IPv4 Exhaustion and the IP Address Marketplace

The official story is that IPv4 ran out in 2011. The practical story is that companies still buy and sell IPv4 blocks every day in 2026, prices have climbed to $40-60 per address, and a small but liquid marketplace exists for the IPv4 inventory that nobody is using.

This post explains how IPv4 exhaustion actually played out, why we haven’t fully moved to IPv6 twenty years later, who buys and sells IP blocks, and what this means for anyone running infrastructure.

What “Exhaustion” Actually Means

IANA — the global authority — handed out its last /8 IPv4 blocks to the five RIRs in February 2011. Each RIR then started handing out its remaining inventory to local registrars and ISPs:

• APNIC (Asia-Pacific) exhausted its general pool in 2011.
• RIPE NCC (Europe / Middle East / Central Asia) in 2012, with a small reserve.
• LACNIC (Latin America) in 2014.
• ARIN (North America) in 2015.
• AFRINIC (Africa) in 2017, with reserves still being slowly distributed.

“Exhaustion” doesn’t mean every IPv4 address is in use. It means the RIRs no longer have unallocated blocks to give. The addresses themselves still exist; they’re just all currently held by some organization.

Why the Internet Didn’t Break

A few mechanisms kept the internet running despite no new IPv4 inventory:

CGNAT

Mobile carriers and ISPs put thousands of users behind a single public IP via carrier-grade NAT. This compresses the IPv4 needs of new customers from 1:1 to 1:N.

IPv6 deployment

Slowly but steadily, IPv6 became practical. By 2026, ~50% of global traffic carries IPv6 in parallel with IPv4 (dual-stack). New deployments — especially mobile — can be IPv6-only.

Reclamation

Many large /8 allocations from the early 1990s sat mostly unused. Companies with too many IPs started selling them to companies that needed them. The IP marketplace was born.

Better utilization

Cloud providers became dense. A 2010 server farm used 10,000 IPs for 10,000 servers. A 2026 cloud uses 100 IPs for 10,000 containers (via internal NAT and load balancers).

The combined effect: IPv4 exhaustion is a slow squeeze, not a sudden break. The cost of an IPv4 address has gone up; access has not been cut off.

The IPv4 Marketplace

The marketplace operates in three layers:

RIR transfers

The RIRs maintain transfer policies. An organization with surplus IPs can sell them to another organization; the transfer is registered with the RIR. This makes the transfer official — the new owner becomes the WHOIS-listed entity, BGP announcements can be made, etc.

Brokers

A few specialized firms (IPv4 Market Group, Hilco Streambank, Avenue4) broker transactions between buyers and sellers. They handle due diligence (verifying the seller can legally transfer, the IPs aren’t on blocklists, etc.) for a percentage cut.

Leasing

Some organizations lease IPs rather than sell. You pay monthly to use the block; the underlying registration stays with the lessor. Common for short-term needs or for organizations that want IP flexibility without long-term commitment.

Pricing

The market price for IPv4 has climbed steadily:

• 2015: ~$10 per address
• 2018: ~$15 per address
• 2021: ~$25 per address (post-COVID demand surge)
• 2023: ~$40-50 per address
• 2026: ~$40-60 per address (stabilized)

A /24 (256 addresses) sells for ~$10,000-15,000. A /22 (1024 addresses) for ~$40,000-60,000. Block size matters — bigger contiguous blocks command a premium because routing them is easier.

The leveling-off in 2024-2026 is partly because IPv6 deployment has finally absorbed enough demand that the IPv4 market isn’t growing as fast.

Who Buys IPv4

Major buyers in 2026:

• Hyperscalers (AWS, Azure, GCP, Oracle Cloud) — Constantly need more IPs for customer-facing services. Cloud customers want IPv4 addresses; hyperscalers buy them in bulk.
• VPN and proxy providers — Need diverse, dispersed IPs not associated with hosting blocks. Pay premium for “clean” residential-looking IPs.
• CDNs — Less than you might think (anycast reuses one IP across many POPs), but still meaningful demand.
• Hosting providers — Smaller players servicing customers who need IPv4 specifically.
• Enterprise — Companies migrating workloads to cloud that need stable IPs for partners’ allowlists.

Who Sells IPv4

Major sellers:

• Universities and early internet companies — Got huge /8s in the early 1990s for free, never used them all. Selling now is essentially free money.
• MIT famously sold half its /8 (16 million IPs) to Amazon in 2017.
• The DoD has been releasing some of its long-dormant allocations.
• Bankrupt or shrinking companies — Liquidating IP holdings as part of asset sales.
• ISPs that consolidated — Mergers and acquisitions leave duplicate allocations.

