# Registered but not routed: what a global audit of IPv6 says about deployment

**URL:** <https://community.ipinfo.io/t/registered-but-not-routed-what-a-global-audit-of-ipv6-says-about-deployment/7434>\
**Category:** General\
**Created:** [October 4, 2026, 5:04pm UTC](https://community.ipinfo.io/t/registered-but-not-routed-what-a-global-audit-of-ipv6-says-about-deployment/7434 "2026-10-04T17:04:01Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Abdullah](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.ipinfo.io/abdullah/32/4973_2.png) [@Abdullah](https://community.ipinfo.io/u/Abdullah)\
**Post date:** [October 4, 2026, 5:04pm UTC](https://community.ipinfo.io/t/registered-but-not-routed-what-a-global-audit-of-ipv6-says-about-deployment/7434/1 "2026-10-04T17:04:01Z")

</div>

#### Networks around the world have registered enough IPv6 address space to hand every star in the Milky Way its own network a thousand times over. A large share of it never appears on the internet at all. We measured how much, and who it belongs to.

Every network that wants to use IPv6 has to register its address space with a Regional Internet Registry, then declare which autonomous system will route it, then actually announce that space through the Border Gateway Protocol so the rest of the internet can reach it. Those three steps do not always happen together. A network can complete the first two and never do the third, leaving address space that exists on paper but nowhere in practice.

We wanted to know how often that happens, and whether it happens more in some parts of the world than others. So we compared registration records from the five Regional Internet Registries against a live snapshot of the global BGP routing table, broken down by network operator, country, continent, and a Global North and South split. We also cross-checked the results against the IPv6 Hitlist project, which tracks addresses known to respond to real probes, to get a sense of how much of the registered space corresponds to an actual device on the other end.

The short version: 92.5 percent of registered IPv6 netblocks show up in live routing, but only 79.4 percent of the underlying address space does. That gap of roughly one fifth is not spread evenly. It concentrates in a small number of very large allocations that were registered years ago and never announced, and the pattern of where that happens differs sharply between wealthier and poorer regions.

## **Two ways to measure the same gap**

There are two reasonable ways to ask “how much IPv6 space is actually in use.” You can count netblocks, treating a tiny customer assignment and a continent-spanning carrier allocation as equally important. Or you can weigh by address space, where one enormous dormant allocation counts for more than a million small ones combined. We calculated both, because they tell different stories.

Split the world into a conventional Global North and Global South and the two measures point in opposite directions.

 ![image](https://canada1.discourse-cdn.com/flex035/uploads/ipinfo/original/2X/3/304b92b6511b0c80ddd2dd31ee4459b0df3471c9.png)

> Figure 1. Global North figures cover North America, Europe, Russia, Japan, South Korea, Israel, Australia, New Zealand and Singapore. Everything else is counted as Global South, following the conventional Brandt-line style split used in development economics.

In the Global North, 97.1 percent of registered netblocks are announced, but that share drops to 77.5 percent once you weigh by address space. In the Global South, it runs the other way: only 69.4 percent of netblocks are announced, yet 84.6 percent of the address space is.

The explanation seems to be structural rather than a simple story about who tries harder. Wealthier regions have accumulated millions of small, individually registered netblocks over two decades of IPv6 rollout, and nearly all of them ride along under their parent ISP’s announcement without needing one of their own. What drags the address-space figure down is a much smaller number of very large legacy allocations, some dating back to the early IPv6 era, that were registered and then simply never used.

The Global South’s IPv6 space tends to be concentrated in a handful of enormous mobile carrier allocations, the kind national telecom operators receive to serve hundreds of millions of subscribers. Once one of those is announced, it accounts for a large share of the region’s total address space in one move. But many smaller assignments in the same regions never get an individual route, which is consistent with address space managed behind carrier-grade NAT rather than routed to individual customers.

## **A regional breakdown, and one clear outlier**

Looking at continents instead of a North/South split sharpens the picture, and turns up one striking anomaly.

 ![image](https://canada1.discourse-cdn.com/flex035/uploads/ipinfo/original/2X/b/b302a50de2908fefa0d6b3b22f948a718e781cfe.png)

> Figure 2. Share of registered IPv6 address space announced in live BGP, by continent.

Africa has the highest share of announced address space of any continent, at 92 percent, ahead of North America, South America, Europe and Asia, which all cluster between 79 and 89 percent. Oceania is the exception, at just 34.8 percent.

That number is misleading on its own. It comes almost entirely from a single legacy allocation, registered to Orange S.A. under a New Zealand registration handle, that holds a vast amount of address space and announces only 0.14 percent of it. Remove that one network from the calculation and Oceania’s figure rises to roughly 81 percent, in line with the rest of the world. It is a useful reminder that regional aggregates can hide a lot in a handful of very old, very large records.

## **Does size buy better habits?**

One natural question is whether the largest, most sophisticated networks keep their registrations cleaner than smaller ones. We plotted each ranked network’s position in APNIC Labs’ IPv6 population estimates against the share of its address space that is actually announced.

 ![image](https://canada1.discourse-cdn.com/flex035/uploads/ipinfo/original/2X/3/39810af46ab1183ee26a6ff8eefc6352e0b7a751.png)

> Figure 3. Each point is one network with a resolvable IPv6 registration record, ranked by APNIC Labs’ estimate of its IPv6 user population. If larger networks kept cleaner records, the chart would trend upward from left to right. It largely does not.

