V100$0.190/hrT4$0.343/hrA16$0.471/hrMI300X$0.500/hrL40$0.690/hrL40S$0.790/hrL4$0.917/hrA100$1.00/hrM60$1.14/hrP100$1.42/hrMI25$1.86/hrGH200 Grace Hopper$1.99/hrRadeon PRO V710$2.00/hrRadeon PRO V620$2.54/hrMI355X$2.59/hrH100$2.69/hrA10$3.20/hrH200$3.59/hr
Live · 5 official price feeds
What GPUs actually cost.
Every rate here is the provider's own published price divided by the number of GPUs in the instance — never an estimate, and the arithmetic is shown on each chip's page. 43 GPUs, 4,395 offers, 5 providers.
Data-centre GPUs
22 priced · cheapest firstfrom
$0.190
per GPU / hour
16 GBData centre2 exact variants$1.52 per PFLOP-hr
RunPod
Tesla V100
16 GB
$0.190LOW STOCK
RunPod
V100 SXM2
16 GB
$0.230LOW STOCK
Azure
V100
16 GB
$3.06
All 3 rates and the arithmetic →
2 providers
NVIDIA T4
from
$0.343
per GPU / hour
16 GBData centre$5.28 per PFLOP-hr
Azure
T4
16 GB
$0.343
The rate and the arithmetic →
1 provider
NVIDIA A16
from
$0.471
per GPU / hour
16 GBData centre
Vultr
A16
16 GB
$0.471
The rate and the arithmetic →
1 provider
AMD MI300X
from
$0.500
per GPU / hour
192 GBData centre
RunPod
MI300X
192 GB
$0.500LOW STOCK
Vultr
Instinct MI300X
192 GB
$3.99
Azure
Instinct MI300X
192 GB
$6.00
All 3 rates and the arithmetic →
3 providers
NVIDIA L40
from
$0.690
per GPU / hour
48 GBData centre$3.81 per PFLOP-hr
RunPod
L40
48 GB
$0.690LOW STOCK
The rate and the arithmetic →
1 provider
NVIDIA L40S
from
$0.790
per GPU / hour
48 GBData centre2 exact variants$2.18 per PFLOP-hr
RunPod
L40S
48 GB
$0.790LOW STOCK
Vultr
L40S
48 GB
$1.67
Scaleway
L40S
48 GB
$1.71
Scaleway
L40S
memory not stated
$2.39
All 4 rates and the arithmetic →
3 providers
NVIDIA L4
from
$0.917
per GPU / hour
24 GBData centre$7.58 per PFLOP-hr
Scaleway
L4
24 GB
$0.917
The rate and the arithmetic →
1 provider
NVIDIA A100
from
$1.00
per GPU / hour
40–80 GBData centre5 exact variants$3.21 per PFLOP-hr
RunPod
A100 SXM 40GB
40 GB
$1.00LOW STOCK
RunPod
A100 SXM
80 GB
$1.39LOW STOCK
Vultr
A100
80 GB
$2.40
Vultr
A100
80 GB
$2.80
All 6 rates and the arithmetic →
3 providers
Read it another way
$
Rate index
Every GPU on one logarithmic scale, so you can see the market's whole price structure at a glance instead of one chip at a time.
Open the index → ◲By provider
What each cloud rents, what it charges, and how each of its prices compares with the cheapest published elsewhere.
5 providers → ≡Like for like
The 5 GPUs rented by more than one provider in the same exact variant — the only honest comparisons in the index.
Compare → ✓Method & coverage
7,714 rows read, 4,395 kept, and every refusal counted by reason. Nothing is dropped silently.
How this works →