Cloud Cost Calculator: AWS vs Azure vs GCP vs Oracle
Compare a workload's cost across AWS, Azure, Google Cloud, and Oracle.
Normalized to each provider's nearest region.
Reserved/Spot apply an approximate discount.
730 = always-on.
Microsoft Azure is the cheapest that prices every requested component, at $351.80/mo, about 4% below the next complete quote. Oracle Cloud shows a lower $151.32/mo but could not price load balancer.
AWS
Microsoft Azure
Lowest costGoogle Cloud
Oracle Cloud
Monthly cost by provider
| Component | AWS | Microsoft Azure | Google Cloud | Oracle Cloud |
|---|---|---|---|---|
| Compute | $294.34 | $280.32 | $319.74 | $134.32 |
| Database | $0.00 | $0.00 | $0.00 | $0.00 |
| Block storage | $8.00 | $8.03 | $11.00 | $4.25 |
| Object storage | $11.50 | $10.40 | $10.00 | $12.75 |
| Egress | $36.00 | $34.80 | $59.88 | $0.00 |
| Load balancer | $16.43 | $18.25 | $18.25 | — |
| Monthly total | $366.27 | $351.80 | $418.87 | $151.32 |
Prices last updated 2026-06. Estimates use USD public on-demand list prices and exclude taxes, free-tier credits, support plans, and negotiated discounts. NoSQL databases are not modelled (their per-request pricing is not comparable to vCPU/RAM). Always confirm with each provider's own calculator before committing.
What this tool does
Cloud pricing is hard to compare because every provider uses different instance families, units, and names for the same thing. This calculator lets you describe a workload once, in plain terms (how many instances, how much vCPU and RAM, how much storage, how much outbound traffic), and then estimates the monthly and annual cost of the nearest equivalent on AWS, Azure, Google Cloud, and Oracle Cloud at the same time. The breakdown shows exactly where the differences come from.
How the comparison is built
Each abstract input maps to the closest real service on every provider: compute to EC2 / Virtual Machines / Compute Engine / OCI Compute, block storage to EBS / Managed Disks / Persistent Disk / Block Volume, object storage to S3 / Blob / Cloud Storage / Object Storage, and so on.
For compute, the hardest part is matching an instance. AWS, Azure, and Google Cloud
sell fixed shapes, so the tool selects the smallest general-purpose shape whose vCPU
and RAM both meet or exceed your request and reports the shape it used (for example
m7i.large or n2-standard-2). Oracle Cloud sells flexible shapes priced per OCPU
and per GB of memory, so its compute cost is computed directly from your request.
Pricing models
On-Demand uses the exact published list price. Reserved (1-year and 3-year) and Spot apply each provider’s published committed-use or spot discount to the on-demand price. Those discount figures are approximate and clearly flagged: real reserved pricing depends on the exact commitment and SKU, and spot prices change continuously.
What is included and what is not
Included: compute, block storage, object storage (Standard, Infrequent Access, and Archive tiers), internet egress with each provider’s free allowance, a base load balancer, and an optional managed relational (PostgreSQL) database.
Not included: taxes, free-tier credits, support plans, data-processing charges on load balancers, inter-region or inter-zone transfer, and any negotiated or enterprise discount. NoSQL databases are not modelled because their per-request and throughput pricing does not map onto a vCPU and RAM workload.
A worked example
Suppose you run two always-on general-purpose instances (4 vCPU, 16 GB each), 100 GB of block storage, 500 GB of Standard object storage, 500 GB of monthly egress, and a load balancer, in US East. The tool sizes each provider’s nearest 4 vCPU / 16 GB shape, multiplies the hourly rate by 730 hours and two instances, adds the storage and egress (after each provider’s free allowance) and the load balancer, and ranks the four totals so you can see which provider is cheapest for that exact shape and by how much.
Methodology
Every price in the dataset is a USD public on-demand list price taken from an official provider source, stored with that source and an “as of” date. A price with no source is never shipped. Because list prices change, the tool shows when the data was last updated and links you back to each provider’s own calculator for a binding quote.
Frequently asked questions
Which cloud providers does it compare?
Amazon Web Services (AWS), Microsoft Azure, Google Cloud (GCP), and Oracle Cloud Infrastructure (OCI). You describe a workload once and see the estimated monthly and annual cost on all four side by side.
Are these the exact prices I will pay?
No. They are public on-demand list prices in USD, gathered from each provider's pricing pages and dated in the tool. Your real bill depends on taxes, free-tier credits, support plans, sustained-use and committed-use discounts, and any enterprise agreement. Treat the result as a like-for-like starting point, then confirm with each provider's own calculator.
How does it match an instance across providers?
Instance shapes do not line up one to one. For AWS, Azure, and Google Cloud the tool picks the smallest general-purpose shape whose vCPU and RAM both meet or exceed your request, and shows which shape it chose. Oracle Cloud uses flexible shapes, so it is priced per OCPU and per GB of memory (1 OCPU equals 2 x86 vCPUs).
What do Reserved and Spot do?
On-Demand uses the exact published list price. Reserved (1 or 3 year) and Spot apply each provider's published, approximate committed-use or spot discount to that list price. Spot prices in particular move in real time, so treat those figures as indicative only.
Why is a provider or component sometimes blank?
The tool never invents a number. If a price for a given region, service, or tier is not in the sourced dataset, that cell is left blank and excluded from the total rather than guessed. NoSQL databases are not modelled at all, because their per-request pricing is not comparable to a vCPU and RAM workload.