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

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

Rig A

AWS · cloud-model

Change

Rig B

NVIDIA · accelerator

Change

0 of 35 fields comparable

Spec	AWS Trainium2	NVIDIA H100
BF16 TFLOPS (sparse)	—	1,979
Bf16 Tflops PER Chip	667	—
Chips PER Trn2 Instance	16	—
Chips PER Ultraserver	64	—
Form factor	—	SXM
FP16 TFLOPS (dense)	—	989.5
FP16 TFLOPS (sparse)	—	1,979
FP32 TFLOPS	—	67
FP64 Tensor TFLOPS	—	67
FP64 TFLOPS	—	34
FP8 Dense Petaflops PER Chip	1.3	—
FP8 Sparse Petaflops PER Chip	5.2	—
FP8 TFLOPS (dense)	—	1,979
FP8 TFLOPS (sparse)	—	3,958
HBM (GB)	—	80
HBM Bandwidth TB S PER Chip	2.9	—
HBM Capacity GB PER Chip	96	—
HBM Type	HBM3	—
INT8 TOPS (sparse)	—	3,958
Memory bandwidth (TB/s)	—	3.35
Memory type	—	HBM3
Neuroncores PER Chip	8	—
NVLink bandwidth (GB/s)	—	900
TDP max (W)↓ better	—	700
TF32 TFLOPS (sparse)	—	989
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	—