Most networks, large and small, sit close to full announcement. But there is no clear trend by size. Some of the world’s largest carriers, serving tens of millions of users, still carry meaningful unannounced address space, while plenty of smaller, less prominent networks route everything they hold. Registration hygiene looks more like an organizational habit than a function of scale.

## **Who is sitting on the most dark address space**

Ranking networks by the absolute amount of unrouted address space, rather than by percentage, surfaces a different set of names. A number of them are hosting or VPS providers, which is consistent with a known pattern in that industry: registering address space in bulk with little intention of routing most of it.

| Network | Country | Share announced |
| --- | --- | --- |
| Orange S.A. (AS5511) | New Zealand | 0.14% |
| [GoDaddy.com](http://GoDaddy.com), LLC (AS26496) | Singapore | 0.07% |
| WorkTitans B.V. (AS209847) | Europe (regional record) | 9.80% |
| Vault Host (AS46573) | Cyprus | 0.00% |
| ULTRA ONE (AS399989) | United States | 0.00% |
| LLC Vpsville (AS204916) | Russia | 56.10% |
| Tencent (AS132203) | China | 49.80% |
| Datacamp Limited (AS212238) | Poland | 60.90% |
| KIEVLINE LLC (AS29491) | Ukraine | 3.40% |
| RCN (AS6079) | United States | 69.90% |

**A paradox at the country level**

The netblock-versus-address-space split shows up even more starkly inside individual countries. India announces only 4.8 percent of its registered netblocks, yet 95.6 percent of its address space, because a couple of mobile carriers hold the vast majority of the country’s IPv6 space and route it in full, while a long tail of smaller registrations never gets an individual announcement. South Korea shows almost the same pattern: 28.7 percent of netblocks but 98.8 percent of address space. China sits in between, at 48.4 percent and 88.9 percent.

None of this means those countries are routing IPv6 badly. It means their address space is organized around a small number of very large carriers rather than a broad base of individually routed networks, which produces exactly this kind of divergence between the two metrics.

## **What is actually there**

Knowing that a block of address space is announced tells you it is reachable. It does not tell you anything is actually listening on the other end. For that, we turned to the IPv6 Hitlist project run by the Technical University of Munich, which maintains the largest known list of IPv6 addresses confirmed to respond to real measurement probes. Full scans of the IPv6 address space are not computationally feasible, so this is the closest thing available to a device census.

After removing about 38,600 addresses that only appeared responsive because they sit inside known “every address answers” prefixes, which is a common misconfiguration rather than a sign of real devices, we were left with 28.1 million genuinely distinct, confirmed-live hosts.

 ![image](https://canada1.discourse-cdn.com/flex035/uploads/ipinfo/original/2X/a/a6e813bcbcd53310224709802850590f99f9e856.png)

> Figure 4. Genuine live IPv6 devices confirmed by the IPv6 Hitlist project, by continent, after removing aliased addresses. Note the logarithmic scale

Twenty-eight million confirmed devices sounds substantial until you set it against the scale of the address space involved. The portion of registered IPv6 space that is never announced at all works out to roughly 212 trillion dark /64 subnets, more than a thousand times the number of stars in the Milky Way. Expressed as /48 blocks, the size typically assigned to a single home network, that unused space would be enough to give about 40 percent of humanity a subnet of their own, unused.

Africa is a particular case worth noting. It has the highest share of announced address space of any continent, at 92 percent, yet the fewest confirmed live devices by a wide margin, just 52,000. That combination suggests address space that is properly registered and routed, but where relatively few devices are directly reachable from the open internet, likely due to a mix of limited deployment, firewalling, and carrier-grade NAT even on IPv6 networks.

* * *

## **Methodology**

We built the “registered” side of this analysis from whois netblock records published by the five Regional Internet Registries, covering roughly 1.5 million individual IPv6 netblocks. For each netblock, we resolved the network operator’s declared origin autonomous system using Internet Routing Registry route6 objects, which record who intends to route a given block independently of whether they currently do. We then checked that declared network’s actual, live BGP table, built from a recent RIPE RIS and RouteViews full-table snapshot and filtered for routes invalidated by RPKI or IRR records, to see whether it covers that block. A netblock that is reachable only because it falls inside a larger announcement from its parent ISP counts as announced, which is the normal and expected situation for the great majority of individual customer assignments.

Network rankings and estimated user counts come from APNIC Labs’ IPv6 population estimation project, which infers per-network user counts from advertising-based measurement. Where a single autonomous system appeared under multiple countries, which happens for satellite and global content delivery networks, we summed its user estimates and used its best available rank.

The Global North and Global South grouping follows a conventional, simplified split used in development economics: North America, Europe, Russia, Japan, South Korea, Israel, Australia, New Zealand and Singapore are counted as Global North, and everything else as Global South. This is offered for legibility, not as a claim about any individual country’s status.

Device counts come from the IPv6 Hitlist project’s list of addresses confirmed responsive to real probes, after excluding addresses that fall inside ranges flagged as aliased, meaning every address in the range responds regardless of whether a real host is present. Because scanning the full IPv6 address space is not computationally feasible, this list reflects previously observed and inferred-active addresses rather than a literal, complete census, and should be read as a lower bound and an illustration of scale rather than a precise population figure.