In 1981, 4.3 billion addresses seemed infinite. The world had 4.5 billion people and a few thousand networked computers. The idea that we’d need an IP address for a thermostat, a car, a watch—that would have sounded like science fiction.

By 2011, we were out of addresses. Not running low. Out.

The Ceiling Was Always There

IPv4 uses 32 bits per address: 2³² combinations, which equals 4,294,967,296 addresses total.

But the usable number is lower. Large chunks are reserved—private networks (192.168.x.x), loopback testing, multicast, documentation ranges. The ceiling was fixed before the first device connected.

What wasn’t fixed was how many devices would want addresses.

The Exhaustion Timeline

February 3, 2011: IANA allocated the last IPv4 blocks to regional registries. The global pool was empty.

Then each region ran dry:

Region	Registry	Exhaustion Date
Asia-Pacific	APNIC	April 2011
Europe	RIPE NCC	September 2012
Latin America	LACNIC	June 2014
North America	ARIN	September 2015
Africa	AFRINIC	April 2017

By 2019, every registry had exhausted its supply. Yet the Internet kept growing.

How We Survived

NAT changed what an address means. Network Address Translation didn’t just extend IPv4’s life—it redefined the relationship between devices and addresses.

Before NAT, every device needed its own public address. Your computer had a number, and the Internet knew that number. After NAT, your router has a public address, and your devices hide behind it. The router remembers which internal device requested what, translates responses back, and the Internet sees one address where dozens exist.

NAT has costs—peer-to-peer connections get complicated, some applications break, every connection must be initiated from inside—but it transformed the economics completely. One address could suddenly serve a household, an office, an apartment building.

NAT didn’t extend IPv4’s life. It changed what an address means—from "who you are" to "who speaks for you."

A market emerged for the addresses that remained. Organizations sitting on blocks from the 1990s realized they owned something valuable. In 2011, Microsoft paid $7.5 million for 666,624 addresses—about $11 each. By early 2024, prices exceeded $50 per address. In 2025, large blocks trade around $24, smaller ones at $31-35.

IPv4 addresses became real estate. Finite, transferable, appreciating.

IPv6 offers the permanent fix. With 128-bit addresses, IPv6 provides 340 undecillion addresses—enough to give every atom in a human body its own address, for every human who will ever live. Exhaustion becomes mathematically absurd.

But IPv4 and IPv6 don’t interoperate. You can’t just upgrade. Networks must run both protocols simultaneously during migration, which means double the complexity. So IPv6 adoption creeps forward while IPv4 limps along on NAT and traded addresses.

What the Architects Couldn’t See

The people who designed IPv4 made a reasonable bet. They built for universities and research labs, maybe eventually businesses. They designed a system that could address every computer on a planet that barely had computers.

They weren’t wrong. They just couldn’t see a future where computers would be cheap and addresses would be expensive.

The lesson isn’t about their failure. It’s about hard limits: 32 bits gives you 2³² addresses. No amount of clever engineering changes that ceiling. You can work around it—NAT proves that brilliantly—but you can’t eliminate it.

We’re still in the long transition. Most networks run dual-stack, supporting both protocols. IPv4 continues on borrowed time. IPv6 grows slowly, inevitable but unhurried.

Eventually this will be history—the story of the day the Internet ran out of addresses and kept running anyway, through ingenuity, market forces, and patient migration.

But the ceiling was always there. We just couldn’t see it until we hit our heads on it.