The story of IPv6 is one of the strangest in internet history. The replacement for IPv4 was finalized in 1998. It solves the problem everyone agreed was urgent in 1995. And here we are, in 2026, with IPv4 still doing most of the work. Roughly half the internet has IPv6 connectivity at this point, and half still doesn't.
This article walks through what the two protocols actually are, why the transition has been so slow, and what — if anything — you need to do about it as a regular user.
Why IPv6 had to happen
IPv4 uses 32-bit addresses. Thirty-two bits can express about 4.3 billion unique values. When this was designed in the 1970s, that sounded like a comfortable cushion — there were maybe a few hundred computers on the entire ARPANET. By the mid-1990s, with the consumer internet booming, it was already obvious that 4.3 billion wouldn't last.
The official "we're out" milestone was February 3, 2011, when IANA (the central registry) handed out its last blocks of IPv4 addresses to the five regional registries. Those regional registries then started running out themselves, one by one. ARIN (covering North America) exhausted its free pool in 2015. RIPE (Europe) followed in 2019. Today, getting new IPv4 address space requires buying it from someone who has more than they need, at market prices that have hovered around $40-50 per address for several years.
The two protocols, side by side
| IPv4 | IPv6 | |
|---|---|---|
| Address length | 32 bits | 128 bits |
| Notation | Decimal, dot-separated | Hexadecimal, colon-separated |
| Example | 192.0.2.42 | 2001:db8::42 |
| Total addresses | ~4.3 billion | ~3.4 × 1038 |
| Header size | 20 bytes (variable) | 40 bytes (fixed) |
| NAT typically used? | Yes, almost always | No, not normally |
| Configuration | DHCP or manual | SLAAC or DHCPv6 |
| Built-in encryption | No | IPSec support standardized (rarely used as base) |
| Year standardized | 1981 (RFC 791) | 1998 (RFC 2460), updated 2017 (RFC 8200) |
What IPv6 changed (and what stayed the same)
The headline number — 128 bits instead of 32 — gets all the attention, but IPv6 isn't only "IPv4 with more digits." A few other things changed:
The header was simplified
The IPv4 header has a bunch of fields that turned out to be either obsolete (like the checksum, which lower layers also calculate) or rarely useful in practice. IPv6's header is fixed-size and skinnier, which makes it easier for routers to process at line rate.
Fragmentation moved to the endpoints
In IPv4, if a packet is too big for the next hop's link, a router can break it into smaller fragments. In IPv6, that's not allowed — the sender is expected to figure out the right packet size in advance (using a technique called Path MTU Discovery) and never send anything that needs to be fragmented mid-route. This makes routing faster and simpler.
Autoconfiguration was built in
IPv6 has SLAAC (Stateless Address Autoconfiguration), which lets a device generate its own address based on the local router's prefix plus a unique identifier — no DHCP server required. This was a big deal in the design vision: every device, anywhere, could just plug in and get a globally unique address.
NAT became optional, and discouraged
In IPv4, basically every home and office runs NAT because you don't have enough public IPs for every device. In IPv6, you have so many addresses that every device on your network gets a real public IP, and NAT (which complicates a lot of protocols) becomes unnecessary. This is good for things like VoIP, peer-to-peer apps, and gaming, which all suffer when there's a NAT in the path.
Why has the transition been so slow?
Given that IPv4 ran out a decade ago and IPv6 has been stable for almost as long, you'd expect the migration to be done by now. It isn't. Adoption hovers around 45% globally as of mid-2026, with huge regional variation. Several things explain the slowness.
NAT was too good
The original urgency around IPv6 was "we're out of addresses!" But NAT — the trick where many devices share one public IP — turned out to be remarkably effective at letting IPv4 stretch further than anyone expected. Carrier-grade NAT (CGNAT), where the ISP itself runs a big NAT for its customers, is now common, especially on mobile networks. CGNAT is operationally ugly, but it works. The "we're out" emergency never quite arrived for end users.
The transition isn't really a transition
IPv4 and IPv6 aren't compatible. An IPv4-only device can't talk to an IPv6-only device. That means any network operator who wants to support IPv6 has to run both protocols (called "dual stack") for a long time, until the IPv4-only holdouts die off. Running both costs twice the configuration work, twice the security review, twice the troubleshooting overhead. Many smaller networks have looked at that and said "later".
Legacy gear and software
A lot of equipment from the 2000s either doesn't support IPv6 at all, or supports it badly. Some industrial systems, point-of-sale terminals, building automation controllers, and embedded devices are still IPv4-only. Replacing them costs money. Until someone forces the issue, they stay.
It's not visible to users
Here's the underrated factor: IPv6 doesn't make your browsing measurably better. Your YouTube video doesn't load faster (much), Slack doesn't connect more reliably (much), your VPN doesn't behave differently (much). For users, IPv6 is invisible when it works. There's no marketing pull, only operator push.
Where IPv6 is winning
Despite the slow average, IPv6 has reached dominance in some places:
- Mobile networks. US mobile carriers like T-Mobile and Verizon have been almost entirely IPv6 internally for years. Your phone is probably using IPv6 right now, even if you didn't notice.
- India. Reliance Jio's massive rollout was IPv6-first, pulling India to one of the highest national adoption rates.
- Large content providers. Google, Facebook/Meta, Netflix, and Cloudflare all serve IPv6 by default. If you're on a dual-stack network, those connections are probably using v6.
- Some ISPs in Europe and Asia. Comcast (US), Sky (UK), Free (France), Deutsche Telekom (Germany), and many Japanese ISPs have done significant IPv6 deployments.
Where IPv6 is stuck
- Small and mid-size enterprises. Internal LANs at most companies are still v4-only.
- Older home routers. Many devices from before ~2018 either don't support v6 or do so flakily.
- VPN gateways and cloud configurations. Lots of corporate VPNs and a surprising amount of cloud infrastructure is set up assuming v4 only.
- Africa and parts of Latin America. Adoption is lower in regions where ISPs prioritize buying more v4 over the dual-stack project.
What does this mean for you?
For a normal user, the practical answer is: very little, most of the time. Your operating system and browser handle the choice between v4 and v6 automatically. If a site has both, your device prefers v6 (this is called "Happy Eyeballs" — the algorithm tries both and uses whichever connects first).
A few situations where it matters:
- Hosting your own services. If you run a home server and want it reachable from the outside, IPv6 (when your ISP supports it) avoids CGNAT headaches.
- Online gaming and voice/video. Direct peer-to-peer connections are easier without NAT in the path. IPv6 helps a lot of games behave better.
- Privacy. Some IPv6 addresses use a stable identifier based on your device's hardware address. Modern OSes default to "privacy extensions" that randomize this, but check the setting if you care.
The future, briefly
IPv4 is not going away in the next decade. It's too cheap to keep running, and too many systems depend on it. What's likely is a long tail: more and more of the internet's edge — the connections between users and big content providers — runs over IPv6, while v4 quietly persists for legacy reasons in the middle and the corners. Eventually, perhaps in the 2030s, you'll meet ISPs that don't sell v4 connectivity at all. Until then, dual stack is the world.
Related reading: What is an IP address, really? · How IP geolocation actually works