AWS cost · EC2

Older EC2 generations cost more for the same specs

An instance type is chosen once, in a hurry, and then never looked at again. AWS keeps launching hardware that does the same work for less money, and it does not reprice what you are already running. So the box keeps costing what it cost on day one, while the catalogue lists the same specifications at a lower price.

An m4.xlarge in eu-central-1 is $175.20/month. An m6a.xlarge - the same 4 vCPU and 16 GiB - is $151.11. That is $24.09/month for a stop, a type change and a start. Measured against the live price list, not estimated.

Two swaps, priced today

The t3a.micro is the more interesting row: cheaper and twice the vCPUs. The price falls while the machine gets better.

The family map does not work

The obvious way to hunt for this is a mental map: m4 becomes m5, t2 becomes t3, and so on. AWS contradicts that map at both ends. Ask the API which types are current generation and it says m5.xlarge is not, while t2.micro is.

So the useful question is not whether your family is old. It is whether AWS sells the same machine cheaper today - same memory, at least the vCPUs, lower price. That needs no lineage map and cannot go out of date.

One warning if you script it: the price list has its own attribute called current generation, and it answers Yes for m4.xlarge. The EC2 catalogue is the source that tells the truth.

How to check your own instances

Take the memory figure of what you are running, then ask the catalogue what else it sells at that size:

# what is running, and on what
aws ec2 describe-instances \
  --query 'Reservations[].Instances[?State.Name==`running`].{Id:InstanceId,Type:InstanceType,Az:Placement.AvailabilityZone}' \
  --output table

# current-generation types with the same 16 GiB and at least 4 vCPUs
aws ec2 describe-instance-types \
  --filters Name=current-generation,Values=true Name=memory-info.size-in-mib,Values=16384 \
  --query 'InstanceTypes[?VCpuInfo.DefaultVCpus>=`4`].{Type:InstanceType,VCpus:VCpuInfo.DefaultVCpus,Ena:NetworkInfo.EnaSupport}' \
  --output table

Match memory exactly rather than taking anything larger - loosening it turns a shortlist into hundreds. The cheapest type with more memory than you asked for is a different machine, not a better deal.

What the swap costs

It is a stop and a start, so there is downtime, anything on instance store is lost, and an auto-assigned public IP is replaced. That is the predictable part.

The unpredictable parts are worth checking first. An older instance may report ENA as unsupported - t2.micro does - and every current generation is Nitro, which needs the ENA and NVMe drivers present before the instance will start on the new type. A t-family target defaults to unlimited credit mode and bills surplus credits. And if the old family is covered by a Reserved Instance or an EC2 Instance Savings Plan, the commitment stays behind on a family you no longer run. That can turn a saving into a loss until it expires.

# MUTATING - the instance must be stopped first
aws ec2 stop-instances --instance-ids i-0123456789abcdef0
aws ec2 modify-instance-attribute \
  --instance-id i-0123456789abcdef0 \
  --instance-type Value=m6a.xlarge
aws ec2 start-instances --instance-ids i-0123456789abcdef0

Two questions to ask in the same sitting: whether the instance should be smaller rather than newer, and whether the workload could move to Graviton, which saves more but is a port rather than a swap.

Frequently asked questions

More AWS cost breakdowns