AWS cost · Graviton / ARM
Graviton (ARM) is ~20% cheaper than x86 for the same work
There is a 20% discount sitting on most AWS accounts that almost nobody claims, and it is not a commitment or a contract - it is a different chip. AWS builds its own ARM processor, Graviton, and running the same workload on it costs about 20% less than x86 at equal or better performance. For Lambda it is a single dropdown. The reason it goes unclaimed is inertia: x86 was the default when everything was set up, and “it works” is a hard thing to revisit. But the work to switch is usually far smaller than the saving.
AWS Graviton (ARM) instances cost about 20% less than comparable x86 at equal or better performance, and arm64 Lambda is 20% cheaper for a single dropdown change. Interpreted code mostly runs on ARM unchanged; the only real work is recompiling native binaries or building multi-arch images. Start with Lambda, end with EC2.
Where does the Graviton discount apply?
Across the services you already run, ordered from least effort to most:
| Move | Effort | Note |
|---|---|---|
| Lambda (x86 → arm64) | One dropdown | no code change for most runtimes |
| RDS (MySQL/PostgreSQL/MariaDB) | Modify instance class | move to a db.*g class |
| Containers (ECS/EKS) | Build a multi-arch image | then schedule on Graviton nodes |
| EC2 (x86 → *g family) | Recompile native binaries | most interpreted code just runs |
Why is Graviton cheaper and not slower?
Because AWS designs the chip itself, it is not paying a third-party margin, and it passes part of that on as a lower rate. The performance is the point that surprises people: for typical workloads Graviton matches or beats the x86 equivalent, so the 20% is not a trade against speed - it is the same or better work for less. That is what makes it different from most cost levers, which ask you to give something up.
How do I find what to migrate?
Start by listing what is still on x86. Every running instance with an x86_64 architecture is a candidate - the biggest and longest-running ones are where the 20% adds up fastest. This lists them so you can pick the obvious wins first:
# running instances still on x86 - your Graviton candidates
aws ec2 describe-instances \
--filters Name=architecture,Values=x86_64 Name=instance-state-name,Values=running \
--query 'Reservations[].Instances[].{ID:InstanceId,Type:InstanceType,Arch:Architecture}' \
--output tableGraviton discounts the rate; it stacks with a Savings Plan on steady usage and with right-sizing, since a Graviton Fargate task is cheaper again.
Want to know how much you would save moving to Graviton? Connect your account read-only and see what it is wasting, in real dollars.
The role can only read - Get, Describe, List, nothing else. Read the exact permissions before you deploy it, and delete the stack whenever you want.
Frequently asked questions
How much cheaper is AWS Graviton than x86?+
Graviton (ARM) instances cost about 20% less than comparable x86 instances at equal or better performance, and AWS positions them at up to 40% better price-performance for suitable workloads. arm64 Lambda functions are 20% cheaper than x86 for the same memory and duration.
What is AWS Graviton?+
Graviton is AWS's own ARM-based processor. Instance families with a 'g' in the name - t4g, m7g, c7g, r7g - run on Graviton, as do arm64 Lambda functions and Graviton RDS classes. It is the same managed services you already use, on cheaper silicon.
Is it hard to migrate to Graviton?+
It depends on the workload. arm64 Lambda is a single dropdown for most runtimes. Interpreted code - Python, Node.js, Java, .NET - generally runs on ARM without changes. The only real work is recompiling native binaries or building multi-architecture container images, which is straightforward for most applications.
Which workloads should move to Graviton first?+
Start with Lambda - it is a zero-effort 20% cut. Then containerized workloads once you build multi-arch images, then RDS by changing the instance class, and finally EC2 where you may need to recompile. Lambda first, EC2 last, is the usual order of least effort to most.
Will my application run slower on Graviton?+
For most workloads, no - the 20% price cut comes with equal or better performance, not a slower chip. Newer generations widen the gap further. The only workloads to test carefully are ones with hand-tuned x86-specific code or dependencies that lack ARM builds.
Can I use Graviton with a Savings Plan?+
Yes. Graviton instances are covered by Compute and EC2 Instance Savings Plans just like x86. The savings stack: the ~20% Graviton discount on the rate, then the Savings Plan discount on top for steady usage.
Related cost breakdowns
Why EKS is expensive: the $73/month control plane
Every EKS cluster bills $0.10/hour - about $73/month - for the control plane before a single node runs, and $0.60/hour once it falls out of standard support. Idle and forgotten clusters are pure waste.
Why your stopped EC2 instance is still being charged
A stopped EC2 instance keeps billing for its attached EBS disk and any Elastic IP. One 8 GB volume quietly cost me ~$70 over six years. Here is what still charges when an instance is off.
Spot instances: what is safe to run on them, and what is not
Spot instances are up to 90% cheaper than on-demand - I moved almost everything to spot. The catch is a 2-minute interruption notice. Here is what is safe to run on spot and what is not.
How to find idle and oversized EC2 instances draining your bill
An EC2 instance sized for a peak that never comes idles at under 10% CPU while billing full price. Here is how to spot the oversized ones and right-size them down.