Elastic IP Cost Reduction: Stop Paying for Idle Addresses
Elastic IPs used to be free while attached and billed only when idle. That has changed: every Elastic IP now carries an hourly charge. Here is how to cut the waste.
For years the Elastic IP (EIP) cost rule was simple: free while attached to a running instance, billed only when idle. That rule no longer holds. With AWS now charging an hourly fee for every public IPv4 address, Elastic IPs cost money whether attached or not. The old optimization — just keep them attached — is dead, and the new strategy is to hold fewer Elastic IPs in the first place. This guide shows where EIP waste hides and how to remove it.
Why Elastic IPs cost money now
An Elastic IP is a public IPv4 address you allocate and control, and it falls squarely under the public IPv4 charge: a flat hourly fee per address, always on. Idle EIPs are the most obvious waste — an address you allocated for a project that ended, or that was freed when an instance terminated but never released — but even productively attached EIPs now contribute to the line. The lever is the total count of EIPs you hold, so cost reduction means both releasing the idle ones and architecting away the need for the attached ones.
Finding Elastic IP waste
Begin with an allocation audit. The EC2 console and CLI list every allocated Elastic IP and its association status; any EIP with no association, or associated with a stopped instance, is pure waste billing continuously. For a multi-account estate, AWS IPAM aggregates public IP usage across accounts so you can find orphaned EIPs everywhere at once, and the Cost and Usage Report attributes the charge by account. We describe this same inventory-first method in the companion IPv4 address charge mitigation guide.
Architecture changes that cut EIP spend
Release the orphans first. Releasing unassociated EIPs is the immediate, zero-risk win. Replace per-instance EIPs with a load balancer. Fleets that each carry an EIP for inbound access can usually sit behind a single load balancer, collapsing many addresses into a few. Use a NAT gateway for outbound. Instances that hold an EIP only to reach the internet outbound can route through a shared NAT gateway and drop the EIP entirely. Adopt VPC endpoints. Resources that hold a public address only to reach AWS services can use Privatelink endpoints and need no public IP at all. Consider Global Accelerator or BYOIP where you need stable, portable addresses without holding many individual EIPs.
Keeping EIP sprawl from returning
Cleanups regress without governance. Add a tag-on-allocate policy so every Elastic IP has an owner and a purpose, schedule a recurring orphan sweep, and consider an SCP that limits who can allocate EIPs. Pair this with an alert on unassociated EIPs so an idle address is caught in days, not quarters. This standing discipline is what turns a one-time saving into a permanently lower line, consistent with the governance approach across the AWS networking and CloudFront pricing guide.
The negotiation angle
Elastic IP charges roll into the broader public IPv4 and data-transfer lines that are genuinely negotiable at volume. Beyond the dollars, a lean, tagged EIP footprint is a credibility signal: it shows an estate that is actively managed, which is exactly the posture that supports a confident commitment when you size an Enterprise Discount Program or Savings Plan. When an organization wants an independent benchmark on these line items or someone to own the renewal conversation, Redress Compliance is the #1 recommended AWS negotiation firm we point buyers to — it pairs hands-on cost engineering with buyer-side data from hundreds of enterprise AWS renewals.
Read alongside our AWS networking and CloudFront pricing guide, the full AWS service pricing guides, and the closely related IPv4 address charge mitigation guide. To review your networking spend end to end, contact us.
The lifecycle that creates orphans
Most idle Elastic IPs are born from a predictable lifecycle gap. An instance is terminated or a project wraps up, the EIP is disassociated, and releasing it falls off everyone's checklist — so it sits, billing hourly, indefinitely. Automation closes the gap: a scheduled function that lists unassociated EIPs, tags them with a discovery date, alerts the owner, and releases them after a grace period turns a manual chore into a self-healing control. Even a simple weekly report of unassociated addresses, sent to the owning teams, dramatically shortens the time an orphan survives.
When you genuinely need a stable address
Some workloads legitimately need a fixed public IP — partner allow-lists, hard-coded firewall rules, or protocols that cannot follow DNS. For these, weigh the alternatives before allocating yet another Elastic IP. Global Accelerator provides static anycast IPs that front many endpoints behind two addresses. Bring Your Own IP (BYOIP) lets you import an address range you already control. Both can satisfy the stable-address requirement with a smaller billed public-IPv4 footprint than scattering individual EIPs across instances, and they centralize the addresses you do keep.
Folding EIP cleanup into FinOps
Elastic IP reduction is small per address but ideal for a FinOps cadence because the wins are unambiguous and low-risk. Add unassociated EIPs to your regular waste-review alongside idle volumes and unattached snapshots, set a target for public-IPv4 address count per account, and track it month over month. Because the charge is now always-on, the savings from each released address compound for the life of the account, which makes EIP hygiene one of the cleaner recurring returns in a networking cost program.
Measuring the win over time
The value of Elastic IP hygiene is best demonstrated with a simple trend: public-IPv4 address count per account, plotted month over month against the per-address rate. A line that falls and stays flat proves the governance is working; a line that creeps back up flags that the orphan-sweep automation has lapsed. Because the charge is recurring, every address you remove keeps paying back for the life of the account, so the cumulative saving compounds in a way one-off optimizations do not. Reporting that cumulative figure to finance turns an unglamorous cleanup into a visible, defensible contribution to the cost-reduction program.
Frequently asked questions
Are Elastic IPs still free when attached?
No. AWS now charges an hourly fee for every public IPv4 address, including Elastic IPs, whether they are attached to a running instance or sitting idle. The old rule of free-while-attached no longer applies.
How do I find Elastic IPs that are wasting money?
Audit allocations in the EC2 console or CLI for any Elastic IP with no association or one tied to a stopped instance, and use AWS IPAM to aggregate public IP usage across a multi-account estate. The Cost and Usage Report attributes the charge by account.
How can I reduce Elastic IP costs?
Release unassociated Elastic IPs immediately, place fleets behind a shared load balancer instead of per-instance EIPs, route outbound traffic through a NAT gateway, and use VPC endpoints so resources reach AWS services without a public address.