AFRINIC's Remaining Pool and the Cost of Being Last

Summary

AFRINIC became the last of the five Regional Internet Registries to enter its final /8 on 31 March 2017. By then APNIC, RIPE NCC, LACNIC and ARIN had already restricted or exhausted ordinary IPv4 supply, giving AFRINIC's remaining stock a global option value that its regional rules were not originally designed to price.
Phase 1 permitted allocations as large as a /13, about 524,288 addresses, based on demonstrated need. AFRINIC's 2018 report says it allocated about 6.1 million addresses that year and ended with 0.40 of a /8. Its 2019 report records another 4,574,464 addresses and only 0.16 of a /8 remaining, including the /11 reserved for Phase 2.
Phase 2 began on 13 January 2020 when approval of a prefix left no more than one /11 of non-reserved space in the final /8. The minimum became /24 and the maximum /22, but there was no explicit limit on the number of additional requests; applicants still faced an eight-month planning test, utilization review and contractual compliance checks.
Cross-region attention was visible before the final /8. AFRINIC's own archive contains proposals about foreign membership, out-of-region use, inbound transfers, regional transfers and tighter soft-landing limits. Those debates do not prove that every large or externally connected applicant was abusive. They show that geography, corporate identity and end use became distributive questions once other regions lacked comparable free supply.
Review conflict intensified because a compliance finding could do more than correct records: recovery could return valuable addresses to the free pool. A resource-review proposal ran through eight versions between 2016 and 2019, contemplated random, periodic and whistleblower-triggered audits, recovery of non-compliant resources and appeals, and drew both strong support and strong opposition.
Being last did not automatically benefit African operators. A remaining pool offered lower cash acquisition cost, but uncertain review, regional-use restrictions, repeated small requests, delayed transfers and institutional instability could impose a scarcity premium through legal risk, financing cost, provider dependence and slower network expansion.
The proper registry response is narrow and auditable: verify authority, uniqueness, utilization under the adopted rule and accurate records; publish the demand funnel and decision reasons; separate applicant domicile from routing and customer geography; and prevent recovered scarcity value from expanding administrative discretion beyond objective technical and contractual requirements.

The last pool was an asset and a burden

When ARIN announced the end of its ordinary free pool in 2015, AFRINIC became exceptional. It was the only Regional Internet Registry still able to make substantial allocations without operating entirely through a final-ration regime, a waiting list or transfers. The distinction appeared favourable to Africa: operators in the region still had a path to addresses that networks elsewhere increasingly had to buy, lease or obtain in small one-time blocks.

That advantage was real. It was also unstable. A free-pool allocation can spare a network the cash cost of a market purchase and the dependency of provider-assigned space. Once the rest of the world lacks the same option, however, the remaining stock attracts attention. Every definition around access becomes valuable: who counts as a regional applicant, what use is sufficiently African, how need is measured, whether related companies count separately, when utilization is adequate and whether a prior allocation was compliant.

AFRINIC described the moment in an article titled Last RIR Standing. It records that APNIC entered its final /8 in 2011, RIPE NCC in 2012, LACNIC in 2014 and ARIN exhausted its free pool in September 2015. AFRINIC did not enter Phase 1 until 31 March 2017. The article also explains that Phase 1 allowed a maximum /13 and that, at the time, no explicit rule limited the number of requests, provided need could be justified.

The official phrase "last RIR standing" captured the inventory sequence but understated the institutional consequence. AFRINIC was not standing outside the global market. It was holding the last large marginal supply inside a world market. An address allocated in Mauritius, Kenya, Nigeria or South Africa could support customers and infrastructure whose economic relationships crossed borders. Corporate groups could operate in several regions. Cloud, hosting and content networks did not confine every packet or customer to the jurisdiction where the member was incorporated.

The remaining pool therefore made AFRINIC's regional boundary do work it had not previously needed to do at the same value. When addresses were broadly available at every registry, a disagreement about out-of-region use concerned policy consistency. When AFRINIC alone retained millions, the same disagreement determined access to scarce capital.

This did not mean the registry should open the pool without conditions. African operators had a legitimate interest in preserving access during transition. Fraud, duplicate claims and fictional need required controls. It meant the controls needed more objective evidence, more review and more public accounting as the value of a favourable decision increased.

Being last was therefore not a windfall. It was a stress test of whether a member association designed to register number resources could administer a scarce global input without turning every factual judgment into territorial economic policy.

