What is an IP address?

The Internet Protocol (IP) is the set of rules that moves data between networks. It chops everything you send into packets, and every packet carries two addresses in its header: the address it came from and the address it is going to (RFC 791). Routers along the way read only the destination and forward the packet one hop closer; the reply travels back the same way to the source address. That is the whole job of an IP address: to be the where in "send this where".

Two properties follow from that job. First, an address must be unique on the network it is used on; two devices with the same address on the same network cannot both receive their packets. Second, addresses are handed out in blocks that follow the shape of the network, so routers can keep short tables: a router in Frankfurt does not store a route to every device in Tokyo, it stores one route to the block that contains them.

Addresses are not permanent property of a device. Most are leased for a while by a DHCP server (your router at home, your ISP for the router itself) and can change; a laptop gets one address on your home Wi-Fi and a different one at a café. The address identifies a point of attachment, not the hardware itself.

An IPv4 address is written as four decimal numbers separated by dots, each from 0 to 255, because the address is 32 bits split into four 8-bit groups: 203.0.113.42, 192.168.1.20, 8.8.8.8. An IPv6 address is 128 bits written as eight groups of four hexadecimal digits separated by colons, with leading zeros dropped and one run of zero groups shortened to ::: 2001:db8:85a3::8a2e:370:7334, or simply 2001:db8::1. The examples in this guide use the ranges reserved for documentation (203.0.113.0/24, 2001:db8::/32), which never appear on the real internet.

At the top sits IANA (the Internet Assigned Numbers Authority), which hands large blocks to five Regional Internet Registries: ARIN for North America, RIPE NCC for Europe and the Middle East, APNIC for Asia-Pacific, LACNIC for Latin America and AFRINIC for Africa. The registries allocate smaller blocks to internet providers, hosting companies and large organizations, which is why a lookup of any address can name the network that holds it. Your provider assigns one address from its block to your connection; your router then invents private addresses for the devices behind it. Nobody "owns" a public address outright: it is registered to a network, and the registration is public information.

IPv4, defined in 1981, has 32-bit addresses: 2³², or about 4.3 billion, possible values. That seemed enormous for a research network and turned out to be far too few for a planet of phones, laptops, cars and light bulbs. The central pool of unassigned IPv4 blocks ran dry in February 2011, and the regional registries followed over the next decade. Addresses did not stop working; they became scarce, traded second-hand, and shared through NAT (below).

IPv6 (RFC 8200) was designed as the long-term fix: 128-bit addresses, roughly 3.4 × 10³⁸ of them, enough that every device can have a public address of its own and every home network gets a block of 2⁶⁴ addresses or more. The two protocols do not talk to each other directly, so the internet runs both side by side ("dual stack"); your browser picks IPv6 when a site offers it. In March 2026 Google's own measurement of visitors reaching it over IPv6 crossed 50 % for the first time. Your connection may already use both: the IPv6 test shows which, and IPv4 vs IPv6 goes into the differences.

Early IPv4 was split into classes by the first bits of the address: class A networks (first number 0–127) held 16 million addresses each, class B (128–191) 65,536 and class C (192–223) 256, with D for multicast and E reserved. The scheme wasted address space badly, so in 1993 it was replaced by Classless Inter-Domain Routing (CIDR), where a network can be any size that is a power of two and is written with a prefix length such as /24. Classes survive only as vocabulary: people still say "a class C" for a /24. The subnet calculator shows the class, the prefix and the usable range for any address; IP address classes has the full history.

A public address is one that is unique on the whole internet and reachable from anywhere; it is what this page shows at the top and what every website you visit sees. A private address is only meaningful inside one local network. Three IPv4 ranges are set aside for private use by RFC 1918: 10.0.0.0–10.255.255.255, 172.16.0.0–172.31.255.255 and 192.168.0.0–192.168.255.255. Every home router uses one of them, which is why your laptop is 192.168.1.20 in millions of houses at once without conflict: those addresses never leave the house.

The device that bridges the two worlds is the router, using Network Address Translation (NAT). When your laptop at 192.168.1.20 opens a website, the router rewrites the packet so that it appears to come from the home's single public address and a port number it chose, remembers the mapping, and rewrites the reply on the way back. From the outside, a household of thirty devices looks like one address. Many mobile and some fixed-line providers add a second layer, carrier-grade NAT, which shares one public address between many customers using the 100.64.0.0/10 range; if your router's "WAN address" starts with 100.64–100.127, that is what you have. Public vs private IP addresses covers the ranges and the special cases (127.0.0.1, 169.254.x.x) in detail.

A dynamic address is assigned automatically for a limited time by DHCP (Dynamic Host Configuration Protocol, RFC 2131). The lease is renewed quietly while the device stays connected, so in practice the address often stays the same for weeks; it tends to change after a long disconnection, a router reboot or maintenance at the provider. Almost every home connection and every phone has a dynamic public address, and every device on your Wi-Fi has a dynamic private one.

A static address is fixed by configuration and does not change. Servers need one so that DNS can point at them reliably; a home user might want one for remote access to a camera or a game server, or because a service allow-lists addresses. Providers usually sell static public addresses as an add-on; inside your home you can achieve the same effect for free with a DHCP reservation in the router, which always hands the same private address to the same device. If you are unsure which kind you have, note the address shown on the homepage, restart the router, and compare; static vs dynamic IP explains the trade-offs.

An address on its own tells a website four things with reasonable confidence: the network that holds it (your provider or a company), the country, the approximate area the provider serves from (a city or region, from commercial geolocation databases), and the type of connection (home broadband, mobile, business, data center, VPN). That is enough for a shop to show prices in your currency, for a streaming service to apply its regional catalog, and for a fraud system to notice that your card is registered in one country and your address sits in another.

It does not reveal your name, your street address, your phone number or what you do on other sites. Geolocation is a statistical estimate of where the provider's equipment is, typically accurate to a city for fixed lines and often only to a region for mobile networks; it can be wrong by hundreds of kilometers. Linking an address to a person requires the provider's subscriber records, which are released only through legal process. What a stranger can and cannot do with the number is covered in what can someone do with your IP; the IP location finder shows how several databases disagree about the same address, and the card above shows what this site sees about you right now.