AFRINIC and the economics of fee incidence and regressivity

The invoice under the generator

An AFRINIC invoice is rarely the largest bill on an African network operator's desk. For a small ISP, a university network, a municipal fibre project, a data-centre operator or a local public-sector network, diesel, grid instability, transit, international capacity, equipment imports, spectrum charges, wayleave costs, customer churn and currency weakness usually dominate the month. The registry fee arrives in another category. It is a dollar-denominated charge attached to the administrative record that makes the network legible to counterparties. It is not a fuel bill, a router lease or a transit contract. Yet it sits beside all of them because the network cannot operate as a recognised participant in the global numbering system without the registry relationship behind its address space and autonomous system number.

That is where the economics of fee incidence begins. The question is not whether a regional internet registry needs revenue. AFRINIC must run records, member systems, WHOIS and RDAP publication, reverse DNS, routing registry functions, RPKI services, support desks, security controls, audits, meetings, governance processes and ordinary administration. The harder question is who actually bears fixed registry costs, legal-budget shocks, reserve rebuilding, enforcement overhead and scarcity-era administrative friction when the members are unequal and cannot exit the monopoly ledger. The invoice identifies the legal payer. It does not identify the economic bearer.

The cost can be borne by downstream customers through higher prices, by staff through deferred hiring, by local users through lower redundancy, by universities through delayed network upgrades, by shareholders through lower returns, by governments through a smaller connectivity budget, or by the region through a risk discount on AFRINIC-administered IPv4 resources. When the registry is healthy and narrowly run, that incidence may be modest. When governance failure and litigation convert member fees into crisis financing, the incidence changes shape.

AFRINIC is a hard test because its members are not equal economic units. A national carrier, a mobile group, an internet exchange point, a university, a small wireless ISP, a public-sector network, an enterprise end-site, a data-centre operator and a company holding millions of IPv4 addresses do not absorb the same dollar cost in the same way. Some can spread annual fees across millions of customers. Some cannot. Some can hire counsel, follow litigation, send people to meetings and model policy risk. Some have one engineer doubling as billing administrator. Some hold address space whose market value dwarfs their annual registry fee. Some need a small block only to keep local services reachable.

Equal-looking fees, equal deadlines and equal compliance rules can therefore be regressive even when the table is tiered by address holdings. A fixed annual charge falls harder on a low-margin network than on a large incumbent. A late-payment penalty in dollars is not the same burden when one member earns in hard currency and another earns in a depreciating local currency. A months-long uncertainty over transfers, reverse DNS or good standing is not the same cost when one member has spare address inventory and another is rationing customer growth. A legal budget spread across all members is not neutral if the dispute concerns a theory of control that benefits some business models and threatens others.

This is not another argument about reserve constitutions, member rights or board oversight in the abstract. The fee-incidence problem is narrower and more distributive: how costs move through a captive registry relationship when the registry is a settlement utility, IPv4 is scarce, and operators have very different margins, currencies, legal capacity and dependence on continuity.

The mechanism is sharper because AFRINIC is not an ordinary supplier. A dissatisfied operator cannot take the same AFRINIC-administered prefixes to another regional registry. It cannot ask ARIN, RIPE NCC, APNIC or LACNIC to become the authoritative recordkeeper for the same African resources merely because AFRINIC's governance has become expensive or uncertain. The registry is closer to a settlement utility: a shared ledger whose job is to keep uniqueness, recognition and administrative continuity stable. Its fees therefore resemble a quasi-tax on participation in a necessary coordination system.

That does not make fees illegitimate. It makes the budget test stricter. A compulsory registry fee can be justified when it pays for the narrow ledger: accurate records, verified authority, publication services, reverse DNS, RPKI, transfer recording, security and continuity. It becomes harder to justify when the same fee base is asked to fund prolonged litigation, discretionary enforcement, institutional positioning, election repair, governance crisis and broad programmes whose benefits do not track the members paying the bill. In that setting, regressivity is not only about how many dollars appear on the invoice. It is about how much unchosen institutional risk is embedded in the invoice.