The global calendar changed the value of an African allocation

AFRINIC did not create global IPv4 scarcity. IANA's ordinary pool ended in 2011, each RIR received a final /8, and regional communities chose different soft-landing rules. The resulting calendar nevertheless changed incentives inside Africa.

An operator in another region facing a one-time /22 or no free-pool access could value a large AFRINIC allocation far above its administrative fee. An African member serving multinational customers could expand using registry-issued space instead of buying at market price. An address holder could see growing value in unused or weakly used inventory. A consultant able to navigate needs documentation acquired more valuable expertise.

The change can be expressed without assuming misconduct. Suppose an applicant can receive a /13 in Phase 1. That prefix contains 524,288 addresses. Market evidence reviewed by APNIC indicates that reported IPv4 prices rose from about USD 10 per address in 2016 toward USD 20 by 2019 and stood near USD 22 in 2019 and the first half of 2020. Those are global transfer-market observations, not AFRINIC allocation prices. Applied only as an order-of-magnitude benchmark, USD 20 per address makes a /13 correspond to more than USD 10 million of gross address value before transaction costs, reputation differences or policy restrictions.

The comparison does not prove that a recipient could sell the block, that AFRINIC transferred property, or that the registry should have charged market rates. It shows why an approval carried an economic benefit much larger than the service invoice. The institution was distributing access to a scarce input with capital value while continuing to describe the decision through need and membership.

That gap creates scarcity rent. The rent is the difference between the cost of obtaining addresses through the free pool and the cost or value of the nearest alternative. It can accrue to an efficient African network that uses every address. It can accrue to downstream customers through lower prices. It can also accrue to a holder, intermediary or related company. The existence of rent says nothing by itself about abuse. It says the allocation decision deserves controls proportionate to the value distributed.

It also changes the cost of errors. An overly generous approval can remove hundreds of thousands of addresses from later applicants. An overly restrictive denial can force a genuine African operator into a market purchase, address sharing or provider lock-in. A delayed review can destroy a launch window. A mistaken recovery can disrupt running customers and erase financing value.

The remaining pool thus raised the stakes on both sides. It increased the cost of leakage and the cost of defensive overreach. The registry's challenge was not to choose suspicion or generosity. It was to make each high-value decision reproducible.

Phase 1 made demonstrated need a capital-allocation decision

AFRINIC's Phase 1 began on 31 March 2017 when an otherwise valid request could no longer be satisfied without drawing from the final /8. Under the published terms, the maximum request was a /13. Existing members seeking more space had to demonstrate at least 90 percent efficient use of prior resources. The planning period was reduced to eight months.

The AFRINIC exhaustion description records a layered review. Applications entered a ticket channel and were handled first come, first served. A hostmaster evaluated a complete request, another hostmaster reviewed it, and the Registration Services Manager gave final approval in a weekly batch. Approved prefixes could be reserved for up to 45 days while payment and, for new members, the signed Registration Service Agreement were completed. If the requirements were not met, the prefix returned to available inventory and the applicant had to submit again.

This arrangement combined queue order with judgment. The timestamp decided which complete case came first. Hostmasters decided whether evidence demonstrated need and prior utilization. Contractual checks decided whether the member was in good standing. Management approval decided whether the recommendation became a reservation. Payment and signature decided whether reserved stock left the pool.

Each step is reasonable in isolation. Together they form a capital-allocation gate. A /13 could support large access networks, hosting platforms, mobile systems or customer estates. The decision was not equivalent to checking a form. It selected which applicant could avoid millions of dollars of alternative acquisition cost.

The weekly approval batch added another boundary. Batching can improve consistency because comparable cases receive final review together. It can also create a cliff when the remaining pool is close to a policy threshold. A request approved before the Friday batch that triggers Phase 2 may receive treatment different from a similar case still under evaluation. AFRINIC's 2019 notice explicitly said that all applications still under evaluation when the /11 threshold was reached would be assessed under the new clause.

The institutional question is therefore what event fixed the applicant's substantive regime. Receipt did not. Completeness did not necessarily. Pre-approval did not necessarily. The inventory after final approval did. An applicant could enter under Phase 1 expectations and emerge under a /22 maximum.

That transition rule protected the reserved remainder from being consumed by pending large cases. It also transferred threshold risk to applicants. The public needed to know how many requests were caught, their requested size, their time under evaluation and whether the registry's own delay contributed to the change.

