Summary

  • ARIN implemented its first specified-recipient transfer policy on 1 June 2009, more than six years before its free pool reached zero. By October 2009, two transfers had completed and another was pending. The institutional route for market exchange therefore existed well before formal exhaustion.
  • ARIN's optional Specified Transfer Listing Service, available from August 2010, openly connected organizations seeking addresses, holders willing to release them and facilitators. It reduced search costs but did not publish transaction prices or turn registry approval into commercial settlement.
  • The 2011 Nortel-Microsoft transaction made the economic reality visible. Nortel's court record identified 666,624 IPv4 numbers sold for USD 7.5 million, while ARIN and Microsoft separately agreed to place transferred resources under a registration services agreement. The sale and the registry update were connected but distinct events.
  • Completed ARIN specified-recipient transfers rose from 24 in 2011 to 42 in 2014 and 105 by the end of September 2015. Inter-RIR transfers with APNIC had also reached 83 completions by then. These figures show a functioning and accelerating market before the exhaustion announcement.
  • Formal exhaustion changed the outside option. Organizations could no longer compare a priced transfer with a plausible free-pool allocation of the required size. ARIN's own announcement directed approved applicants to the waiting list or the transfer market and removed an exhaustion-linked restriction on how often a source could seek specified-recipient transfers.
  • Policy lagged operating demand because allocation-era needs tests, regional compatibility limits and incomplete public statistics were carried into private transactions. The market supplied price discovery, search and risk allocation; ARIN supplied identity checks and authoritative record changes. Confusing those roles made a registry decision look like the whole transaction and concealed delays, failed deals, prices and operational consequences.

The date on the calendar was an inventory event, not a market origin

On 24 September 2015, ARIN announced that it had issued the final IPv4 addresses in its free pool. The sentence was exact and important. A particular inventory held for ordinary allocation had reached zero. Yet the date has acquired a larger meaning than the event can bear. It is often treated as the day IPv4 became scarce in North America, the day transfers became necessary or the day addresses became economic assets.

None of those propositions fits the chronology. Buyers and sellers had already met. Transfer requests had already been approved. A registry-supported listing service had already operated for five years. A bankruptcy estate had already sold a large block for a disclosed price. Brokers had begun assembling holders, buyers, diligence and settlement. Address demand had already crossed ARIN's regional boundary toward APNIC. The market did not wait for ARIN's inventory to print zero.

Exhaustion did something narrower and more powerful. It ended the expectation that an applicant able to satisfy ordinary criteria might receive the required block from the free pool at a service charge rather than a market price. Before exhaustion, a buyer could compare a transfer with waiting, resizing, improving utilization or returning to ARIN. After exhaustion, the waiting list depended on returns, revocations and later IANA distributions. For an operator that needed a predictable quantity on a predictable date, the transfer market ceased to be an unusual alternative and became the main acquisition channel.

This distinction matters because an inventory event and a price-forming market answer different questions. ARIN could announce how many unissued addresses remained. It could not announce the full demand for addresses, because some operators leased, used upstream space, deployed address sharing, acquired companies, bought transfers or never submitted a request they expected to fail. The free-pool count was visible. The economic shortage had already spread through decisions outside that count.

The useful history therefore begins before September 2015. It asks when policy first recognized a seller with unneeded space and a named recipient with demand; when commercial parties could find one another; when a court treated address rights as valuable sale consideration; when transfers crossed regions; and when request volume revealed that operators were no longer planning around ordinary allocation. Those are the events that formed the market.

Demand split from the free pool before the pool was empty

A free pool can coexist with scarcity. The relevant question is not whether any addresses remain, but whether the pool can deliver the size, timing, location and certainty an operator requires. Fragmentation appears before the last unit disappears. Large blocks become difficult to match. Requests face deeper review. Forecast horizons shorten. An applicant that expects a data-centre opening, mobile deployment or cloud service cannot treat an uncertain future allocation as equivalent to a block available under contract.

