What Is IPv6? How It Works, Why It Matters, and What Changes
What is IPv6, and why does it matter? Learn how IPv6 addresses work, why IPv4 ran out, what a /48 means, and when you need your own prefix.
Introduction
You've probably seen it without recognizing it: 2a04:4e42::347 — a string of hex digits and colons that looks like someone mashed a keyboard. That's an IPv6 address. And at some point in the next few years, your servers, your customers, and your infrastructure will need to speak it.
Not because IPv6 is "faster." Not because it's "the future." But because IPv4 — the protocol that has carried Internet traffic since 1983 — has run out of room. Every new device, every new service, every new network that wants a public address on the Internet now competes for a pool that stopped growing years ago.
This guide explains IPv6 from the ground up: why it exists, how it works, what a prefix actually is, and when you need more than what your cloud provider hands you. No assumptions beyond basic Internet literacy.
IPv6 (Internet Protocol version 6) — the successor to IPv4, designed to replace its 32-bit addressing scheme with a 128-bit address space, restoring end-to-end connectivity without Network Address Translation.
IPv6 replaces address scarcity with enough room to make network structure clear and deliberate.
Why IPv4 Ran Out
IPv4 uses 32-bit addresses. That gives you about 4.3 billion possible addresses — a number that seemed inexhaustible in 1983 when the Internet connected a few hundred research institutions. By 2011, IANA had allocated the last free /8 blocks to the Regional Internet Registries. By 2019, RIPE NCC — the RIR for Europe, the Middle East, and parts of Central Asia — announced it had no remaining IPv4 space to distribute to new members.
The workaround was NAT: Network Address Translation. One public IPv4 address, many private devices behind it. NAT kept the Internet growing, but at a cost:
| Problem | What NAT Does | Why It Hurts |
|---|---|---|
| End-to-end broken | Hides devices behind a shared public IP | Peer-to-peer protocols, VoIP, and direct connections require complex workarounds |
| State overhead | Every connection needs a mapping table entry | Large NAT gateways become performance bottlenecks and single points of failure |
| Address conflict | Private ranges (10.0.0.0/8, 192.168.0.0/16) collide | Merging networks or connecting to partners means renumbering |
| Carrier-grade NAT | ISPs layer NAT on top of NAT | Shared public IPs mean one user's bad behavior blacklists everyone sharing that address |
| Cost | IPv4 addresses now trade on secondary markets | A single /24 (256 addresses) can cost thousands of euros; a larger block costs tens of thousands |
NAT didn't solve the IPv4 shortage. It bought time. IPv6 solves the problem at the protocol level.
What IPv6 Actually Changes
IPv6 replaces 32-bit addressing with 128-bit addressing. The headline number — 340 undecillion addresses — is less interesting than what the larger address space enables:
End-to-end addressing becomes practical again. Every device can have its own globally unique, publicly routable address, without using NAT merely to conserve public addresses. That removes a major source of port-forwarding workarounds and address conflicts when networks connect. It does not remove the need for firewalls: public reachability should still be an explicit security decision.
Address autoconfiguration is built in. IPv6 devices generate their own addresses through Stateless Address Autoconfiguration (SLAAC), using router advertisements and the device's own interface identifier. No DHCP server required, though DHCPv6 exists for organizations that want centralized address management. A device plugs in, hears a Router Advertisement, and gets an address.
Routing scales better. IPv4's routing table carries over a million entries and grows constantly — every new network means another route. IPv6 was designed with address aggregation in mind: Regional Internet Registries allocate large contiguous blocks to ISPs, who allocate sub-blocks to customers. Routes summarize cleanly. A single /32 covers an entire ISP's customer base rather than hundreds of scattered /24s.