AFRINIC's crisis gives the analysis concrete ground. Public reporting has described address-record corruption, the dispute with Cloud Innovation, provisional freezing of bank accounts, years without a functioning board, court-supervised receivership, annulled elections, allegations over powers of attorney, renewed board legitimacy questions, legal spending anxiety, and continuing conflict over IPv4 leasing, transfers, winding-up and institutional continuity. These are not decorative details around a fee schedule. They are the conditions under which members are asked to pay. They determine whether a fee is a clean charge for registry service or a levy on unequal operators to finance the consequences of institutional failure.

A monopoly ledger is not a normal vendor

A normal vendor relationship contains a discipline that the regional registry model largely lacks. If a software provider raises prices, a buyer can migrate. If a conference becomes wasteful, sponsors can stop attending. If a trade association loses credibility, companies can leave and still conduct business. Exit is never costless, but it is available often enough to make pricing and service quality more accountable. AFRINIC's members do not enjoy that kind of market exit for the number resources already recorded under the African registry.

That absence of exit is the foundation of fee incidence. The registry relationship is not just a contract for service. It is a recognition relationship. The member needs the authoritative record to be accurate. It needs contact data, reverse DNS, routing-security material, WHOIS or RDAP publication, transfer records, dispute status and member standing to be credible to customers, upstreams, peers, banks, auditors and other registries. Even when the routing of packets is technically independent of an office in Mauritius, the administrative record supplies trust around who controls which identifiers.

This is why the narrow ledger idea matters. AFRINIC's defensible core is the maintenance of a clean, secure and reliable record of number-resource recognition. That record is the settlement layer for scarce identifiers. It reduces transaction costs because counterparties do not have to reinvent proof of control for every prefix. It supports security because contact data, certification and registry-publication systems can be checked. It supports markets because transfers, leases, mergers and network expansions need a reliable statement of who is recognised for what. It supports public connectivity because universities, governments, ISPs and enterprises can build on a stable administrative base.

The narrower the registry remains, the easier the fee is to defend. Members pay because the shared ledger must exist and must be reliable. The wider the registry's discretion becomes, the more the fee starts to resemble a tax funding policy choice, legal strategy or economic gatekeeping. If a registry reviews customer geography, reopens business models, blocks transfers on broad regional theories, or fights for institutional authority in court, it may be doing so from funds contributed by members whose interests are not aligned. The invoice then finances not only recordkeeping but a contested view of what the registry should be.

AFRINIC's official materials are useful as limited factual exhibits. The organisation is a nonprofit, member-based body registered in Mauritius, serves Africa and parts of the Indian Ocean, and manages IPv4, IPv6 and autonomous system numbers. Its public materials list reverse DNS, WHOIS, IRR, DNSSEC and RPKI services, and its fee page ties fees to operations, organisational financial health and board validation. Those facts establish the formal frame, not the economic one.

Institutions naturally describe fees as operational support, policies as community decisions and litigation as defence of the registry. Incidence asks a different question: when AFRINIC spends a dollar, whose economic burden does that dollar become, and what risk is the payer being made to insure? A fee funding RDAP uptime and key-management controls is different from a fee funding a discretionary fight over whether a holder's customers are too far from the service region. A fee funding election verification is different from a fee funding the reputational cost of an annulled election. A fee funding ordinary counsel is different from a fee funding a multi-year legal posture with consequences for all holders.

Captive membership also changes the moral status of cross-subsidy and crisis costs. In a voluntary association, broad programmes and emergency spending are easier to tolerate because dissatisfied payers can leave. In a monopoly ledger, the member may be forced to fund activities it did not choose in order to preserve a record it cannot move elsewhere. If bank accounts are frozen, elections fail or litigation consumes management attention, the costs diffuse into delayed service, higher legal bills, deferred programmes, staff time and future fee pressure.

A settlement utility should be designed to minimise that diffusion. Its core services should keep working when board politics fail. Its reserve should protect the ledger, not discretionary institutional ambitions. Its legal budget should be classified so members know whether they are funding continuity, governance repair, ordinary corporate advice, resource disputes or policy defence. Its mandate should be bounded so compulsory fees cannot be turned into a general levy for whatever the current leadership calls community protection.

AFRINIC's problem is not that it charges members. The problem is that the crisis has blurred the line between fees for registry service and fees for institutional survival. Once that line is blurred, regressivity becomes unavoidable. The smallest and least liquid operators bear not only their own invoice but the uncertainty created by a registry that is both essential and contested.