ARIN's final-year record shows this separation clearly. A 2016 ARIN presentation recorded 414 IPv4 requests in June 2015, the peak of the final rush. The first unmet request arrived on 1 July because the organization qualified for a block size that was no longer available. Within weeks, only individual /24 blocks remained in the general pool. Full depletion came nearly three months later. For that first unmet applicant, functional exhaustion had already occurred.

The same logic operated years earlier for parties seeking a block that ARIN could not or would not supply on their commercial schedule. A company could have genuine operating demand and still prefer a transfer because acquisition had to align with a merger, a deployment window or the availability of a particular clean and aggregatable block. Another company could control legacy space no longer central to its business and face a choice between leaving it idle, returning it without sale proceeds or finding a buyer.

That pair of incentives is enough to create exchange even while a registry has residual inventory. The free pool suppresses the market price because some buyers retain a lower-cost alternative. It does not eliminate the market where timing, block size or qualification differ. A market can be thin, negotiated and administratively constrained and still be a market.

The policy debate that began in 2008 was itself evidence that operators and ARIN understood this. ARIN's assessment of a proposed IPv4 transfer policy described an organization with excess space transferring it to an organization needing more, together with a listing service intended to bring the parties together. The proposal was framed around anticipated depletion, speculation and documentation. The economic actors were present in the design before the final inventory disappeared.

The 2009 policy created a lawful corridor, not the underlying incentive

ARIN adopted Transfer Policy 2009-1 on 28 May 2009 and implemented it on 1 June. The rule allowed transfers to specified recipients within the ARIN region under stated conditions. At the October 2009 public policy meeting, ARIN reported two completed transfers and one pending. That is a modest beginning, but it fixes the sequence: recorded specified-recipient transfers began more than six years before exhaustion.

Policy did not manufacture the economic incentive. It created a recognized corridor through which the registry would update records after a source and recipient satisfied its conditions. The distinction is easy to lose because a transfer that ARIN refused to recognize would carry lower operational and commercial value. Accurate registration, reverse DNS, contact authority and later routing-security services make recognition consequential. But consequence is not origin. Holders wanted value for surplus capacity, and operators wanted usable addresses, whether or not the policy manual had caught up.

The first rule also shows how policy remained attached to allocation logic. The recipient had to demonstrate need. The source faced eligibility and timing constraints. The recipient was initially limited to the ARIN region. These conditions were intended to deter speculation, preserve conservation principles and keep registry records accurate. They also meant that a buyer and seller could agree commercially without knowing whether the authoritative record would change.

That uncertainty placed ARIN inside the transaction as a condition to completion. It did not make ARIN the seller, buyer, price setter, broker, escrow agent or operator. The registry reviewed identity and eligibility, accepted an agreement and changed records. Commercial parties still had to identify the block, value it, warrant authority, investigate disputes and reputation, coordinate payment, plan routing and handle tax and accounting.

This division explains why policy could lag demand even when it arrived before exhaustion. The policy recognized a transfer but viewed it through the same administrative concepts used to ration unissued space. The market was moving from allocation to exchange. The rule treated exchange partly as a special allocation in which a source named the next recipient.

A matching service acknowledged that search had become a real cost

In August 2010, ARIN began offering the Specified Transfer Listing Service. The service connected organizations seeking IPv4 space, organizations offering it and facilitators interested in helping a transaction. ARIN later described it as an optional service for in-region and inter-RIR specified-recipient transfers. Its existence is one of the clearest pieces of evidence that a market architecture preceded exhaustion.

Search is not a minor inconvenience in a thin asset market. A holder's public registration does not show willingness to sell. An apparently unused block may support private infrastructure, future expansion, disaster recovery or a corporate option. A buyer does not simply need addresses; it may need a particular minimum size, registry status, route history, geolocation profile and delivery date. Publicly approaching every holder can reveal strategy and attract fraud.

