Elastic IP

A regular public IP assigned to an EC2 instance is released the moment you stop or terminate it — if you restart the instance, you get a new public IP, breaking any DNS records or firewall rules pointing at the old one. An Elastic IP (EIP) solves this by giving you a static public IPv4 address that you own and control independently of any instance’s lifecycle — you can attach it, detach it, and re-attach it to a different instance at will.

• An Elastic IP is allocated to your account within a specific region — it cannot be attached to a resource in a different region.
• It can be attached to (and moved between) any instance or network interface within that same region, regardless of AZ.
• Elastic IPs are also required by, and attached to, other regional resources like NAT Gateways.

graph LR
EIP["Elastic IP (static)"] -.->|"attach/detach"| EC2a["EC2 Instance A"]
EIP -.->|"re-attach later"| EC2b["EC2 Instance B"]
classDef eip fill:#f3ecfb,stroke:#8b3de0,stroke-width:3px,color:#3a1a5c

An Elastic IP is free while it’s actively attached to a running resource. AWS charges an hourly fee for idle Elastic IPs — ones that are allocated but not attached to anything, or attached to a stopped instance — specifically to discourage hoarding scarce IPv4 addresses.

• By default you’re limited to 5 Elastic IPs per region per account (a soft limit, increasable via support request).
• Remembering to release unused Elastic IPs is a classic cost-optimization exam scenario — “why is my bill higher after I stopped all my instances?” is almost always an idle EIP (or an idle NAT Gateway).
• An EIP is not the same as a public IP auto-assigned at launch — the auto-assigned one changes on stop/start, an EIP does not.
• Moving an EIP between instances is nearly instant and is a common technique for quick failover — pointing traffic at a standby instance without changing DNS.