The fee table and the hidden incidence

AFRINIC's public fee schedule is not a flat poll tax. It is tiered by resource category, and that matters. LIR annual IPv4 fees range from small categories into much larger bands. The published table lists, among other entries, $1,000 for an LIR Micro holding from /24 to below /23, $1,400 for Extra Small, $6,400 for Small, $12,800 for Medium, $30,000 for Very Large and $38,400 for Extra Large. End-site annual fees are lower, running from $200 for Micro to $2,500 for Extra Large, with an ASN-only end-user fee of $50. Academic and research institutions may receive a 50% discount if they qualify. Critical infrastructure requests may receive a full discount under stated conditions.

At first glance this seems progressive: larger holders pay more, end-sites pay less, discounts exist for academic, critical infrastructure and IPv6-only cases, and the table avoids the crudest flat-fee regressivity. But incidence is not measured only by the schedule's slope. It is measured by the cost relative to the payer's ability to absorb it and by the non-price burdens bundled with the fee relationship.

A small LIR paying $1,400 may face a far higher burden relative to free cash flow than a large holder paying $38,400. The large holder can amortise the fee over address inventory, customer base, financing capacity or leasing revenue. The small operator may be paying in hard currency while collecting retail subscriptions in a weak local currency, absorbing payment frictions, generator costs and churn. The table rises with address holdings, but the ability to absorb fixed institutional risk rises much faster.

The billing calendar reinforces the point. AFRINIC's fee page says renewal invoices are issued on November 1, with a 5% discount for payments received by December 31 and a January 31 due date. A 28-day moratorium runs to the end of February. Fees received in March, April and May attract 5%, 10% and 15% late-payment penalties, and on June 1 the closure process starts.

For a well-capitalised operator, this is ordinary discipline. For a smaller operator, the calendar can collide with seasonal revenue, public-sector procurement cycles, capital controls, foreign-exchange shortages, bank delays or local tax obligations. The formal due date is the same. The practical cost of meeting it is not. A member paying from a country with dollar scarcity may face a higher effective price than a member whose revenue is naturally dollar-linked. A university or government network may have a budget year that does not map neatly onto the registry's billing cycle. A data-centre operator may be waiting for customers who pay late. The fee table does not show those constraints, but they determine who bears the cost.

There is also administrative incidence. Paying the invoice is only the most visible act. Members must keep billing contacts current, maintain registry files, respond to resource checks, handle paperwork for transfers or additional resources, monitor policy changes, verify representatives, manage credentials and understand the consequences of good standing. For a large carrier, these tasks are distributed across departments. For a small ISP, they may fall on the founder or a senior engineer. The opportunity cost is higher when managerial bandwidth is scarce.

Allocation, assignment and transfer rules create the same hidden incidence. Under scarcity, evaluation is an economic bottleneck: need justification, payment windows, utilisation requirements and Phase 2 limits can delay business plans even before a fee is paid. AFRINIC's fee page says transfers between existing members did not at that time attract IP or ASN transfer fees, but all parties had to be in good standing. The real cost is therefore not only a transfer fee. It is due diligence, good-standing risk, needs justification, the registration service agreement, possible loss of legacy status and uncertainty over whether the registry will treat the transaction as a narrow record update or a broader policy event.

This is the hidden incidence of a registry fee schedule. The visible charge buys access to a system whose discretion can impose invisible charges. When the system is predictable, those hidden costs are small. When the system is in crisis, they expand. A member paying $1,400 may be paying a modest invoice and an enormous uncertainty premium at the same time.

Why equal treatment can be regressive

Regressivity in registry economics does not require malice. It can arise from formally equal rules applied to members whose constraints are radically unequal. A late-payment penalty is equal in percentage terms but unequal in practical burden. A good-standing condition is equal in wording but unequal in bargaining effect. A request for documentation is equal in form but unequal in staff time. A legal dispute that delays institutional recovery affects all members but not equally. Those least able to hedge the risk often pay the highest relative price.

Consider a small wireless ISP in a market where households churn quickly because electricity costs and food prices squeeze disposable income. It needs public IPv4 addresses because carrier-grade NAT has operational costs, customers still use IPv4 services, enterprise clients require routable endpoints, abuse handling depends on traceability, and upstreams ask for clean registration. It also needs RPKI, reverse DNS and accurate contacts because reputation matters. The registry fee is not its largest cost. But it is a hard-currency, compliance-linked obligation tied to services that can affect revenue. If the registry's legal troubles delay support or raise uncertainty over future fees, the operator has little capacity to absorb the shock.