The listing service lowered that first barrier. It allowed parties to signal supply or demand without pretending that ARIN had created an exchange with a common price and binding bids. Facilitators could add market knowledge, identify likely counterparties and guide parties through registry requirements. The service gave institutional visibility to a role that would later expand into specialist brokering.

Its limits are just as revealing. A listing did not establish title, clean reputation, routeability or a final price. It did not secure funds, allocate tax risk or guarantee approval. ARIN's later retirement of the service in 2023 did not end the market because private search institutions had matured. The service was important in 2010 precisely because the private market was still sparse and the registry knew counterparties needed a meeting place.

The public record does not disclose how many potential sellers declined to list, how many matches failed, how asking prices differed from settlement or how long a party waited. Those absences prevent a complete welfare assessment. They do not erase the institutional fact. By offering to connect supply, demand and facilitators five years before zero inventory, ARIN acknowledged that allocation alone no longer described address acquisition.

Nortel made the price visible without settling the legal theory

The Nortel transaction in 2011 turned an obscure transfer mechanism into a visible capital event. A Nortel bankruptcy disclosure filed in the United States records that Nortel Networks Inc. and Microsoft entered an asset sale agreement on 16 March 2011 for 666,624 Internet protocol numbers at a price of USD 7.5 million. The bankruptcy court approved the sale on 26 April, and the transaction closed on 11 May. The arithmetic was roughly USD 11.25 per address.

ARIN's own announcement addressed a separate but connected matter. On 15 April, ARIN said it and Microsoft had agreed that certain resources moving from the Nortel bankruptcy would be placed under a registration services agreement. ARIN described the specified-transfer policy as allowing market incentives to improve utilization. The statement showed an accommodation between a court-supervised sale and registry recognition.

The transaction should not be simplified into a sweeping judicial declaration that every IPv4 address is property in every context. Bankruptcy approval, contractual rights, registry policy and operational use are not interchangeable legal categories. The court record does, however, establish a paid sale of identified IPv4 numbers from an estate, and the ARIN record establishes a negotiated route for recognizing the resulting registration change. Those are powerful facts without embellishment.

The price was especially important because transfer logs generally omitted consideration. One disclosed sale supplied a public comparable for holders, buyers, advisers and later analysts. It made clear that a large legacy block could produce material estate value. It also showed that registry vocabulary could not prevent commercial parties and a court from assigning economic value to the ability to control and use addresses.

The event revealed the layered nature of the market. The estate sold. The court approved the estate transaction. Microsoft contracted and paid. ARIN evaluated the registration consequences and obtained an agreement. Operators then had to make addresses usable in networks. No single actor performed all those functions. The architecture emerged from their intersection.

A transfer is a contract, a registry change and an operating change

Much confusion about 2015 comes from describing a transfer as if it were one instantaneous act. In practice, at least three events must be separated.

The first is the commercial event. Parties agree on the block, price, warranties, conditions, payment, confidentiality, liability and remedies. A broker or adviser may assist. A court, lender or corporate board may need to approve. This event can exist before ARIN is ready to update any record.

The second is the registration event. ARIN verifies that the source is the registered holder or has an acceptable chain of authority, assesses the recipient under then-current policy, obtains agreements and fees, and changes the authoritative registration. This is the event most visible in a transfer log. It is indispensable to market confidence, but it does not disclose the whole bargain.

The third is the operating event. Routes change or remain with the same origin under a new commercial arrangement. Reverse DNS is moved. Route objects and later ROAs are updated. Geolocation and reputation services learn slowly. Customer allowlists, mail systems, cloud platforms and security tools may continue to remember the previous user. A block can be registered to a buyer yet not be fully usable on the planned date.

These events can occur in different orders and can fail separately. Funds can be committed before registry approval. A registration can change before routes are ready. A company acquisition can transfer control while public records remain stale. A block can route through a service provider without a sale. That is why BGP changes alone produce false positives when researchers try to infer market transfers.

