Free calculator
AWS Graviton savings calculator
Graviton lists about 20% below the equivalent x86 instance. The useful question is not what 20% of your compute bill is, it is what 20% of the portable share is.
x86, on-demand Linux.
Running around the clock.
Interpreted runtimes and managed services: high. Anything with native x86 dependencies: low.
On m5.xlarge
$840.96/month
$10,091.52/year · ≈ £656/month
If all moved to m7g.xlarge
$714.82/month
$8,577.84/year · ≈ £558/month
Saved at 60% ported
$75.68/month
$908.16/year · ≈ £59/month · 9% of this line
m5.xlarge is $0.19/hour against $0.16/hour for m7g.xlarge, both on-demand Linux at 730 hours a month. Compute Savings Plans cover both architectures, so an existing Compute plan follows you across. EC2 Instance Savings Plans and standard Reserved Instances are locked to a family and do not.
The scan lists every instance running an x86 type that has a like-for-like Graviton equivalent, with the saving per instance and the commitment you already hold against it.
Estimate the whole billWhat ports easily, and what does not
- →Managed services: trivial. RDS, ElastiCache, OpenSearch and MemoryDB are a modify operation. Nothing in your application knows or cares, which makes these the first things to move.
- →Lambda: one setting. Switch the architecture to arm64. For a runtime with no native dependencies it is a one-line change for about 20% off, and functions often run slightly faster, which reduces the duration charge too.
- →Containers on interpreted runtimes: a build change. Node, Python, Ruby and JVM images generally build for arm64 without incident if the base images publish multi-arch manifests. The work lands in CI rather than in the application.
- →Native dependencies: a project. C extensions, vendored binaries and anything compiled needs rebuilding and retesting. Usually possible, occasionally blocked outright by a dependency that never shipped an Arm build.
- →Licensed commercial software: check first. Some vendors do not support Arm and some licence per architecture. Find that out before the migration plan, not during it.
Frequently asked questions
How much cheaper is Graviton than x86?
Roughly 20% on list price for a like-for-like instance. A t4g.medium is $0.0336 an hour against $0.0416 for a t3.medium, and m7g against m5 is a similar gap. AWS also claims up to 40% better price-performance, which depends entirely on the workload; the 20% list gap is the part you can count on.
What is hard to migrate to Graviton?
Anything with compiled native dependencies, vendored x86 binaries, or commercial software licensed per architecture. Interpreted runtimes such as Node, Python and Ruby, JVM applications, and managed services such as RDS and ElastiCache are usually a configuration change rather than a port.
Do my Savings Plans still apply after migrating?
Compute Savings Plans cover any instance family, size, region and architecture, so they follow you to Graviton automatically. EC2 Instance Savings Plans and standard Reserved Instances are locked to a family, so migrating away from x86 while holding one means paying for a commitment you no longer use. Check which you hold before planning the migration.
Should I migrate to Graviton before or after rightsizing?
After. A 20% discount on an instance twice the size it needs to be is a worse outcome than the correct size on x86, and porting a workload you were about to delete is wasted effort. Delete the idle, size the rest, then port the survivors.
Is Graviton available for RDS and Lambda?
Yes. RDS, ElastiCache, OpenSearch and MemoryDB all offer Graviton instance classes and the change is a modify operation. Lambda supports arm64 with a single configuration setting, which is the cheapest 20% anyone will ever offer you if your runtime has no native dependencies.