AFRINIC did publish pool inventory and policy notices. It did not publish a full historical demand funnel sufficient to reconstruct all lost or reduced requests. Without that denominator, outsiders can observe the pool falling but cannot distinguish rising genuine African demand, accelerated applications, cross-region structures, repeated requests, review backlog or withdrawals.

The capital-allocation power was visible in the block sizes. Its exercise remained much harder to observe.

The 2018 and 2019 drawdown shows why the threshold mattered

The annual figures make the pace concrete. AFRINIC's 2018 annual report says the registry allocated about 6.1 million IPv4 addresses during the year and ended with the equivalent of 0.40 of a /8. It had 1,666 members and its Member Services team handled more than 35,000 request tickets across its responsibilities.

The ticket total is not an IPv4 application count. It includes service interactions of different types and cannot be divided into allocated addresses to produce a meaningful rate. The 6.1 million figure nevertheless shows that the pool was moving in large increments under Phase 1.

AFRINIC's 2019 annual report records 4,574,464 addresses distributed that year. The available pool fell from 0.40 of a /8 in January toward 0.16 in December. The year-end inventory included one /11, one /17, eight /18s, twenty-one /19s, twenty-three /20s, thirteen /21s, seventy /22s, four /23s and forty-three /24s. The report notes that the total included the /11 reserved for Phase 2.

These numbers demonstrate two distinct scarcity problems. Total quantity was shrinking, and prefix composition was fragmenting. A headline such as 0.16 of a /8 can imply one divisible pool. In practice, available blocks had different sizes. An applicant seeking a large contiguous allocation could not be satisfied by an arithmetically equivalent pile of small prefixes without operational consequences.

Fragmentation increases the value of inventory management. The registry decides when to preserve a larger block, when to fill a request with multiple prefixes and when to protect the designated Phase 2 reserve. Those choices affect route aggregation, applicant cost and the date at which the threshold is declared.

The reported drawdown also explains why soft-landing revisions became contentious. Some entities wanted tighter maximums, recurrence limits and dedicated reserves for late entrants or IPv6 transition. Others argued that holding addresses back delayed productive use and would not solve IPv6 adoption. Both positions addressed a real intertemporal choice: use addresses now for demonstrated networks or preserve them for unknown future entrants.

No policy could create more IPv4. A reserve changed who bore scarcity and when. A smaller cap increased the number of possible recipients but could make each allocation operationally marginal. A larger cap supported scale but allowed a small number of approvals to consume the pool. A recurrence limit restrained repeat access but might punish a fast-growing operator. A geographic condition protected a regional narrative but could make multinational African networks difficult to assess.

The 2018-2019 figures do not tell the reader which policy was best. They show why every parameter had distributive force. Millions of addresses could move in one year, and the remaining inventory was already small enough that one large decision affected the regime for everyone behind it.

Phase 2 reduced the ration but did not eliminate repeat demand

AFRINIC announced Phase 2 on 13 January 2020. Approval of a prefix had left no more than one /11 of non-reserved space in the final /8, and the extended statistics showed 102.192.0.0/11 as the only available prefix at the trigger.

The maximum allocation or assignment became /22 and the minimum /24. The planning period remained eight months. Existing members seeking additional space still had to demonstrate 90 percent efficient utilization. Contractual compliance checks continued. Crucially, the policy contained no explicit limit on how many times an organization could request additional IPv4 space.

The immediate reduction was dramatic. A /13 has 524,288 addresses; a /22 has 1,024. The per-request maximum fell by a factor of 512. Yet the absence of a recurrence cap meant the policy constrained transaction size rather than a member's total annual access. A qualifying organization could return with another request.

That distinction matters for both fairness and administration. Repeated small applications increase review volume. They reward organizations able to document utilization quickly and maintain continuous interaction with the registry. They can also permit an operator with genuine rapid growth to obtain incremental capacity instead of receiving one speculative large block.

A 2026 proposal on recovered space and priority illustrates that the issue remains live. Its problem statement says several members had received more than eight /22s within a calendar year by making multiple requests up to the maximum. It also reports, attributing the figures to AFRINIC staff, 783,872 currently available addresses, more than three million recovered addresses and a separate /12 reserve that could later return to the pool.

Those statements come from a proposal under discussion, not an audited allocation report or adopted rule. They should not be treated as proven misconduct or final inventory policy. They do show that the original distinction between per-request maximum and cumulative access has become a current governance question.

