	—
Total Nvlink Bandwidth TB S	14.4	—
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