ARIN's 2015 exhaustion announcement concerned only the supply side of one part of this system: ordinary unissued inventory. It did not change the fact that commercial, registration and operating events had already been coordinated for years. What it changed was the relative bargaining power of the alternatives.

The transfer counts were rising before the announcement

ARIN's October 2015 members meeting supplied a concise pre-exhaustion series. The registry reported 24 completed specified-recipient transfers in 2011, 23 in 2012, 24 in 2013 and 42 in 2014. By the end of September 2015, the count had reached 105. The annual figure had already more than doubled the previous year's total even though exhaustion occurred only six days before the September cutoff.

The sequence does not show a market appearing overnight. It shows a thin channel operating at roughly two completions per month for several years, nearly doubling in 2014 and accelerating sharply during the final run on the pool. Some increase likely reflects improved familiarity, broker capacity and better registry processes. Some reflects declining free-pool certainty. Some reflects demand in other regions. The public totals do not allow each cause to be isolated.

They do establish direction. Parties were not waiting for a ceremonial declaration. They were closing transactions while ARIN still had addresses to issue. The market thickened as the probability of receiving the desired block through ordinary allocation fell.

Even these counts understate economic activity. A completed transfer is the survivor of search, diligence, negotiation and registry review. It excludes inquiries that never became tickets, tickets that closed without approval, buyers that turned to leases or upstream addresses, and sellers that withdrew. ARIN's current statistics warn that a processed ticket means a ticket was closed regardless of final status. A request count is therefore not a completion count, while a completion count is not a demand count.

The missing denominator is central to policy evaluation. A rise from 42 to 105 completions may represent a more liquid market, a larger shortage, improved approval, more applications or all four. Without submitted, withdrawn, rejected and timed-out cases linked to block size, public reporting cannot identify how much demand policy filtered out.

APNIC demand globalized ARIN supply before ARIN ran out

Inter-RIR transfers widened the market's geography. ARIN states that its inter-RIR policy became effective on 31 July 2012. By October 2015, ARIN reported 83 completed transfers between its region and APNIC, with almost all early movement going from ARIN toward the Asia Pacific. At that meeting, only two had come into the ARIN region, and the first ARIN-to-RIPE NCC request had recently arrived.

This direction is economically coherent. APNIC had entered its final /8 regime in April 2011. An APNIC account could receive only a small ration from 103/8, while fast-growing networks could require substantially more. ARIN still contained large legacy holdings and residual free-pool supply. Policy compatibility opened a route from one regional stock to another region's demand.

The cross-regional trade undermines a simple exhaustion narrative. ARIN addresses were leaving the region before ARIN's own pool hit zero because scarcity was not synchronized across administrative territories. A seller's location, a buyer's registry account and the network's users did not fit one regional inventory clock.

It also exposed policy as a trade barrier or trade enabler. A technically usable block could move only when both registries maintained compatible rules and coordinated records. If one side required a needs test, imposed a size rule or refused the counterpart's policy, the deal could fail despite willing parties. Regional policy differences therefore affected liquidity and price without changing the 32-bit protocol.

ARIN's free-pool date was regional; the market was already global. The relevant supply for an APNIC buyer included willing ARIN holders. The relevant demand for an ARIN seller included networks outside North America. Once that became true, a registry-by-registry exhaustion calendar could no longer describe the full economy.

The final rush exposed a timing market before the quantity reached zero

June and July 2015 are more informative than the September headline. June's 414 requests showed applicants attempting to secure the remaining option. On 1 July, ARIN started the waiting list for unmet requests after an organization qualified for a size the pool could not provide. In September, an IANA distribution temporarily satisfied 13 waiting organizations, but the general free pool still reached zero on the 24th.

For an operator, this was a market in time as much as in addresses. An allocation request offered a low acquisition price but uncertain block availability and delivery. A transfer offered a negotiated price but potentially greater control over size and closing. A waiting-list position offered an option on future returns or distributions but no guarantee that timing would match a deployment.