Header simplification. The IPv6 header is fixed at 40 bytes, compared to IPv4's variable-length header (20–60 bytes with options). Most extension headers are processed only at the destination, so intermediate routers handle a simpler, faster path.
| IPv4 | IPv6 | |
|---|---|---|
| Address length | 32 bits (4.3 billion addresses) | 128 bits (340 undecillion addresses) |
| Address format | Dotted decimal: 192.168.1.1 |
Colon-separated hex: 2001:db8::1 |
| Header size | 20–60 bytes (variable) | 40 bytes (fixed) |
| NAT for address conservation | Common in most deployments | Not required; some deployments still use translation for other reasons |
| Autoconfiguration | DHCP or manual | SLAAC (stateless) or DHCPv6 |
| Broadcast | Yes — ARP, broadcast storms possible | No — replaced by multicast Neighbor Discovery |
| Minimum MTU | 576 bytes | 1,280 bytes |
| IPsec | Optional extension | Originally mandatory, now recommended |
IPv6 Addresses: What They Look Like
An IPv6 address is 128 bits written as eight groups of four hexadecimal digits, separated by colons:
2001:0db8:0000:0000:0000:ff00:0042:8329
Two rules compress this significantly:
Omit leading zeros within each group: 2001:db8:0:0:0:ff00:42:8329
Replace one contiguous run of zero-groups with :: (once per address): 2001:db8::ff00:42:8329
The :: is a "fill in the zeros from here" marker. It can only appear once — otherwise the address becomes ambiguous. The RIPE NCC documentation address 2001:db8::1 expands to 2001:0db8:0000:0000:0000:0000:0000:0001.
Common Address Types
| Type | Prefix | Purpose | Example |
|---|---|---|---|
| Global Unicast | 2000::/3 |
Publicly routable Internet addresses — the IPv6 equivalent of a public IPv4 address | 2a04:4e42::347 |
| Link-Local | fe80::/10 |
Valid only on a single network segment; every IPv6 interface has one automatically | fe80::1%eth0 |
| Unique Local (ULA) | fc00::/7 |
Private addresses for internal networks — roughly equivalent to IPv4's 10.0.0.0/8, but with a globally unique random prefix to avoid collisions | fd12:3456:789a::1 |
| Loopback | ::1/128 |
The local machine — equivalent to IPv4's 127.0.0.1 | ::1 |
| Multicast | ff00::/8 |
One-to-many delivery to subscribed listeners | ff02::1 (all nodes on the local link) |
A practical note: you'll mostly encounter Global Unicast addresses in production. Link-local addresses appear in routing tables and neighbor tables (fe80:: next-hops are normal in BGP). ULA is for internal-only services that must never reach the Internet. If you see an address starting with 2 — that's a real, publicly routable IPv6 address.
Prefixes, /48, and /64: Building Intuition
In IPv4, you're used to thinking about individual addresses: "I need a public IP for this server." In IPv6, the smallest unit you typically work with is a prefix — a block of addresses.
A prefix is written in CIDR notation: 2001:db8:abcd::/48. The /48 means "the first 48 bits identify the network; the remaining 80 bits are available for subnetting and host addresses within that network."
Here's the mental model shift: organizations receive a prefix, not individual addresses. An IPv6 /48 allocation is a block from which you carve out smaller subnets — normally /64s — and assign addresses within those subnets.
| Prefix Length | What It Represents | Typical Use |
|---|---|---|
/32 |
ISP-level allocation | An ISP allocating /48s to thousands of customers |
/48 |
Standard end-user organization allocation | Your network's address block — carve into /64 subnets |
/56 |
Smaller allocation (sometimes from ISPs to residential customers) | A home network or small office |
/64 |
Standard LAN subnet | One VLAN, one broadcast domain — the unit SLAAC requires |
/128 |
Single host address | A loopback interface, a BGP router ID, or a specific anycasted service |
A /48 contains 65,536 /64 subnets. That's enough to assign a unique /64 to every VLAN in a campus network, every VPS in every data center you operate, and every point-of-presence in your anycast deployment — with room for a century of growth. You don't subdivide because you need the space now. You subdivide because the convention of giving every LAN a /64 makes addressing predictable, and the abundance of space makes it free.