If repeat requests are allowed, accountability requires a member-level distribution: number of recipients with one, two, three or more approvals; total annual addresses by beneficially controlled group; time between requests; utilization evidence; and treatment of entities. Without it, a /22 cap can create the appearance of broad rationing while substantial volume accumulates through repetition.

If recurrence is capped, the rule should also confront operational reality. A mobile or broadband network can use 1,024 addresses quickly even with aggressive sharing. Forcing it to wait a fixed period may raise customer cost without preserving meaningful long-run supply. The right design depends on measured demand and alternatives, not the moral appeal of equal block counts.

The 2021 annual report provides a later checkpoint. It records 282,624 addresses issued during the year and lists an available inventory composed of one /12, one /13 and a series of smaller prefixes down to /24. That lower annual issue volume is not evidence that regional need fell by the same proportion: the per-request ceiling, available prefix mix and review conditions had all changed. It does show how quickly the institution moved from multi-million-address Phase 1 years to a smaller and more fragmented Phase 2 supply.

Phase 2 therefore did not end discretion. It moved discretion from the size of one large approval toward repeated proof, related-party identity, utilization timing and the sequencing of many small approvals.

Cross-region attention was predictable, not proof of foreign capture

AFRINIC's policy archive shows that cross-region pressure was not invented after a single dispute. It was discussed openly for years.

A 2011 proposal titled Limited Out of Region Allocation of IPv4 Resources contemplated allowing foreign entities to become AFRINIC members after depletion in other regions. It proposed a doubled membership fee, a /16 limit per application and a ceiling on the stock available to that channel. The proposal was not adopted, but its existence is revealing: entities anticipated that regional exhaustion differences would generate external demand and debated whether Africa should capture some benefit rather than pretend the demand would not appear.

By 2015, public meeting minutes on out-of-region use record a different response. The proposal sought to deter organizations outside Africa from acquiring AFRINIC space for external use and considered a limit under which an African member could use up to 40 percent outside the region. The debate had shifted from whether to admit foreign members to how to police the geography of use.

Neither position supplies a simple rule for modern networks. Applicant incorporation, infrastructure location, route announcement, customer residence, service control and economic benefit can all point to different places. An African company may use a global cloud. A content network may announce the same prefix in several countries. A carrier may serve roaming or multinational customers. A hosting provider may own facilities in Africa while customers sit elsewhere. Geolocation data can be stale or commercially inferred.

The current soft-landing text stated that AFRINIC resources used outside the service region should do so solely to support connectivity back to the African region. The phrase expresses a regional priority but requires judgment. What degree of support is enough? Does a global backbone link count? Must each address have an African endpoint? How should anycast be treated? What evidence proves customer location without exposing confidential data?

Scarcity increases the temptation to answer these questions through suspicion. If an applicant has foreign directors, customers or routes, reviewers may infer extraction. That inference can harm legitimate African operators with global ambitions. The opposite temptation is to accept a thin African corporate presence while most economic use occurs elsewhere. That can defeat the regional priority.

The solution is not a slogan about keeping addresses in Africa. It is an evidence matrix published before application: corporate authority, facilities, network control, traffic function, customer service, routing design, address plan and related entities. Each factor should have a declared weight and a chance for correction. Routing geography alone should never establish control or beneficial use.

Cross-region attention was a foreseeable consequence of unequal scarcity. It justified better evidence. It did not justify treating every internationally connected African applicant as an adversary.

Review power became more valuable because recovery replenished the pool

Ordinary registry review protects record accuracy and ensures that allocations follow adopted policy. Under scarcity, review can also create supply. If a finding leads to return or recovery, the addresses can be allocated again. That makes the reviewer's judgment economically consequential to both the current holder and applicants waiting behind it.

The policy record shows how this pressure surfaced. The Internet Number Resources Review proposal ran through eight versions from 2016 to 2019. The 2019 staff assessment described random, periodic and whistleblower-triggered reviews of IPv4, IPv6 and AS number resources, with recovery of non-compliant resources after an effort to restore compliance and an appeal route for dissatisfied members.

At the 2018 public meeting, the proposal did not move cleanly to adoption; minutes record a need for further discussion. The 2019 meeting account describes both strong support and strong opposition, with the proposal returning for more work. Entities could agree that fraudulent applications and inaccurate records were problems while disagreeing over audit triggers, discretion, recovery and procedural protection.