A large incumbent faces the same registry but a different incidence. It may pay more in the table, yet it can spread the cost across a large subscriber base, commercial accounts, wholesale contracts and finance functions. It may have lawyers who read bylaws, staff who monitor policy lists, and executives who can attend meetings. Its marginal cost of registry compliance is lower relative to revenue. It may also benefit from rules that restrain new entrants, limit address liquidity or preserve scarcity rents around existing holdings. A fee table that charges it more can still be regressive if the associated institutional system protects incumbent advantage while imposing fixed costs on smaller networks.

A university or research network experiences another pattern. It may qualify for a discount, which is important. But a discount on the invoice does not remove the burden of public budgeting, procurement approval, currency conversion, audit rules and technical staff scarcity. If AFRINIC's governance crisis turns member communications into a stream of legal and election uncertainty, the university is not well placed to investigate every claim. If the registry's systems or processes slow, the university's ability to serve students, researchers and public institutions can be affected. A lower nominal fee can coexist with high dependence.

A data-centre operator sees the cost through customers. Customers ask for IP addresses, reverse DNS, abuse contacts, clean routing, service-level confidence and contract certainty. If the operator holds AFRINIC space, a registry dispute can become a sales objection. The customer may not understand AFRINIC's bylaws or the history of Cloud Innovation, but it understands uncertainty. It may demand indemnities, prefer a provider with space from another region, or discount the service. That discount is a form of fee incidence. The registry crisis becomes a private cost borne by a member that did not cause it.

Public-sector networks face still another incidence. A government department or municipal project may hold resources to support health, education, identity, tax, emergency or local-service systems. It may not be a profit-maximising network, but the registry relationship still has costs. If fees rise to replenish reserves or pay legal expenses, the burden may come from public funds. If administrative friction delays changes, public users bear service degradation. If the registry's crisis invites geopolitical or intergovernmental intervention, public-sector operators may face pressure unrelated to technical need. Equal treatment in the registry can hide unequal political exposure.

These cases show why address quantity alone is an incomplete measure of burden. A small address holding can be operationally critical. A large holding can be financially diversified. A small annual fee can be regressive if it is fixed, compulsory and attached to a legal monopoly. A larger annual fee can be less burdensome if it sits on top of asset-like holdings whose market value is large and whose holder can monetise or finance them. The same schedule can look progressive by prefix size and regressive by business resilience.

The problem becomes sharper when litigation and governance failure are included. If AFRINIC spends on legal defence, election repair or court-supervised continuity, those costs ultimately come from somewhere: existing fees, reserves, deferred projects, staff time, external assistance or future fees. The burden falls on the whole membership, but the dispute may concern a narrow set of actors or a contested theory of registry authority. A small ISP may pay for the consequences of a fight over millions of addresses held by others. A university may help finance legal positions about commercial leasing that it neither practices nor understands. A government network may bear risk from a corporate-law dispute in Mauritius because the registry cannot be easily replaced.

This is the economic reason to keep the registry narrow. The broader the mandate, the more ways there are for unequal members to be forced into cross-subsidising each other's conflicts. A narrow ledger still requires fees, but the benefits are more common: accurate records, service continuity, security and uniqueness. A broad gatekeeper creates distributive politics. One class of members pays for rules that advantage or disadvantage another. Regressivity then becomes political as well as financial.

AFRINIC's crisis should therefore be read less as a morality play and more as an incidence map. The visible payers are members. The bearers include customers, staff, public agencies, smaller networks, buyers, sellers, lessees, lenders and the region's future entrants. The more unequal the membership, the more carefully the registry must justify every compulsory dollar.

IPv4 scarcity changed the tax base

AFRINIC's fee schedule was not designed for an era in which the remaining IPv4 pool, transfer markets and leasing models would turn registry recognition into a high-value economic interface. The official fee page still reads like an administrative cost-recovery document. The exhaustion page reads like a scarcity-management document. The market reads both through a different lens: IPv4 is scarce, operationally necessary and priced. That change alters fee incidence even if the fee numbers stay the same.