Geographic Aspects

IPv4 transfers aren’t uniformly easy across regions. Each RIR has different transfer policies:

• ARIN requires the buyer to demonstrate need.
• RIPE has more open transfers but stricter rules on cross-region transfers.
• APNIC has the most liquid market in absolute terms.
• AFRINIC has had governance disputes that complicated transfers.

Cross-region transfers happen but require coordinated RIR approval. Moving a block from ARIN to APNIC is more work than within either.

What This Means for Application Developers

For most application developers, the IPv4 marketplace is invisible. Your cloud provider deals with this. You request a load balancer; the cloud provider allocates an IPv4 from its (paid-for) pool.

Where it matters:

Cloud bills include IPv4 cost

AWS started charging $0.005/hour ($3.65/month) per public IPv4 in 2024. Other clouds followed. If your architecture uses many public IPv4 addresses for stateful services, this is a real line item — and it’s been a meaningful nudge toward IPv6.

IPv6 deployment becomes economic

Pre-2024, “deploy IPv6” was a “nice to have” tech debt item. Post-2024, the cost of not deploying IPv6 is now in your bill every month. For high-density deployments (containers, ephemeral workloads), this becomes meaningful.

IPv4-only services are increasingly fragile

Mobile-first markets (India, parts of Africa, Southeast Asia) are increasingly served by IPv6-only mobile carriers with NAT64 gateways for IPv4 access. An IPv4-only service works, but you’re seeing a meaningful fraction of users through translation layers that can occasionally fail.

Geolocation gets weirder

IP blocks transfer hands. A block that was Argentinian in 2018 might be owned by a US company in 2026 — but geolocation databases might lag. Reassigned IP blocks are one of the consistent sources of geolocation inaccuracy.

Why We Don’t Just Use IPv6

A reasonable question: with IPv4 expensive and IPv6 free, why isn’t everyone IPv6?

The honest reasons:

• IPv4 still works. Inertia is powerful. Existing systems don’t need to be touched.
• Operational complexity. Dual-stack deployments need monitoring, firewalls, and logging set up for both protocols.
• Some software is still IPv4-only. Legacy enterprise software, certain network appliances.
• Skills gap. Many network engineers were trained primarily on IPv4. IPv6 has different conventions.
• Customer requirements. Some customers demand IPv4 endpoints because their tooling assumes them.

The transition is happening, but it’s been slower than predicted at every point since 1995. See IPv4 vs IPv6: the great transition for the broader context.

Practical Advice in 2026

If you’re starting a new service:

• Default to dual-stack. IPv4 + IPv6 from day one. Most cloud platforms make this easy.
• Don’t design for IPv4 scarcity. Use load balancers and internal networking; don’t allocate public IPs per pod.
• Audit your IPv4 footprint quarterly. Reclaim unused IPs; consolidate where possible.
• Watch your cloud bill for public IPv4 charges. They’re a leading indicator of waste.

If you’re operating a large existing service:

• Inventory your IPv4 holdings. You may have unused IPs you can lease or sell.
• Evaluate IPv6 deployment as a cost reduction project, not just modernization.
• Talk to a broker if you have surplus /22 or larger blocks. The current market is liquid.

The Long Term

IPv4 won’t go away in 2026, 2030, or arguably ever — there’s too much installed base. But its role is shrinking:

• More mobile carriers are IPv6-first internally, with IPv4 as a gateway service.
• New cloud regions launch with IPv6-priority addressing.
• Some IPv4 blocks will eventually become economic-only assets (companies hold them as inventory).
• IPv6 keeps growing. Adoption was ~40% globally in 2024, ~50% by 2026 estimates.

A reasonable prediction: by 2035, IPv4 will be a legacy compatibility layer like IPX or AppleTalk — still around, but only relevant for older systems.

TL;DR

• IPv4 ran out at the RIR level around 2011-2017. Addresses themselves still exist, just all currently allocated.
• A liquid IPv4 marketplace trades blocks between organizations. Prices ~$40-60 per address in 2026.
• Major buyers: hyperscalers, VPN providers, hosting. Major sellers: universities, shrinking companies, governments.
• AWS started charging for IPv4 in 2024. Other clouds followed. The economics finally favor IPv6 deployment.
• IPv4 reassignments cause geolocation drift — a US-registered block might have been Brazilian a few years ago.
• For new services in 2026: default to dual-stack.
• For existing services: audit IPv4 footprint; consider IPv6 deployment for cost reduction.

The IPv4 marketplace is one of those niche industries that exists because of internet architectural constraints. Most people will never deal with it directly. But the consequences — higher IPv4 costs, slow IPv6 transition, geolocation drift from reassignments — show up everywhere. For the broader IP picture, see everything you need to know about IP addresses and the Ip2Geo API for staying current on what each IP actually represents in 2026.