The final rush also rewarded administrative readiness. Organizations with documented utilization, forecasts and current records could submit quickly. Those without staff to assemble evidence faced a higher risk of missing the remaining blocks. The queue did not simply rank underlying social value. It ranked requests that reached ARIN in an acceptable form while matching the fragments then available.

Market acquisition shifted some of that contest from chronological access to purchasing capacity. It solved timing for buyers able to pay and created a return for holders willing to sell. It also imposed a new entry price on networks without inherited space. Neither mechanism is distributionally neutral.

The important conclusion is not that a market is inherently fairer than a queue. It is that the queue and the market were already substituting for one another before official exhaustion. Policy debate that treated transfers as a post-zero exception missed the period in which operator planning had already moved between both systems.

Exhaustion removed one restriction and strengthened another kind of control

ARIN's zero-pool announcement did more than mark inventory. It stated that approved requests could be fulfilled through the waiting list or the IPv4 transfer market. It also explained an immediate policy effect: source organizations were no longer limited by an exhaustion-linked restriction on how often they could request specified-recipient transfers.

That change increased potential market supply. Before exhaustion, a source that had transferred space could face a period of ineligibility for new allocations or assignments, designed to prevent an organization from selling and returning to the pool. Once the pool was gone, the free-pool side of that anti-cycling rule lost its practical basis.

At the same time, ARIN's role in recipient qualification became more economically important. When free allocations existed, a refusal could send an applicant back to planning or another request. After exhaustion, registry approval stood between a privately negotiated purchase and an authoritative record change. The same needs doctrine now operated over capital supplied by a seller rather than inventory supplied by ARIN.

That was the central institutional transition. ARIN moved from deciding how to distribute its own residual pool to deciding whether parties could complete recognized exchanges of privately negotiated value. The technical need for identity verification and accurate records remained. The justification for carrying every allocation-era demand judgment into a paid transfer became harder to state.

Policy discussions immediately reflected the tension. In late 2015 and 2016, proposals sought to simplify or remove needs evaluation, particularly for smaller or post-depletion transfers. ARIN's comparison material said existing requirements asked recipients to document current utilization and a 24-month plan, including detailed network information. The market could agree a price before the registry completed that test.

Exhaustion therefore did not make ARIN irrelevant. It shifted the basis of ARIN's power. Inventory power declined to zero; approval, record and compatibility power became more valuable.

Why allocation-era need did not map cleanly onto a paid transaction

A needs test has an understandable purpose when an administrator is distributing a common pool below market price. Every approved request consumes inventory unavailable to the next applicant. Requiring evidence can deter waste and stretch the pool. The administrator is rationing its own stock.

In a specified-recipient transfer, the source has selected a buyer and the parties have allocated the purchase price. ARIN is not choosing between two applicants for the same free block. It is verifying a change from one holder to another. Applying a forward-looking need test in that setting does not conserve the depleted pool. It controls which buyer may deploy capital and how much it may acquire.

There can still be legitimate concerns. Fraudulent recipients, disputed authority and incomplete records threaten the registry. A buyer that immediately resells could contribute to manipulation in a thin market. A huge dormant purchase might reduce liquidity. But those risks require evidence and tailored controls. They do not automatically justify treating every private acquisition as a free allocation.

The test also creates forecasting problems. Operators acquire addresses for uncertain growth, contingency, customer contracts and platform launches. Requiring a fixed 24-month subnet plan may favor established firms with historical demand over entrants whose customer base depends on securing addresses first. It can encourage applicants to present their business in the categories the rule recognizes rather than the way demand actually develops.

This is one reason policy lagged operating demand. The market priced optionality, timing and risk. The registry asked for a deterministic need narrative. A company might rationally buy more than immediate utilization because future transfer costs, clean-block scarcity or deployment risk made reserve capacity valuable. Allocation doctrine could classify the same decision as insufficiently justified.

The mismatch did not mean ARIN should approve every claimed transaction. It meant the approval question needed to narrow: is the source authorized, is the recipient real, is the block undisputed, will the record remain unique and accurate, and are technical services ready to change? Those questions protect the ledger without pretending to allocate the buyer's capital.

