What is IPv4 exhaustion (and what's actually happening with it)?

You've heard "the internet is running out of IPv4 addresses." Technically that happened a decade ago. But the internet kept growing — through 2026, it's still growing — without grinding to a halt or forcing universal IPv6 migration. So what's actually going on?

The story of IPv4 exhaustion is one of the more interesting near-misses in internet history. It explains why your home network might be behind CGNAT, why some ISPs charge for static IPs, why IPv4 addresses now trade like commodities, and why IPv6 deployment has been steady but slow.

The two-line summary

IPv4 has 4.3 billion possible addresses. The world has more than 5 billion internet users plus tens of billions of connected devices. The math stopped working in the 2010s. We've adapted by sharing IPs (CGNAT), trading allocations (the IPv4 market), and slowly shifting to IPv6.

A quick recap: how addresses get allocated

IANA is the global authority for IPv4. They distribute large blocks (/8s, each containing ~16.7 million addresses) to the five regional internet registries (RIRs):

ARIN (North America)
RIPE NCC (Europe, Middle East, Central Asia)
APNIC (Asia Pacific)
LACNIC (Latin America, Caribbean)
AFRINIC (Africa)

The RIRs sub-allocate to ISPs and large organizations. ISPs assign to subscribers. Large orgs assign internally.

When the system was designed in 1981, 4.3 billion addresses seemed sufficient. They aren't.

When did exhaustion actually happen?

The first exhaustion was at IANA — the global pool. In 2011, IANA distributed its last /8 blocks to the RIRs. After that, RIRs had only what they had on hand.

Then RIR by RIR:

APNIC — exhausted in 2011.
RIPE NCC — exhausted in 2012 (with reserves rationed).
LACNIC — exhausted in 2014.
ARIN — exhausted in 2015.
AFRINIC — exhausted around 2017–2019, with extra contention.

By the late 2010s, every RIR had effectively no allocations left. New ISPs and large enterprises couldn't simply request a /16 of fresh IPv4 addresses anymore.

But — paradoxically — the internet kept growing. Here's why.

Adaptation 1: NAT and CGNAT

NAT (Network Address Translation) lets many devices share one public IP. Originally a workaround at the home/office level — your router gives every device on your network a private IP, and translates them all to one public IP outbound.

That alone bought decades of growth: instead of 1 IP per device, 1 IP per household.

When even that wasn't enough, CGNAT (Carrier-Grade NAT) extended NAT to the carrier level. Many subscribers share one public IP at the ISP level. This is why mobile carriers can serve millions of users on a small pool of IPv4. It's why your "public IP" sometimes appears in the 100.64.0.0/10 reserved range.

CGNAT is the reason IPv4 exhaustion didn't translate into "no more new internet customers." But it has costs: port forwarding breaks, peer-to-peer apps suffer, and many subscribers share IP reputation, leading to mistaken bans and CAPTCHAs.

Adaptation 2: The IPv4 market

The unexpected outcome of exhaustion: a real, functioning market.

When RIR allocations stopped, organizations holding unused IPv4 blocks realized they had value. A company that received a /16 (65,536 addresses) in 1995 for $25 might have only used a tiny fraction. They could now sell the rest.

The market emerged in the 2010s. Per-IP prices:

2014: ~$10 per IPv4 address
2017: ~$15
2020: ~$25
2022: ~$50
2024: ~$45 (slight cooling)
2026: ~$40 (continued slow decline as IPv6 grows)

A /16 (65,536 addresses) sells for around $2.5 million in 2026 — down from peak but still substantial. Brokers (IPv4.Global, Hilco Streambank, others) facilitate transactions; RIRs handle the registration transfer.

This market is the main way anyone gets new IPv4 in 2026. Cloud providers, large hosting companies, and specialty buyers participate. Small businesses generally don't — they rent IPs from cloud or hosting providers instead of buying.

Adaptation 3: IPv6

The "real" answer to IPv4 exhaustion is IPv6 — 128-bit addresses, effectively unlimited supply. Designed in 1998, deployed slowly since the early 2000s.

IPv6 adoption in 2026:

~50% of Google's traffic is IPv6.
Mobile carriers lead — most major mobile networks are IPv6-first with IPv4 fallback.
Major content (Google, Facebook, Cloudflare, Akamai) is dual-stack.
Home ISPs vary — some are dual-stack, others are still IPv4-only.
Enterprise networks lag — corporate networks are notoriously slow to adopt IPv6 because of legacy gear.

Adoption grows steadily by a few percentage points per year. Full IPv6-only is decades away (if it ever happens — IPv4 may stick around as a long-tail forever, like fax machines).

Why we'll never just "switch to IPv6"

Two reasons:

They're not directly compatible. IPv4 and IPv6 traffic can't talk without a translator. So IPv6-only servers are unreachable to IPv4-only clients (and vice versa). Every public service needs to support both for the foreseeable future.
Legacy is real. Old hardware, old software, old configurations don't all support IPv6. Migrating an enterprise network is years of planning. Migrating the entire global internet is generational.

Result: dual-stack is the steady state for the next 10–20 years at minimum. IPv4 keeps working, IPv6 grows, and the two coexist.

What this means for users

Day to day: nothing changes. Your devices use whichever protocol works for the connection. Most users have both available without knowing it.

Specific impacts:

You may be behind CGNAT. Self-hosting, port forwarding, and certain peer-to-peer apps may not work. Workaround: Cloudflare Tunnel.
Static public IPv4 costs more than it used to. ISPs charge $5–25/month or more.
IPv6 is automatic when available. Modern OSes and apps prefer IPv6 if both work. You're using it without configuring anything.
Some sites still don't support IPv6. Mostly older or smaller sites. Browsers transparently fall back to IPv4.

Quick FAQ

Are we really "out" of IPv4? Out of new RIR allocations, yes. But the secondary market is active and most ISPs maintain pools. There's no "no more new IPs" cliff in 2026; it's a slow squeeze.

Will my home IP run out? No. Your ISP maintains a pool of IPs and rotates among customers. Even as the global supply tightens, your specific IP renewals don't fail.

Why don't we use private IPs more aggressively? Inside organizations and homes, we already do (192.168.x.x, 10.x.x.x, 172.16.0.0/12). The shortage is at the public layer where global addressability matters.

Will IPv4 ever be deprecated? Probably not in the way "deprecated" usually means. It'll keep being used as long as legacy systems need it. The internet may eventually be IPv6-primary with IPv4 as a backwards-compat layer, like ASCII still works alongside Unicode.

What about new TLDs and IP-related innovations? TLDs (like .io, .dev) are about domain names, not IP addresses. They don't help with IPv4 exhaustion. Innovations to expand the IP space at scale would essentially be IPv6 — which we already have.

TL;DR

IPv4 has 4.3 billion addresses. Demand exceeded supply by the 2010s.
We adapted by NAT, CGNAT, an active IPv4 market (~$40/IP in 2026), and slow IPv6 deployment.
The internet kept growing without a hard cliff because of these adaptations.
IPv6 is steadily taking over for new infrastructure but won't fully replace IPv4 for decades.
For users: mostly invisible, with occasional pain points around hosting, port forwarding, and static IP costs.

A protocol designed in 1981 is still the foundation of internet addressing in 2026. The fact that it's still working — through clever workarounds and a market it was never designed to need — is one of the quiet engineering wins of our era.