Short answer: IPv4 addresses are scarce and expensive, and most of the waste doesn't sit with users — it sits in how hosting providers design their networks. We attack it from two sides: every VPS gets a routed /32 — one address, fully usable, without the three addresses of overhead a classic subnet costs — and if you don't need IPv4, you run IPv6-only with us. That keeps addresses available for servers that genuinely need them, and means you don't pay for addresses that do nothing.
Here's why this matters and how it works technically.
IPv4 isn't "almost gone" — it's gone
RIPE NCC — the organisation that hands out IP addresses in Europe — made its final regular IPv4 allocation in November 2019. What remains is a waiting list: an LIR can request a single /24 (256 addresses) from address space that trickles back into the pool, and applicants queue for months to get it. Need more, or need it sooner? Then you buy on the second-hand market, where a single IPv4 address now costs tens of euros. That's not a one-off purchase that earns itself back; it's dead capital that gets priced into every hosting bill.
There are two things you can do about that. Pass it on and move on — or be careful with what exists. We choose the second, because the maths below shows how much there is to gain.
Where the waste sits: the classic subnet
The traditional way to give a server an IP address is its own subnet. And a regular subnet comes with fixed overhead: a network address, a broadcast address and a gateway address. Three addresses gone before your server uses a single one.
| Subnet | Total addresses | Overhead | Usable | Waste |
|---|---|---|---|---|
| /31 point-to-point (RFC 3021) | 2 | 1 | 1 | 50% |
| /30 | 4 | 3 | 1 | 75% |
| /29 | 8 | 3 | 5 | 37.5% |
| /28 | 16 | 3 | 13 | 18.75% |
| /32 routed | 1 | 0 | 1 | 0% |
The leanest classic option is a /31 point-to-point (RFC 3021): it skips the network and broadcast addresses, but the gateway still claims one of the two — half gone. The smallest regular subnet is a /30, of which three quarters is lost. Multiply that by thousands of VPSes and you see why entire ranges around the world sit occupied without a single packet ever crossing them.
Solution 1: a /32 per VPS
That's why VPSes with us don't get a subnet but a routed /32: exactly one address, and that one address is fully usable. There's no network or broadcast address, and the gateway lives outside your own address — your server learns it through an on-link route, which your operating system or cloud-init handles automatically. For you it works exactly like it always has; there's just nothing lost anymore.
This is possible because our network is an EVPN-VXLAN fabric: every address is distributed through the network as an individual host route, so an address doesn't need to sit in a contiguous block next to its neighbours. A bonus: if your VPS moves to another node or another datacenter, your address simply moves with it.
It helps on the security side too. In a shared subnet, customers sit together on the same Layer 2 segment, with everything that comes with that — ARP spoofing, broadcast traffic from your neighbours. With a /32 per customer, that segment simply doesn't exist.
Solution 2: IPv6-only for those who don't need IPv4
The second question is more honest: does every server even need an IPv4 address? Often not. A database only reached by your own application, a worker processing queues, a development environment, a backend behind a load balancer or CDN — those are perfectly fine with IPv6 alone. Every large network you want traffic from (Google, Cloudflare, mobile carriers) has spoken IPv6 for years.
That's why we also offer VPSes as IPv6-only. You get a generous IPv6 allocation — more addresses than any setup could ever need — and the IPv4 address you don't use stays available for a server that can't do without one. If that server still occasionally needs to reach an IPv4-only destination, there are clean solutions for that, from NAT64 to a reverse proxy on a shared address; tell us your situation and we'll think along.
And if you need a dedicated IPv4 address later after all? You just add one. You start lean and scale up when you must — not the other way around.
Beyond that: no sleeping ranges
Being careful doesn't stop at allocation. Addresses that come free go back into our pool, not into a drawer; ranges reserved "for later" don't exist with us. And we hand out extra addresses the way RIPE intended: based on actual usage, not as bundle filler.
Already have address space of your own, say your own /24? Bring it — on a VPS or in colocation we'll announce your own range. And if you don't have your own space yet but want it, our LIR services will get you your own allocation or ASN.
What you'll notice
Very little — and that's exactly the point. Your VPS works the way you're used to. But under the hood it means: no dead capital in your monthly price, addresses available the moment you scale up, and a host that won't knock on your door in five years because its own supply ran dry.
IPv4 will never get cheaper or more plentiful. The only sustainable route is to use it carefully and make the road to IPv6 easy — and that's simply part of decent infrastructure, just like the network itself. Curious what this means for your setup? Have a look at our VPSes or ask us in the chat.