We Designed the Fix in 1995. It's 2025. We're Halfway There.

We've known the Internet would run out of addresses since the 1990s. We designed the fix in 1995. It's 2025, and we're still only halfway there.

This isn't a story about technical failure. It's a story about how infrastructure actually changes: as long as the old way still works, the new way stays optional.

The Economics of "Good Enough"

IPv6 solves a real problem—we literally ran out of IPv4 addresses. But Network Address Translation (NAT) provided a workaround good enough to delay the inevitable. Why spend millions and years migrating your enterprise network when you can just... not?

That calculation is finally changing. In February 2024, AWS started charging $43.80 per year for each public IPv4 address¹. Cloud providers are running IPv6-only datacenters. The scarcity is no longer theoretical—it's hitting the balance sheet.

This is how infrastructure transitions actually happen. Not through technical superiority, but when the pain of staying becomes greater than the pain of changing.

Where We Actually Are

Global IPv6 adoption: 45-49% as of late 2025, growing 3-4% annually². The US just crossed 50% earlier this year. At this pace, we might finish by 2045.

But those averages hide the real story:

The leaders: France (78%), Germany (75%), India (74%)³. These countries didn't wait for perfect conditions—they pushed through the complexity.

The mobile miracle: T-Mobile USA runs over 90% IPv6. Reliance Jio in India launched IPv6-native in 2016. Mobile carriers understood something: when every device needs a real address, NAT is a liability, not a solution.

The enterprise gap: IPv6 usage spikes on weekends when people use residential and mobile networks, then drops on weekdays when corporate networks dominate. Your phone is living in the future while your office is stuck in 2005.

The China puzzle: Internal government measurements show 834 million active IPv6 users—about 75% of China's Internet population⁴. External platforms measure anywhere from 22% (Akamai) to 45% (APNIC). The gap isn't contradiction; it's methodology. Measure users with IPv6 capability, you get one number. Measure IPv6 traffic reaching global content networks, you get another. Both reflect real aspects of a partially complete transition.

The Invisible Trap

If you're reading this on your phone, there's an 80% chance you're already using IPv6. You didn't notice because you weren't supposed to.

This invisibility is both the triumph and the trap. Dual-stack networks speak both protocols, applications choose automatically, users see nothing. It works so seamlessly that there's no pressure to finish. We can run both forever, like a city that never tears down the old bridge even after the new one opens.

The transition is designed to be invisible. Which is why it might never actually complete.

What This Means If You Build Things

For most users: Nothing. Your devices handle it. Your apps don't care. The Internet works.

For network operators: You're running two Internets in parallel, for decades. Every router, every firewall, every monitoring tool, every policy—doubled. This is your reality.

For anyone building new infrastructure: IPv6 isn't optional anymore. Not because the old addresses are gone (they're not—they're just expensive), but because the Internet's center of gravity is shifting. Build IPv4-only and you're building backward.

The Long Arc

We might complete this transition by 2045. We might not complete it at all. IPv4 could become like Latin—technically dead but embedded in everything that matters.

The transition won't happen because it's technically correct. It will happen when running the old system becomes more expensive than replacing it.

IPv6 is crossing that threshold. Slowly. Unevenly. Inevitably.

The question isn't whether we'll finish. The question is whether "finished" means 100% adoption or permanent coexistence. History suggests the latter. The economics are starting to suggest the former.

Either way, the halfway point isn't the middle of the journey—it's where momentum finally builds enough that the end becomes visible.

We're there now.

Sources