We'll go deeper into subnetting in the third article in this series — for now, the key intuition is: the number after the slash is the network boundary, and a /48 is a big box of /64-size compartments.
IPv6, ASNs, and BGP: How the Pieces Fit
An IPv6 prefix is address space. An ASN is a network identity. BGP is the protocol that announces your prefix to the rest of the Internet, saying "this address space lives here, reachable through my network."
The relationship is:
- You receive an IPv6 prefix — typically a /48 — from a Regional Internet Registry (directly or through a Sponsoring LIR)
- You have an Autonomous System Number that identifies your network
- You configure RPKI Route Origin Authorizations that cryptographically bind your ASN to your prefix
- You use BGP to announce your prefix to upstream providers, who propagate the route globally
- Traffic destined for your prefix finds your network through the Internet's routing table
Without an ASN, you can still use IPv6 — your ISP or cloud provider gives you addresses from its own allocation. But to originate your own prefix with BGP and build multihomed infrastructure, you normally need an ASN. Provider-assigned space is usually tied to that provider, so moving providers often means renumbering. The ASN and a suitable prefix are the combination that gives an operator routing autonomy.
Do I Need My Own IPv6 Now?
The answer depends on what you're building, not on how large your organization is.
You probably don't need your own IPv6 prefix if:
- Your application runs entirely on a cloud platform (AWS, Vercel, Fly.io) that provides IPv6 connectivity
- You have a single server or VPS and your provider already gives you IPv6 addresses
- You don't operate BGP, don't have an ASN, and don't plan to
You probably do need your own IPv6 prefix if:
- You operate or plan to operate an ASN — the prefix is what you announce
- You need a portable, provider-independent prefix and understand the sponsoring-LIR arrangement that comes with it
- You're building infrastructure that spans multiple data centers, providers, or regions
- You're deploying anycast services that need the same IP reachable from multiple locations
- You need consistent routing policy and IP reputation that isn't tied to a single provider's allocation
For most organizations, the progression is: start with your provider's IPv6 → get your own ASN when you outgrow single-provider → get your own prefix with the ASN. The two resources work together.
The next article in this series covers exactly how to get that IPv6 /48: the paths available, the trade-offs between PA and PI space, and which one fits your situation.
Frequently Asked Questions
Is IPv6 faster than IPv4?
Not by design. IPv6 has a simpler header structure and eliminates NAT processing overhead, which can reduce per-packet work on routers. In some deployments, IPv6 paths show slightly lower latency because they take different — sometimes more direct — routes. But IPv6 itself is not a performance optimization; it's an addressing and routing architecture improvement. If your IPv4 path is already efficient, switching to IPv6 won't suddenly halve your latency.
Can I use IPv6 alongside IPv4?
Yes. This is called dual-stack operation: every device has both an IPv4 and an IPv6 address, and applications choose which to use based on destination address availability. Most modern operating systems and network stacks prefer IPv6 when both are available (Happy Eyeballs, RFC 8305). Dual-stack is the standard transition strategy — you don't turn off IPv4; you add IPv6 and let traffic migrate over time.
Do I need to reconfigure my firewall for IPv6?
Yes. IPv6 traffic is not automatically subject to the same rules as IPv4. You need separate firewall rules for IPv6, and you need to think about address scope: filtering 2001:db8:abcd::/48 is the IPv6 equivalent of filtering 203.0.113.0/24, not 10.0.0.0/8. Many security incidents involving IPv6 happen because organizations deploy it on their internal networks without applying the same filtering discipline they use for IPv4.
What happens if I don't adopt IPv6?
Your network becomes increasingly reliant on IPv4 addresses — which are scarce, expensive, and subject to secondary-market pricing. New customers and devices in IPv6-only markets (mobile networks in many countries are already IPv6-only with NAT64 for IPv4 reachability) may have degraded access to your services if you don't offer IPv6 connectivity. It's not an emergency for most organizations in 2026, but the cost of inaction rises each year.
