The Address Space Ran Out and It Became a Line Item

The original internet address format allows roughly four billion addresses, which was an unimaginable number when it was chosen and is not one now. The regional bodies that hand out addresses exhausted their available pools years ago and have been operating on waiting lists and recovered fragments since. The consequence is not a technical failure — the internet works fine — but an economic one: addresses became scarce, and scarce things acquire a price. There is now an established secondary market where blocks change hands, complete with brokers, valuations and due diligence on whether a block's reputation has been ruined by previous abuse. Most organisations first encounter this through an invoice rather than an outage. Cloud providers, which hold large allocations and pay to maintain them, began charging explicitly for each public address attached to a resource. The individual amount is small enough to ignore, which is precisely why it grows: an estate that spins up addresses per environment, per test instance, per load balancer, per short-lived experiment, accumulates a charge nobody chose. Teams that audit this typically find a meaningful share of billed addresses attached to nothing at all, reserved by a resource that was deleted around it. The replacement protocol, with an address space large enough that scarcity is not a meaningful concept, has existed for a very long time and is now carried by most mobile networks and a large share of residential connections. Adoption is nonetheless uneven, and the reason is rarely the network layer. It is everything above it: monitoring that parses addresses with assumptions about their shape, access control lists written by hand, geolocation and fraud rules keyed to addresses, logging pipelines with a fixed-width column, licence servers, and a long tail of internal tools written when there was only one format to worry about. The workaround that made scarcity survivable — sharing one public address among many users behind a translation layer — is also the source of a persistent second-order problem. When hundreds of subscribers share an address, that address is a poor identifier: abuse reports implicate innocent users, rate limits catch a whole neighbourhood, and troubleshooting requires correlating logs that most parties do not keep. The practical position for most organisations is dual operation, running both formats for a long transitional period, and the honest budgeting question is not what migration costs but what the delay costs per year in addresses rented rather than owned.