Skip to content

AWS · Azure · GCP · 60 regions · prices with dates and sources

Which cloud region is cheapest for this workload?

Pick an instance shape. Every region is ranked by that instance's on-demand list price, with the exact SKU named beside each number.

The default instance families, Intel or AMD.

One provider only: AWS regions · Azure regions · GCP regions

60 regions · sorted by price

Prices: AWS live 10 Oct 2026 · Azure live 10 Oct 2026 · GCP verified 10 Oct 2026

Cheapest regionfor general purpose onm5.large

AWS

US East (N. Virginia)

Ashburn, United States·us-east-1·6 availability zones

$0.0960
per hour, on-demand Linux list price
$70
per month at 730 h
  • $46/mo less(40%) than Brazil South, the dearest
  • 95 region heuristic

Based in North America, 6 AZs, SOC2 and PCI-DSS coverage.

Cheapest on each provider

  1. AWSUS East (N. Virginia)$0.0960$70/mo
  2. AzureEast US$0.0960$70/mo
  3. GCPUS Central (Iowa)$0.0971$71/mo

Every region on one price scale

The dearest region costs 66% more than the cheapest. Select a dot to find it in the table.

$0.100$0.120$0.140AWSfrom $0.0960Azurefrom $0.0960GCPfrom $0.0971cheapest $0.0960dearest $0.1590

Instances: AWS m5.large, Azure D2s_v3, GCP n2-standard-2. On-demand Linux list price per hour.

60 regions ranked

RegionPin
1
AWSUS East (N. Virginia)
us-east-1|m5.large
$0.0960$70
2
AWSUS East (Ohio)
us-east-2|m5.large
$0.0960$70
3
AWSUS West (Oregon)
us-west-2|m5.large
$0.0960$70
4
AzureEast US
eastus|D2s_v3
$0.0960$70
5
AzureEast US 2
eastus2|D2s_v3
$0.0960$70
6
AzureWest US 2
westus2|D2s_v3
$0.0960$70
7
AzureWest US 3
westus3|D2s_v3
$0.0960$70
8
GCPUS Central (Iowa)
us-central1|n2-standard-2
$0.0971$71
9
GCPUS East (South Carolina)
us-east1|n2-standard-2
$0.0971$71
10
GCPUS West (Oregon)
us-west1|n2-standard-2
$0.0971$71
11
AWSAsia Pacific (Mumbai)
ap-south-1|m5.large
$0.1010$74
12
AWSEurope (Stockholm)
eu-north-1|m5.large
$0.1020$74
13
AzureSweden Central
swedencentral|D2s_v3
$0.1020$74
14
AzureCentral India
centralindia|D2s_v3
$0.1050$77
15
GCPTel Aviv
me-west1|n2-standard-2
$0.1068$78
16
GCPBelgium
europe-west1|n2-standard-2
$0.1068$78
17
GCPMontréal
northamerica-northeast1|n2-standard-2
$0.1069$78
18
GCPToronto
northamerica-northeast2|n2-standard-2
$0.1069$78
19
GCPNetherlands
europe-west4|n2-standard-2
$0.1069$78
20
GCPFinland
europe-north1|n2-standard-2
$0.1069$78
21
AWSCanada (Central)
ca-central-1|m5.large
$0.1070$78
22
AWSEurope (Ireland)
eu-west-1|m5.large
$0.1070$78
23
AzureNorth Europe
northeurope|D2s_v3
$0.1070$78
24
GCPUS East (N. Virginia)
us-east4|n2-standard-2
$0.1094$80
25
AWSEurope (London)
eu-west-2|m5.large
$0.1110$81
26
AzureCanada Central
canadacentral|D2s_v3
$0.1110$81
27
AWSUS West (N. California)
us-west-1|m5.large
$0.1120$82
28
AWSEurope (Paris)
eu-west-3|m5.large
$0.1120$82
29
AzureFrance Central
francecentral|D2s_v3
$0.1120$82
30
GCPTaiwan
asia-east1|n2-standard-2
$0.1125$82
31
AWSEurope (Frankfurt)
eu-central-1|m5.large
$0.1150$84
32
AzureGermany West Central
germanywestcentral|D2s_v3
$0.1150$84
33
AzureUK South
uksouth|D2s_v3
$0.1160$85
34
GCPMumbai
asia-south1|n2-standard-2
$0.1166$85
35
GCPUS West (Los Angeles)
us-west2|n2-standard-2
$0.1167$85
36
AWSAsia Pacific (Seoul)
ap-northeast-2|m5.large
$0.1180$86
37
AWSMiddle East (Bahrain)
me-south-1|m5.large
$0.1180$86
38
GCPSingapore
asia-southeast1|n2-standard-2
$0.1198$87
39
AWSAsia Pacific (Singapore)
ap-southeast-1|m5.large
$0.1200$88
40
AWSAsia Pacific (Sydney)
ap-southeast-2|m5.large
$0.1200$88
41
AzureWest Europe
westeurope|D2s_v3
$0.1200$88
42
AzureUAE North
uaenorth|D2s_v3
$0.1200$88
43
AWSAsia Pacific (Tokyo)
ap-northeast-1|m5.large
$0.1240$91
44
GCPTokyo
asia-northeast1|n2-standard-2
$0.1246$91
45
AzureSoutheast Asia
southeastasia|D2s_v3
$0.1250$91
46
AzureAustralia East
australiaeast|D2s_v3
$0.1250$91
47
GCPLondon
europe-west2|n2-standard-2
$0.1251$91
48
GCPFrankfurt
europe-west3|n2-standard-2
$0.1251$91
49
AWSAfrica (Cape Town)
af-south-1|m5.large
$0.1270$93
50
AzureSouth Africa North
southafricanorth|D2s_v3
$0.1270$93
51
GCPZürich
europe-west6|n2-standard-2
$0.1272$93
52
AzureJapan East
japaneast|D2s_v3
$0.1290$94
53
GCPJakarta
asia-southeast2|n2-standard-2
$0.1306$95
54
AWSAsia Pacific (Hong Kong)
ap-east-1|m5.large
$0.1320$96
55
AzureEast Asia
eastasia|D2s_v3
$0.1320$96
56
GCPHong Kong
asia-east2|n2-standard-2
$0.1359$99
57
GCPSydney
australia-southeast1|n2-standard-2
$0.1378$101
58
AWSSouth America (São Paulo)
sa-east-1|m5.large
$0.1530$112
59
GCPSão Paulo
southamerica-east1|n2-standard-2
$0.1542$113
60
AzureBrazil South
brazilsouth|D2s_v3
$0.1590$116

