Compare GPUs

Pick up to four GPUs to compare live prices, compute per dollar, memory and providers side by side.

H100 SXM×H200 SXM×Instinct MI300X×

+ Add GPU

H100 SXM

Lowest on-demand

$1.82/GPU·h

$1.83 per PFLOP·h (FP16)

80 GB0.989 PF FP1630 providers

H200 SXM

Lowest on-demand

$2.92/GPU·h

$2.95 per PFLOP·h (FP16)

141 GB0.989 PF FP1619 providers

Instinct MI300X

Lowest on-demand

$2.39/GPU·h

$1.83 per PFLOP·h (FP16)

192 GB1.31 PF FP167 providers

Cost per PFLOP-hour by precision

At each GPU's lowest on-demand price. Lower is better; missing bars mean no native support.

FP16 / BF16

$1.83/PFLOP·h

$2.95/PFLOP·h

$1.83/PFLOP·h

FP8

$0.92/PFLOP·h

$1.47/PFLOP·h

$0.91/PFLOP·h

INT8

$0.92/PFLOP·h

$1.47/PFLOP·h

$0.91/PFLOP·h

TF32

$3.67/PFLOP·h

$5.89/PFLOP·h

$3.66/PFLOP·h

FP32

$27.10/PFLOP·h

$43.52/PFLOP·h

$14.63/PFLOP·h

FP64

$27.10/PFLOP·h

$43.52/PFLOP·h

$14.63/PFLOP·h

Lowest on-demand price over time

Daily lowest on-demand price per GPU-hour.

…

Specs and prices

	H100 SXM	H200 SXM	Instinct MI300X
Memory	80 GB HBM3	141 GB HBM3e	192 GB HBM3better
Memory bandwidth	3,350 GB/s	4,800 GB/s	5,300 GB/sbetter
FP16 / BF16 (dense)	0.989 PF	0.989 PF	1.31 PFbetter
FP8 (dense)	1.98 PF	1.98 PF	2.61 PFbetter
FP4 (dense)	—	—	—
INT8 (dense)	1.98 PF	1.98 PF	2.61 PFbetter
TF32 (dense)	0.495 PF	0.495 PF	0.654 PFbetter
FP32 (dense)	0.067 PF	0.067 PF	0.163 PFbetter
FP64 (dense)	0.067 PF	0.067 PF	0.163 PFbetter
Power	700 W	700 W	750 W
Architecture	Hopper (2022)	Hopper (2024)	CDNA 3 (2023)
Lowest on-demand	$1.82better	$2.92	$2.39
Median on-demand	$3.35	$4.47	$2.81better
Lowest spot	$0.74better	$0.85	$1.11
FP16/BF16 $ per PFLOP·h	$1.83	$2.95	$1.83better
Providers	30	19	7

Who offers them

Cheapest on-demand price per GPU-hour at each provider.

Provider	H100 SXM	H200 SXM	Instinct MI300X
Azure 🇺🇸	$12.3/GPU·h	$10.6/GPU·h	$6.00/GPU·h
DigitalOcean 🇺🇸	$4.41/GPU·h	$4.47/GPU·h	$2.59/GPU·h
Oracle Cloud 🇺🇸	$10.0/GPU·h	$10.0/GPU·h	$6.00/GPU·h
RunPod 🇺🇸	$2.69/GPU·h	$3.59/GPU·h	$2.39/GPU·h
Seeweb 🇮🇹	$2.12/GPU·h	$2.92/GPU·h	$2.64/GPU·h
AWS 🇺🇸	$6.88/GPU·h	$7.91/GPU·h	—
Civo 🇬🇧	$2.99/GPU·h	$3.49/GPU·h	—
CoreWeave 🇺🇸	$6.16/GPU·h	$6.31/GPU·h	—
EmpirioLabs AI 🇺🇸	$4.80/GPU·h	$8.40/GPU·h	—
Gcore 🇱🇺	$2.60/GPU·h	$3.50/GPU·h	—
Google Cloud 🇺🇸	$11.0/GPU·h	$10.8/GPU·h	—
GPU.ai 🇺🇸	$2.79/GPU·h	$3.95/GPU·h	—
HHorizon	$1.95/GPU·h	$3.00/GPU·h	—
Hyperbolic 🇺🇸	$3.49/GPU·h	$4.49/GPU·h	—
Hyperstack 🇬🇧	$3.20/GPU·h	$3.99/GPU·h	—
Koyeb 🇫🇷	$2.50/GPU·h	$3.00/GPU·h	—
Lium	$2.50/GPU·h	$4.15/GPU·h	—
Nebius 🇳🇱	$4.50/GPU·h	$5.40/GPU·h	—
Verda (DataCrunch) 🇫🇮	$5.09/GPU·h	$6.50/GPU·h	—
Akash Network	$2.55/GPU·h	—	—
Cudo Compute 🇬🇧	$1.82/GPU·h	—	—
Denvr 🇨🇦	$2.50/GPU·h	—	—
Hot Aisle 🇺🇸	—	—	$2.99/GPU·h
Lambda 🇺🇸	$4.59/GPU·h	—	—
Leafcloud 🇳🇱	$3.53/GPU·h	—	—
Massed Compute 🇺🇸	$3.14/GPU·h	—	—
Oblivus 🇬🇧	$4.31/GPU·h	—	—
Sesterce 🇫🇷	$3.52/GPU·h	—	—
STACKIT 🇩🇪	$9.02/GPU·h	—	—
UpCloud 🇫🇮	$2.50/GPU·h	—	—
Voltage Park 🇺🇸	$1.99/GPU·h	—	—

The link updates as you pick GPUs, so you can share this comparison.