GPU Cloud Pricing

Compare 30 GPU models across AWS, Azure, and GCP.

Sort by:

B300

1 instance configurations

$142.42/hr

AWS

NVIDIA

3 instance configurations

$113.93/hr

Spot: $19.23/hr

AWS

TPU v5p

564 instance configurations

$33.60/hr

Spot: $6.97/hr

GCP

B200

2 instance configurations

$32.61/hr

GCP

TPU v3

16 instance configurations

$16.32/hr

GCP

Gaudi HL-205

2 instance configurations

$13.11/hr

AWS

H200

46 instance configurations

$12.88/hr

Spot: $3.54/hr

AWSGCPAZURE

MI300X

24 instance configurations

$9.60/hr

Spot: $8.87/hr

AZURE

H100-MEGA

13 instance configurations

$9.46/hr

Spot: $1.62/hr

GCP

Qualcomm AI100

2 instance configurations

$8.92/hr

Spot: $2.66/hr

AWS

A100-80GB

26 instance configurations

$5.03/hr

Spot: $1.71/hr

GCPAWS

RTX PRO Server 6000

30 instance configurations

$3.36/hr

Spot: $1.01/hr

AWS

V100

57 instance configurations

$2.86/hr

Spot: $0.38/hr

GCPAWS

TPU v6e

88 instance configurations

$2.70/hr

Spot: $0.28/hr

GCP

L40S

80 instance configurations

$1.86/hr

Spot: $0.49/hr

AWS

P100

21 instance configurations

$1.84/hr

Spot: $0.22/hr

GCP

H100

216 instance configurations

$1.40/hr

Spot: $1.23/hr

AWSGCPAZURE

TPU v5e

128 instance configurations

$1.20/hr

Spot: $0.24/hr

GCP

A10G

136 instance configurations

$1.01/hr

Spot: $0.26/hr

AWS

P4

21 instance configurations

$0.98/hr

Spot: $0.14/hr

GCP

A100

160 instance configurations

$0.73/hr

Spot: $0.68/hr

AWSGCPAZURE

T4g

37 instance configurations

$0.42/hr

Spot: $0.08/hr

AWS

Radeon Pro V520

40 instance configurations

$0.38/hr

Spot: $0.08/hr

AWS

L4

352 instance configurations

$0.20/hr

Spot: $0.03/hr

GCPAWS

V620

29 instance configurations

$0.13/hr

Spot: $0.12/hr

AZURE

T4

296 instance configurations

$0.10/hr

Spot: $0.08/hr

AWSGCPAZURE

A10

153 instance configurations

$0.09/hr

Spot: $0.08/hr

AZURE

V710

20 instance configurations

$0.09/hr

Spot: $0.06/hr

AZURE

TPU v2

5 instance configurations

GCP

GB200

1 instance configurations

GCP

Run a free assessment to identify overprovisioned workloads, idle capacity, and your potential savings, in minutes.

Most clusters are overprovisioned.
Let's prove yours is.