AFRINIC's exhaustion materials state that the region entered Phase 1 of IPv4 exhaustion on March 31, 2017 and Phase 2 on January 13, 2020. They describe justified need, first-come handling, evaluation by hostmasters, peer review, final approval, payment windows, utilisation requirements and Phase 2 limits, including minimum and maximum allocation sizes. The materials also note the broader global context: other RIRs had already exhausted their free pools and were issuing from final reserves. These are factual exhibits, not economic conclusions. The conclusion comes from the market: a registry that once allocated an abundant administrative input now administers a scarce operational asset.

The Internet Governance Project's 2021 analysis put numbers on the change. It reported that transfer-market prices for IPv4 addresses had risen from roughly $8 per address in 2017 to roughly $30 by 2021. At that price, a /16 block carries value in the millions of dollars. The same analysis reported that Cloud Innovation had acquired nearly 7 million AFRINIC IPv4 addresses, paid about $10,000 a year in registry fees and leased addresses to customers at $2 to $3 per address per year. Those figures were part of a contested analysis of arbitrage, but they illustrate the scale mismatch: administrative fees can sit on top of balance-sheet-scale economic positions.

The right answer is not simply to make AFRINIC charge market rent on addresses. That would risk turning a registry into an extraction machine. If the registry's role is narrow service, its fees should reflect the cost of reliable service, not the full market value of every number it records. A registry is not a landlord of the internet. It should not treat member resources as its own tax base merely because scarcity has made them valuable.

But the opposite answer is also insufficient. A registry cannot say fees are purely administrative while exercising broad discretion over transfers, need, use, leasing, regionality or status. If the fee is for narrow recordkeeping, the mandate should be narrow recordkeeping. If the mandate expands into economic control, the fee becomes the price of access to a gatekeeper whose decisions can affect capital value. The member then pays twice: once through the invoice and again through the discount applied to resources administered by a discretionary institution.

Scarcity also changes who benefits from low fees. A very low annual charge can be helpful to a small operator, but it can be much more valuable to a large holder that can monetise address inventory. A large holder paying a capped or tiered association fee may enjoy a high ratio of address value to registry cost. A small operator paying a lower fee may still be stretched because it lacks scale. The same low-price policy can be pro-entry in one case and a subsidy to inventory holders in another.

That is the paradox of regressivity under scarcity. Raising fees across the board can hurt small members. Keeping fees too low can underfund controls, audits and continuity while allowing large holders to capture enormous value from cheap recognition. Funding crisis costs through ordinary fees can punish small members for disputes driven by large-stakes scarcity. Charging transfer or market-activity fees can discipline speculative movement but also obstruct legitimate growth if poorly designed. There is no simple price that solves all of this.

The more practical rule is scarcity-era fee discipline. Mandatory fees should pay for mandatory registry functions, and those functions should be defined narrowly enough to maintain trust across unequal members. If address scarcity creates asset value, the registry should respond with stronger record integrity, faster transfer finality, clearer dispute markers, better authority verification, transparent legal categories and proportional enforcement. It should not respond by turning every fee payer into a financier of broad economic planning.

AFRINIC's transfer policy shows the tension. The consolidated policy manual says transfers within the region apply to organisations with justified need that cannot be satisfied by AFRINIC. It requires the source to be the current rights holder recognised by AFRINIC and not involved in dispute over the resources. It requires the recipient to justify need before AFRINIC, become a member, be subject to current policies and sign the Registration Services Agreement. It also states that transferred IPv4 legacy resources will no longer be regarded as legacy resources. These provisions may serve conservation and order. They also create transaction costs and status risk.

In a liquid asset market, a buyer and seller price their own need, timing and risk. In a registry-controlled transfer environment, the registry's administrative judgment becomes part of the price. A buyer discounts for delay. A seller discounts for eligibility uncertainty. A legacy holder discounts for status conversion. A small operator discounts because it may not have the staff to navigate the process. A large operator may absorb the process but demand a lower price. These are not line items in AFRINIC's fee schedule. They are scarcity-era incidence.

The conclusion is not that AFRINIC should abandon need, conservation or regional policy overnight. It is that scarcity makes the cost of each rule measurable in market behaviour. A rule that once looked like fair allocation can become a tax on liquidity. A fee that once looked like dues can become a price for access to a risk-bearing recognition layer. The more valuable IPv4 becomes, the less credible it is to discuss fees as if they fund only a help desk and a database.