This conflict should not be reduced to honest stewardship against selfish holders. Nor should every review be portrayed as confiscation. There were legitimate reasons to verify that scarce allocations matched submitted plans, that directory records were accurate and that related requests were not duplicative. There were equally legitimate concerns about a private institution conducting broad audits, interpreting business use and threatening live number resources without sufficiently independent adjudication.

The scarcity connection is structural. Before exhaustion, recovering a block corrected the register and increased ordinary supply. After other regions had exhausted, recovery also captured material market value and redistributed opportunity. A decision that a holder was non-compliant could benefit future applicants and the institution's claim to conserve regional space. That does not prove bias. It creates a conflict risk requiring separation of roles.

The staff who investigate should not be the final reviewers of contested facts. The institution that benefits from replenished stock should not control the only appeal. Emergency action should be limited to duplicate registration, fraud, security-integrity threats or binding legal orders, with continuity protected while ordinary disputes are heard. Recovery should identify the exact addresses and exact breach rather than threaten an entire account by association.

Review outcomes should be published in anonymized form: trigger, policy clause, evidence sought, time allowed, finding, addresses affected, appeal and final disposition. If recovered space later returns to inventory, the stock account should link the event without naming the member. This permits the public to see whether review is correcting records or becoming an opaque supply strategy.

Scarcity made review necessary and dangerous at the same time. The answer was stronger independence, not no review and not unlimited registry discretion.

The scarcity rent appeared in time, risk and institutional leverage

Scarcity rent is often imagined as a cash profit on resale. AFRINIC's last-pool position created several less visible rents.

The first was timing rent. An applicant approved under Phase 1 could receive hundreds of thousands of addresses before the /22 limit. Another with similar long-run need, delayed by review or arriving after the threshold, could receive only a small fraction. The difference arose from timing against inventory, not necessarily from difference in network value.

The second was information rent. Organizations that monitored pool data, policy discussions and approval practices could prepare earlier and structure stronger evidence. Advisers with experience in hostmaster review could reduce incomplete-case delay. Information was publicly available in principle, but the ability to interpret and act on it was uneven.

The third was organizational rent. A corporate group could benefit if related entities were treated separately, while an integrated African operator faced one account-level history. Conversely, an overbroad related-party rule could deny genuinely independent affiliates. Beneficial-control definitions determined access.

The fourth was compliance rent. A holder with a clean, well-documented history could obtain additional space more quickly or preserve existing value. That reward can be healthy. It becomes problematic if requirements are ambiguous, change retrospectively or depend on judgments unrelated to uniqueness, accurate records and the adopted allocation standard.

The fifth was gatekeeping rent. The registry controlled the recognized record needed for allocation, transfer, reverse DNS and resource certification. If it could attach broad policy or contractual conditions to that control, the remaining pool increased its leverage over members. The member could not easily take the request to another recognized regional provider.

The sixth was holder rent. Existing address holders benefited as free supply narrowed and transfer prices rose. Some value reflected productive scarcity: an operator had obtained, maintained and deployed a useful input. Some could reflect historical timing rather than current contribution. The registry could not eliminate that distinction through need review alone.

These rents were distributed across different parties. It would be wrong to say AFRINIC captured all of them. Administrative fees did not equal market value, and the public record does not establish a systematic programme to monetize approvals. The larger concern is that high unpriced value flowed through discretionary decisions without a complete public account.

The proper response is not to auction every address. An auction could exclude smaller African entrants and turn regional policy into revenue maximization. The response is to reduce avoidable discretion, publish who receives the benefit by cohort and make transfer markets legible enough that applicants can compare alternatives.

Scarcity rent cannot be abolished while IPv4 demand exceeds available supply. It can be prevented from becoming hidden institutional power.

African operators paid the cost of both leakage fear and global scarcity

The political framing often presents two sides: protect African addresses from external extraction, or allow a global market to allocate them efficiently. African operators can lose under either extreme.

If controls are too weak, an applicant with minimal regional substance may obtain scarce addresses and deploy or monetize them elsewhere. Later African networks then face a thinner pool and higher market prices. Regional development goals are undermined.

If controls are too broad, a genuine African operator can be penalized for serving international customers, using global facilities or choosing a corporate structure that attracts capital. It may wait through repeated review, receive only small increments, remain dependent on an upstream provider or buy addresses at the same global price the policy sought to avoid.