Brokers formed because neither the registry nor a public log supplied the market

The rise of brokers was not an accidental layer of complexity. It answered information problems that ARIN's policy and statistics left open. A buyer needed a willing seller, a suitable block, a credible chain of control, a plausible closing timetable and a price. A seller needed proof that a buyer could fund, qualify and complete. Neither could learn those facts from a registration record alone.

ARIN's listing service provided introductions, but private facilitators could search beyond listed inventory, compare blocks, coordinate pre-approval, help repair legacy records and sequence commercial conditions with registry action. Their repeat experience reduced the cost of interpreting an evolving policy. That experience became valuable before exhaustion and more valuable after it.

Intermediation also created risks. A broker might represent both sides, control information about comparable prices or overstate its influence with the registry. Fees could be hidden in a spread. A seller might mistake facilitator registration for ARIN endorsement. A buyer might rely on a block description without independent route and reputation checks.

Those risks are characteristic of an opaque market, not proof that the market did not exist. Indeed, they show why it did. Where public price, failure and timing data are absent, repeat intermediaries accumulate private knowledge. That knowledge improves execution while concentrating bargaining advantage.

ARIN's role should be legible here. It can identify facilitators that meet program conditions and publish neutral process guidance. It cannot certify a broker's price, loyalty or commercial warranties merely because the broker understands ARIN procedure. Registry recognition of the address change and market trust in the intermediary must remain separate.

The assets were not homogeneous, even when the count was the same

One IPv4 address is numerically interchangeable with another at the protocol layer, but blocks are not economically identical. A contiguous /16 and 256 separate /24s contain the same number of addresses yet differ in routing burden, buyer depth and operational convenience. A block with stable history differs from one associated with spam, fraud or geolocation errors. Legacy registration can require a different authority file from space already under a standard agreement.

This heterogeneity helped the market form before aggregate exhaustion. A buyer might prefer a particular block even when ARIN could offer some addresses because the market block fit routing, geography or deployment. A seller with a large clean legacy range could supply a class of demand the fragmented pool no longer served.

It also means that a single price per address can mislead. Nortel's disclosed USD 11.25 became a useful reference, but size, timing, legal context and quality affected later prices. Large blocks can command a different unit price from small ones. Distressed sales can differ from planned portfolio transactions. A registry transfer log that omits consideration cannot reveal these adjustments.

Market architecture therefore includes diligence about the afterlife of the block. Buyers need route-origin history, abuse reputation, geolocation behavior, reverse-DNS readiness, registration authority and any claims or liens. Sellers need to understand residual obligations and the timing of record removal. These are operating questions, not merely registry form fields.

The exhaustion announcement treated all remaining general-pool inventory as zero. The market treated every available block as a differentiated bundle of rights, records and operational memory. That difference is another reason zero day did not create exchange.

Public transfer logs showed movement but concealed market health

ARIN's statistics are valuable. They identify transfer types, counts and, in later reporting, /24 equivalents moving within and between regions. They make it possible to see direction and acceleration. They do not show price, bid-ask spread, time from agreement to approval, failed transactions, competing bids or post-closing remediation.

This creates a policy blind spot. An institution can say transfers are increasing without knowing whether small buyers face worse terms, whether approval delay produces discounts or whether a few intermediaries control most supply. A large quantity can reflect healthy reallocation or distressed concentration. A low quantity can reflect weak demand, incompatible rules, hidden leasing or administrative blockage.

The 2015 series is particularly instructive. Twenty-four, 23, 24, 42 and then 105 completed transfers demonstrate acceleration. They do not reveal how many addresses moved in each deal, what parties paid or how many requests failed. The 83 APNIC-linked transfers show cross-regional demand, but not the price difference that induced direction.