What is being compared

One instance per provider, matched by shape

Each workload maps to one named instance on each provider, chosen for the same vCPU and RAM. The ranking is that instance's on-demand Linux list price in every region, so a row in the table is always the SKU shown beside it.

The shapes match except for compute optimised on GCP: the C2 family has no 8 GB size, so c2-standard-4 carries 16 GB and is not like for like.

Full methodology, sources and dates

WorkloadAWSAzureGCP
General purpose
2 vCPU · 8 GB
m5.large
2 vCPU · 8 GB
D2s_v3
2 vCPU · 8 GB
n2-standard-2
2 vCPU · 8 GB
Compute optimised
4 vCPU · 8 GB (GCP 16 GB)
c5.xlarge
4 vCPU · 8 GB
F4s_v2
4 vCPU · 8 GB
c2-standard-4
4 vCPU · 16 GB (no 8 GB size)
Memory optimised
2 vCPU · 16 GB
r5.large
2 vCPU · 16 GB
E2s_v3
2 vCPU · 16 GB
n2-highmem-2
2 vCPU · 16 GB
General purpose · Arm
Arm-based instances (AWS Graviton, Azure Cobalt, Google Axion). Usually cheaper per vCPU; not every region offers them.
m7g.large
2 vCPU · 8 GB
D2ps_v6
2 vCPU · 8 GB
c4a-standard-2
2 vCPU · 8 GB
General purpose · Burstable
Cheaper instances that earn CPU credits while idle and spend them under load (AWS t3, Azure B-series). Sustained full CPU costs more than the list price. Google has no burstable type at this size.
t3.large
2 vCPU · 8 GB
B2s_v2
2 vCPU · 8 GB
no reference instance
Storage
general + 100 GB SSD
m5.large
+ 100 GB gp3
D2s_v3
+ 100 GB Standard SSD
n2-standard-2
+ 100 GB pd-balanced
Custom
your vCPU and RAM
Priced from each region's per-vCPU and per-GB rates, fitted by least squares to the provider's general, compute and memory families. An estimate rather than a list price, and labelled as fitted wherever it appears.

What the monthly number includes

In the number

  • The instance's on-demand Linux list price, for 730 hours a month.
  • Internet egress at the region's first paid tier, once you enter a monthly volume.
  • 100 GB of general-purpose SSD for the Storage workload.
  • Your assumed service discount when you set the slider; internet egress stays at list price.

Not in the number

  • Reserved, savings-plan or spot rates. Everything is on-demand until you apply a discount.
  • Load balancers, NAT gateways, managed databases, support plans and taxes.
  • Windows or other OS licences, and inter-region transfer (see the placement tool for that).
  • Negotiated enterprise pricing. Confirm with the provider before you commit.

Questions people ask

Are these prices accurate?
AWS and Azure compute prices are fetched from the providers' price APIs by a script in the repository and carry the date they were fetched; GCP prices are read from Google's pricing pages and carry that date. Every number on the site shows its provenance. Prices are on-demand, Linux list rates with no commitment. Your invoice will differ with discounts, support plans and traffic, so confirm with the provider before you commit.
What is included in the cost comparison?
Compute (vCPU and RAM) for a reference instance per workload type, plus 100 GB of general-purpose SSD for the Storage type, plus internet egress at the first paid tier when you enter a monthly volume. Managed services, load balancers, NAT, support and licences are not included. Instances are matched by capacity across providers, not by identical SKUs.
Does this include reserved or spot pricing?
No. Every price is on-demand. Commitment discounts depend on the service, term, utilisation and billing agreement. Spot capacity has variable prices and interruption risk. Compare on-demand regions here, then model your eligible service discount with the slider or your own provider quote.
What is the difference between compute and memory optimised?
Compute-optimised workloads such as batch jobs, HPC and game servers need more vCPU per GB of RAM; memory-optimised workloads such as in-memory databases and caches need more RAM per vCPU. Each type maps to a different reference instance, and a region that is cheapest for one is not always cheapest for the other.
How often is pricing updated?
A scheduled job refreshes the data daily and opens a change list for review; publication waits for review and CI, and the deployment gate refuses price observations older than 45 days. The date and source of each provider's prices are shown next to the results and on the methodology page.
What does "unknown" mean in the results?
That a region does not price that instance, or a rate was not published by the source. Unknown values are never treated as zero: the region is left out of that comparison and the note under the results says why.