The operator's cost is not limited to the price per address. Carrier-grade sharing requires equipment, logging, abuse response and customer support. Provider space can require renumbering after a contract change. Market acquisitions need legal and technical diligence. Registry uncertainty raises financing cost because a lender or investor cannot treat continued recognized control as predictable. Long review can postpone revenue.

Small operators may suffer most. A large multinational can hire specialists, maintain multiple upstreams and buy on the transfer market. A local provider may depend on one /22, one financing round and one launch season. A rule designed to protect the region can become regressive if compliance has a large fixed cost.

The remaining pool also created expectations that could delay adaptation. As long as AFRINIC still had addresses, networks could hope for another allocation rather than price a transfer or accelerate IPv6. That hope had option value but also uncertainty. Phase 2's repeat-request possibility could encourage incremental dependence on a shrinking source.

IPv6 remains the only protocol-scale answer to address growth. It does not erase near-term IPv4 need. Customers, services and counterparties still use IPv4, and a new IPv6 network often requires translation or dual-stack capacity. Telling an applicant to deploy IPv6 is technically necessary and economically incomplete.

A fair scarcity policy should therefore measure transition burden. How many customers can the requested /22 support under the proposed sharing design? What logging and legal obligations follow? What proportion of services is reachable over IPv6? What would provider dependence cost? These questions can inform policy without turning the registry into a planner of the applicant's business.

The goal is not to guarantee cheap IPv4 forever. It is to ensure African operators do not pay an additional institutional premium because the last pool attracted global concern and expansive review.

Transfer policy became the missing half of soft landing

A free pool and a transfer market are not separate worlds. As the pool contracts, transfers become the source of larger blocks and faster supply. If transfer rules lag, the registry's allocation gate carries more pressure than it can safely absorb.

AFRINIC's record shows prolonged debate over regional transfers, inbound transfers, out-of-region use and later inter-regional arrangements. At the start of Phase 1, the Last RIR Standing notice observed that other regions already had transfer channels while AFRINIC had no comparable route then available across all directions. African applicants could therefore see market supply elsewhere without a simple recognized way to bring it into the regional register.

That asymmetry increased the value of the remaining pool. It also encouraged policy conflict over whether transfers would drain Africa or attract supply. Inbound-only designs could increase African availability but offer sellers outside the region no reciprocal regional path. Outbound restrictions might preserve stock in the short run while discounting the value of African-held resources and discouraging counterparties from entering.

The registry's technical concern should be settlement integrity: the source has authority, the addresses are not duplicated, disputes are identified, the recipient can be recorded, security entities are updated and the change becomes final in a predictable sequence. Needs tests and geographic restrictions add distributive policy. If they remain, their authority and evidence burden should be explicit rather than hidden inside technical review.

Transfer logs should include dates, prefix sizes, source and recipient regions, status, processing intervals and anonymized failure reasons. Price can be reported in aggregate without exposing contracts. Rejected and withdrawn transfers matter as much as completions because they reveal incompatibility and delay.

This evidence would also discipline scarcity claims. If addresses are leaving the region, report quantities and recipient circumstances. If inbound supply is blocked, report where and why. If internal transfers concentrate holdings, show the distribution. Do not infer regional harm solely from an announcement observed outside Africa; anycast and multinational networks complicate geography.

Soft landing without a credible transfer route makes the free pool carry too many objectives: support growth, protect entrants, police geography, induce IPv6, prevent speculation and preserve regional value. A transfer system does not solve every objective. It separates existing-stock reallocation from the terminal reserve, allowing the reserve policy to become narrower.

The cost of being last was partly the cost of postponing that separation.

Institutional instability amplified the scarcity premium

Address scarcity alone does not explain every AFRINIC conflict. Governance disputes, policy appeals, litigation, elections and operational continuity have their own causes. Scarcity made their consequences larger because the institution controlled records attached to valuable, running resources.

An applicant evaluating AFRINIC did not only ask whether addresses remained. It had to ask how consistently need would be reviewed, whether a policy would change, how long approval might take, whether a dispute could affect registration services and whether a transfer would be recognized. A holder had to ask what evidence could trigger review and what remedy existed before recovery.

These questions create an institutional risk discount. A nominally low-cost allocation may be less valuable if its control conditions are uncertain. A market entity may demand a lower price for a block exposed to transfer restrictions or a higher return for financing it. A customer may prefer provider space with simpler immediate service despite weaker long-run independence.