A minimal accountability report would preserve confidentiality while publishing quarterly aggregates by transfer type and block-size band: submitted, approved, rejected, withdrawn and pending cases; median and tail processing time controlled by ARIN; reasons grouped without identifying parties; addresses and prefixes completed; and voluntary, anonymized price ranges collected under a clear method. Broker or research price series could be compared without being treated as official valuation.

The aim is not price regulation. It is to prevent policy from operating on an incomplete picture of the market it affects. ARIN knew its free-pool inventory exactly. It knew far less publicly about the price and delay created when that inventory disappeared.

Policy lag was partly a problem of institutional vocabulary

ARIN's language evolved more slowly than the economic entity. Allocation policy assumed a resource requested from a common pool, justified by forecast use and issued under registry terms. Market exchange involved a holder, a buyer, consideration, diligence, closing conditions and the possibility of court or lender involvement. Both concerned the same numbers, but not the same decision.

When institutions lack a vocabulary for a new market, they stretch familiar categories. A sale becomes a release followed by an allocation. A buyer becomes an applicant. A seller becomes a source. Purchase quantity becomes demonstrated need. Registry recognition is described as the transfer itself.

Those terms can keep the records orderly. They can also obscure where economic authority lies. The price is not an ARIN fee. The block is not supplied from ARIN inventory. The downside of delay falls on the parties and their customers. The registry can prevent recognized completion while assuming none of the commercial risk.

The lag was not simply staff resistance. Policy development had to address speculation, legacy authority, regional compatibility, fraud and rapid depletion in public. Reasonable entities could disagree about how much market freedom would threaten conservation or fairness. The record shows active adaptation before and after exhaustion.

The failure was subtler: adaptation was measured against the allocation regime rather than the operating economy. The question remained how to permit exceptions to rationing when the more useful question was which registry controls remained necessary once rationing no longer supplied the addresses.

The market did not make the registry unnecessary

Recognizing policy lag does not imply that unrecorded trading is safe. A global address system depends on uniqueness and credible control records. A buyer needs assurance that the seller is authorized and that the same block will not be recognized for another party. Networks need contacts, reverse delegation and security assertions to follow the legitimate controller. Fraud becomes more attractive as prices rise.

ARIN therefore retained a necessary role after the free pool reached zero. The market could price scarcity and connect counterparties; it could not unilaterally produce a globally trusted registration change. Courts could approve estate sales; they did not operate the regional registry. Brokers could coordinate closings; they could not guarantee every relying network would accept the result.

The governance boundary lies between protecting the record and governing the business case. Identity, authority, uniqueness, dispute notation, authenticated change and transparent service standards belong inside the registry function. Choosing the buyer's growth model, deciding whether its reserve is economically virtuous or setting the acceptable return on a holder's sale do not follow from recordkeeping.

This narrow boundary is demanding. It requires excellent fraud control, current corporate evidence, consistent reasons, prompt correction and review. A thinner mandate is not a weaker service. It concentrates accountability on the acts only the registry can perform.

Exhaustion made this concentration urgent. Once ARIN stopped being the ordinary supplier, every extra condition on a transfer had to be justified as protection of the record or a clearly evidenced market safeguard, not inherited merely because it once governed free inventory.

September 2015 changed bargaining power more than technology

Nothing in the IPv4 packet header changed on exhaustion day. Routers did not reject addresses because ARIN's pool reached zero. Existing holders did not lose their blocks. Operators continued to use IPv4, deploy sharing, buy transit, lease capacity and add IPv6 according to their own economics.

What changed was bargaining power. Holders controlled a stock that could no longer be replenished through ordinary ARIN allocation. Buyers lost a low-price outside option. Brokers gained demand. Clean, transferable blocks gained strategic value. ARIN's approval and record services became more consequential to transactions while its allocation inventory disappeared.

The waiting list preserved a small alternative based on returned or recovered space, but not one equivalent to a scheduled market purchase. Its uncertainty created option value rather than guaranteed supply. An operator could remain on the list while arranging a transfer, but policy links between the two channels affected strategy.