Does IPv6 have private addresses like 192.168.x.x?
IPv6 has Unique Local Addresses (ULA, fc00::/7) which serve a similar purpose — internal-only addresses that are not routable on the public Internet. The difference is that ULA prefixes include a randomly generated 40-bit Global ID, making collisions between separately managed networks astronomically unlikely. When you merge networks or set up a VPN between two organizations, you won't have the address overlap problems that plague IPv4 private ranges.
Do I need a /48, or is a /56 enough?
A /48 is the standard RIR allocation for organizations that demonstrate their own infrastructure need. A /56 is more commonly delegated by ISPs to residential or small-business customers. The difference: a /56 gives you 256 /64 subnets; a /48 gives you 65,536. Most organizations will never exhaust 65,536 subnets, but the /48 is the policy default for end-user allocations in the RIPE NCC service region. If you're applying through a Sponsoring LIR, a /48 is what you'll typically receive.
Can I get IPv6 without an ASN?
Yes — your ISP, hosting provider, or cloud platform can assign you IPv6 addresses from their own allocation. You can use those addresses for your servers and services. But you don't control the prefix: if you change providers, you renumber. If you want to announce the prefix yourself via BGP and control your own routing, you need an ASN.
Conclusion
IPv6 is not a performance upgrade to IPv4. It's a correction to the architectural decision that capped the Internet's address space at 4.3 billion — a limit that made sense when the Internet connected a few thousand hosts and became a constraint when it connected billions of devices and services.
The protocol change from 32-bit to 128-bit addressing restores what the early Internet had and what NAT took away: every device directly reachable at a globally unique address, every network able to connect to every other network without translation layers, and address space abundant enough that subnet planning is about organization rather than conservation. That's the engineering vision IPv6 delivers. The question for most organizations isn't whether they'll adopt it — it's whether they'll do it before their IPv4 constraints force the decision.
Key Takeaways
- IPv6 replaces IPv4's 32-bit addressing with 128-bit addressing, restoring end-to-end connectivity without NAT.
- A /48 is a common organization-sized IPv6 prefix and contains 65,536 /64 subnets.
- /64 is the standard LAN subnet size, required by SLAAC for stateless address autoconfiguration.
- IPv6 adoption is not about speed — it's about address space, routing scalability, and architectural simplicity.
- To originate your own IPv6 prefix with BGP, you need an ASN; otherwise, you normally use addresses routed by your provider.
- Dual-stack (IPv4 + IPv6 running simultaneously) is the standard transition strategy in 2026.
In One Sentence
IPv6 restores end-to-end Internet addressing at a scale that makes address scarcity obsolete — and owning your own prefix turns it from a connectivity detail into an infrastructure asset.
Continue Reading
- How to Get an IPv6 /48: PA vs PI Address Space Explained — the paths to obtaining your own IPv6 prefix, from ISP delegation to RIR allocation, and how to choose between PA and PI space.
- IPv6 Subnetting Explained: /48, /64, and How Many Addresses You Actually Need — how to carve a /48 into usable subnets, plan your addressing scheme, and avoid the most common mistakes.
- What Is an ASN? A Complete Beginner's Guide — the network identity that turns your IPv6 prefix into a resource you can announce and control.
- BGP Basics: How the Internet Learns Where to Send Traffic — the protocol that distributes your IPv6 routes across the Internet.
- What Is RPKI? How Modern Networks Prevent BGP Hijacking — how to cryptographically bind your ASN to your IPv6 prefix so your routes aren't hijacked.
- What Happens After You Get an ASN? A Complete BGP Setup Guide — the step-by-step sequence that turns an ASN and IPv6 allocation into a working network.
Ready to get your own ASN?
Full RIPE NCC LIR sponsorship — ASN + /48 IPv6, one flat fee. RIPE NCC compliant, UK-based LIR.
See Pricing →