AFRINIC's status as the final large supplier magnified attention to every institutional decision. Actors in other RIR regions watched because the pool affected global supply and transfer possibilities. Policy advocates saw a chance to shape the last substantial reserve. Existing members saw the value of their holdings rise. Staff faced applications whose commercial consequences exceeded ordinary service administration.

The institution could not resolve this by asserting regional stewardship more strongly. Authority claims do not reduce counterparty risk. Predictable procedures do.

Continuity requires a protected registry record even during dispute. Resource certification, reverse DNS and public registration should not become bargaining tools. Contested business use can be marked and reviewed without immediately disrupting operational services. Independent appeal needs enough power to preserve the status quo while deciding facts.

The remaining pool should also be insulated from institutional finance. Recovery and allocation must not become a way to fund broad activity or reward preferred constituencies. Fees should correspond to verifiable registry service, with scarce-stock decisions separated from budget incentives.

Institutional stability is not the absence of challenge. It is the ability to continue accurate, unique and secure registration while challenge occurs. The last-pool period showed that this capability is part of the economic value of IPv4, not an administrative afterthought.

Causation must be stated more carefully than chronology

Did being last cause cross-region attention, review conflict and scarcity rent? The record supports a qualified answer.

It clearly increased scarcity rent. Other regions had restricted ordinary supply, transfer prices were rising, and AFRINIC could still approve large blocks under Phase 1. The value gap between registry access and alternatives was substantial.

It clearly increased cross-region attention. Policy proposals and meeting records repeatedly addressed foreign membership, out-of-region use, inbound supply and transfer direction. The timing and subject matter are consistent with unequal regional exhaustion.

It plausibly increased pressure on review. When recovered resources can return to a valuable pool, compliance findings have a distributive effect. The resource-review proposal's repeated revisions and contested reception occurred during rapid drawdown.

It did not alone cause every conflict. AFRINIC's contractual model, governance structure, policy rules, corporate disputes and personalities mattered. Large requests may have reflected genuine African demand. Out-of-region routing does not prove extraction. Strong opposition to review can reflect fear of overreach or an interest in avoiding legitimate scrutiny. Public aggregate data cannot assign motive.

The distinction is essential. A watchdog article should not convert temporal sequence into accusation. Its task is to identify the mechanism and the evidence needed to test it.

The mechanism is straightforward: unequal scarcity raises the value of access; higher value increases attempts to qualify and efforts to restrict qualification; those efforts increase the importance of review; review outcomes can replenish supply; and the institution's weak substitutes make its decisions hard to escape.

Testing the mechanism requires applicant-level but privacy-protected data. Publish request size, timing, corporate-control band, regional-use evidence category, prior utilization, decision, review, appeal and final allocation. Link recovered stock to anonymized recovery reasons. Compare decision time before and after phase transitions. Report repeated requests by controlled group.

Without those records, both institutional defenders and critics can select anecdotes. One side can call every large allocation development. The other can call every review capture. Neither gives African operators a reliable basis for planning.

Being last raised the risk. It did not relieve analysts of proving the outcome.

A narrower review standard would protect both the pool and running networks

The registry does need a review standard. The standard should be tied to functions that can be observed and defended.

First, verify applicant authority. The legal entity, authorized signatory and beneficial control relevant to duplicate requests should be clear. Corporate groups should not multiply eligibility through paper separation, but minority investment and ordinary commercial relationships should not be treated as control.

Second, verify uniqueness. No prefix should be allocated, transferred or recovered in a way that creates competing recognized claims. Conflicts should be flagged before any new registration.

Third, verify need under the adopted allocation rule using published measures. State the planning horizon, utilization calculation, treatment of shared addressing, infrastructure reservations and customer assignments. Do not change the evidence standard after submission without notice.

Fourth, verify regional nexus through multiple facts rather than IP geolocation alone. Incorporation, facilities, network control, service contracts, routing design and African customer dependency can all contribute. Publish examples and counterexamples.

Fifth, verify record accuracy. Contacts, route data, resource certification authority and assignments should be maintained. Correction should be the first remedy for ordinary defects.

Sixth, separate technical continuity from commercial dispute. A contested use should not automatically invalidate all routing-adjacent records. Interim measures should preserve customer service unless duplicate registration, fraud or a serious security threat requires urgent action.