The date also changed expectations. Before zero, a holder considering sale faced uncertainty about how long buyers could still obtain ARIN space. After zero, the broad direction was clear even though future prices were not. The market's information environment became less ambiguous.

That is why prices and transaction volume can respond to an administrative announcement even when the market predates it. The event changes beliefs and alternatives. It does not need to create the traded entity.

The counterfactual is a shadow market, not an absence of exchange

Suppose ARIN had refused to recognize specified-recipient transfers before exhaustion. Demand would not have vanished. Some organizations would have acquired companies mainly for their address holdings. Some would have leased or obtained provider space. Some would have used private contracts while leaving registration unchanged. Others would have deployed more address sharing or delayed projects.

Those substitutions would have produced less transparent control and more stale records. They would also have shifted value into corporate shells, service agreements and routing arrangements that a transfer log could not see. The absence of a recognized market would not prove the absence of trade; it would prove that trade had been pushed into less legible forms.

The 2009 rule was therefore useful. It gave legitimate exchange a route to accurate records. The problem was not that ARIN recognized transfers. It was that the policy boundary could remain wider than necessary and the public evidence too thin to measure its effects.

The same counterfactual explains the importance of brokers and the Nortel case. Economic actors assembled mechanisms around scarcity because operational demand persisted. Formal recognition reduced risk but followed the incentive. It did not create it.

The lesson of exhaustion is to measure the option that disappeared

An audit of ARIN's 2015 exhaustion should not begin and end with the number zero. It should measure the disappearing outside option. How many applicants could still obtain the size they required from the pool in each month? How long did qualifying requests wait? How many shifted to transfers, leasing, upstream space or corporate acquisition? What premium did certainty command before and after 24 September?

The public record supplies parts of the answer: June's request peak, the first unmet request on 1 July, transfer acceleration, 83 APNIC-linked completions, the 105 specified-recipient completions by September and the earlier market institutions. It does not supply a complete counterfactual price series or rejected-demand denominator.

That limitation should shape confidence, not invite invention. The chronology strongly supports the conclusion that the market existed and accelerated before exhaustion. It supports the claim that free-pool deterioration increased transfer demand. It cannot assign a precise percentage of each price or volume change to ARIN policy rather than global growth, APNIC scarcity, brokerage maturity or IPv6 deployment.

The missing evidence is itself a governance result. Institutions measured the stock they administered better than the market their rules increasingly governed. A post-exhaustion registry should reverse that imbalance by publishing service and decision data sufficient to test whether its controls protect the record at proportionate cost.

Exhaustion day revealed the market's prior construction

ARIN's 2015 announcement remains a genuine milestone. It marked the end of ordinary free-pool allocation in the region and told operators, plainly, that unmet approved requests would depend on a waiting list or the transfer market. It changed expectations, removed an exhaustion-linked source restriction and concentrated demand in paid channels.

But the market's foundations were already laid. The 2009 rule created a recognized corridor. The 2010 listing service lowered search costs. Nortel and Microsoft exposed a public price and a court-registry seam in 2011. Inter-RIR policy connected ARIN supply with APNIC demand in 2012. Brokers accumulated private execution knowledge. Transfer completions accelerated in 2014 and through the first nine months of 2015.

The institutional lesson is not that policy did nothing. Policy made recognized exchange safer and more valuable. The lesson is that policy followed, constrained and interpreted demand that originated in operating networks and asset holders. When it carried allocation-era judgments into paid transactions, it lagged the economy it was trying to govern.

The right post-exhaustion settlement is neither a return to free allocation nor an unrecorded free-for-all. It is a disciplined separation. Markets discover price, match risk and allocate capital. Parties and courts establish commercial authority under applicable law. Operators make addresses useful. ARIN authenticates control, preserves uniqueness, records changes and publishes enough decision data to be held accountable.

On that reading, 24 September 2015 was not the birth of the IPv4 market. It was the day the remaining administrative substitute disappeared and the market that had already been built became impossible to ignore.

Sources