Seventh, provide independent review before irreversible recovery. The reviewer should have access to the evidence, authority to pause action and a duty to publish anonymized reasoning.

Eighth, return recovered space through a declared rule. Quarantine, reputation assessment, prefix preparation and priority should be public. Staff should not choose a recipient ad hoc.

This standard would not satisfy those who want the registry to enforce broad ideas of regional economic justice. A private number registry lacks the electoral mandate, information and remedial capacity to act as a continental industrial planner. Governments can subsidize connectivity, finance networks and regulate competition through public law. The registry should not smuggle those choices into address approval.

Nor would the standard satisfy those who want allocations without proof. Scarce common inventory cannot be distributed on assertion. Objective need and identity checks protect later applicants.

The narrow standard protects both sides because it confines power to evidence the institution can competently evaluate.

The public account should follow demand, not only addresses

AFRINIC has long published pool statistics and delegated-resource files. Those are necessary for conservation. They are not enough for governance.

A complete scarcity account should begin with applications. For each month and phase, publish the number of new and additional requests, requested addresses, complete cases, pending cases, approvals, reductions, rejections, withdrawals and lapses. Separate applicant delay from registry time. Show the weekly approval cohort during Phase 1 and the sequence of repeated /22 requests during Phase 2.

Then publish distribution. Report the share of addresses by recipient-size band, country, operator type and beneficially controlled group, using aggregation where privacy requires it. Show first-time and repeat recipients. Report how many applicants obtained less than requested and whether they accepted.

Next publish the regional-nexus account. Use stable reason categories: facilities, customers, backbone support, anycast, multinational operation, related company, limited public evidence evidence and conflicting evidence. Do not publish sensitive customer lists.

Add the review account. Report random, triggered and member-requested reviews separately; the policy or contract clause; time to cure; addresses in question; recovery proposed; appeal filed; and final result. A whistleblower allegation should never be counted as a violation before verification.

Add the stock account. Reconcile opening inventory, new IANA returns, member returns, recoveries, quarantine, reservations, allocations and closing inventory by prefix size. Identify the /12 and other dedicated reserves separately so headline availability cannot mix policy-restricted and allocable stock.

Finally, add the market account. Publish transfer completions, failed cases, timing and aggregate price bands where parties voluntarily or confidentially report them. Compare free-pool waiting and transfer timing without claiming they are identical services.

The reports should be downloadable and historically immutable. If a figure is corrected, preserve the prior value and reason. Members should be able to reproduce annual totals from monthly events.

This is not excessive disclosure for an institution allocating an input worth millions at market benchmarks. It is the minimum evidence needed to determine whether being last benefited African connectivity or mainly increased institutional conflict.

The cost of being last should shape the next scarcity institution

AFRINIC's experience contains a warning for any future number-resource arrangement. A reserve is not merely technical inventory. Once alternatives become expensive, the operator of the reserve acquires distributive power whether it seeks that power or not.

Good institutional design anticipates the conversion. It defines thresholds before they are reached. It states which event fixes an application's rules. It limits review to objective facts. It reports unsuccessful demand. It builds transfer settlement before the free pool becomes the only practical source. It protects continuity while disputes are heard. It makes the record portable enough that one corporation's instability does not endanger every holder.

The design should also resist territorial simplification. Regions matter because institutions, infrastructure gaps and development priorities differ. Packets, corporate groups and customers cross borders. A registry can require a genuine service-region nexus without pretending that every address has one permanent geography.

Most importantly, the institution should not confuse holding the last pool with owning the region's Internet future. Addresses create value when operators deploy them. The registry creates value by preserving unique, accurate and secure records. Its authority should follow that contribution.

AFRINIC's remaining pool offered African networks a period of lower-cost access after other regions had exhausted. The annual reports show millions of addresses entering use. That is the benefit.

The same pool made a needs approval worth far more than its fee, turned regional-use wording into a trade boundary, gave recovery decisions market consequence and concentrated global attention on a private membership institution. That is the cost.

Both facts can be true. The right conclusion is neither that AFRINIC should have given everything away quickly nor that it should have controlled every subsequent use. It should have distributed the finite stock under narrow, published rules while building a transparent market and independent review around the installed base.

Being last was never going to preserve abundance. It only determined where scarcity first became institutional pressure. The unanswered accountability question is not how many addresses remained on a given day. It is how much unserved African demand, cross-region value and review power stood behind each address the institution still controlled.