Crisis financing and the member-funded downside

AFRINIC's recent history turns fee incidence from theory into a budget problem. The Cloud Innovation dispute, the bank-account freeze, the board failure, receivership, election annulment and later litigation did not remain isolated events. They affected the costs of running the registry, the ability to appoint leadership, the credibility of elections, the need for legal advice, the availability of ordinary services and the perception of AFRINIC-administered resources. Members paid through more than invoices.

In 2021, AFRINIC's conflict with Cloud Innovation escalated after the registry alleged policy and agreement breaches related to the use of IPv4 resources. AFRINIC initiated procedures that could have resulted in withdrawal of Cloud Innovation's resources. Cloud Innovation treated the matter as existential and went to court. The Internet Governance Project reported that the Supreme Court of Mauritius provisionally froze up to $50 million of AFRINIC funds held at two banks after Cloud Innovation sought urgent relief. The same analysis criticised both sides: AFRINIC for an over-aggressive assertion of policy righteousness and Cloud Innovation for legal measures that endangered the registry. That two-sided criticism is useful because incidence does not require one pure villain. It requires a cost path.

The cost path is clear. When a registry threatens a remedy that could destroy a member's business, the member has an incentive to litigate intensely. When the member litigates in a way that freezes accounts or paralyses governance, the registry and other members bear the cost. When the registry fights back, legal costs rise. When governance organs fail, courts and receivers step in. When elections fail, the recovery timetable extends. The original dispute over one holder's address use becomes a levy on the whole ecosystem.

The Register reported in March 2026 that AFRINIC accused Cloud Innovation, Larus and associated advocacy campaigns of trying to paralyse the organisation through litigation and procedural roadblocks. Cloud Innovation's side framed the problem as high-consequence registry power over economically critical number resources without commensurate legal and financial liability. Both accounts describe the same incidence channel from opposite sides. Legal conflict consumes member-funded institutional capacity.

This is not an argument that a registry should never litigate. A registry must be able to defend records, recover from fraud, comply with court orders, enforce contracts and protect continuity. Nor is it an argument that members should lack court access. If a registry threatens unlawful or disproportionate harm, courts are a necessary safeguard. The incidence problem arises when litigation becomes the main mechanism for resolving boundaries that should have been clarified through policy, contract, independent review and proportional remedies before the dispute became existential.

Legal-budget moral hazard appears on both sides. A registry may take broad enforcement positions if it believes legal costs can be funded by member fees and liability will be limited. A large member may pursue aggressive litigation if it can impose operational costs on the registry and thereby increase bargaining leverage. Smaller members fund or suffer the conflict without controlling either strategy. They are the silent taxpayers of the dispute.

The solution is not to starve AFRINIC of legal capacity. A registry without counsel would be fragile and easily bullied. The solution is legal-budget classification. Members should be able to see, at an aggregate level, how much is spent on core service continuity, corporate governance, elections, resource disputes, employment, regulatory engagement, external interventions and strategic litigation. Privilege can be protected. Billing narratives and confidential advice can be redacted. But members need to know whether their fees are preserving the ledger or financing discretionary institutional war.

Reserve policy has to follow the same logic without turning this into a general reserves debate. A continuity reserve that covers payroll, systems, security, data escrow, disaster recovery and critical services is legitimate. A reserve that becomes an unclassified war chest is different. The incidence question is: whose insurance is this? If it insures members against registry service failure, it belongs inside the narrow ledger fee. If it insures management against member pressure or funds an expansive enforcement theory, it becomes a quasi-tax.

AFRINIC therefore needs a continuity firewall. Essential registry services should be financially and operationally insulated from litigation, board turnover and election disputes. A member lawsuit should not threaten reverse DNS for unrelated networks. A board fight should not imperil RDAP publication. A bank-account dispute should not prevent basic security operations. A legal reserve should not be available for every discretionary initiative without classification and authority.

Crisis financing is the moment when equal fees become most regressive. Large members can hedge, litigate, delay, switch suppliers or absorb uncertainty. Small networks cannot. They need the registry to be boring precisely because everything else in their operating environment is not.

Record corruption and the cost of trust repair

AFRINIC's fee incidence cannot be separated from its earlier record-integrity crisis. In 2019, KrebsOnSecurity reported allegations that a senior AFRINIC staff figure, Ernest Byaruhanga, had been linked to companies involved in the sale or movement of African IPv4 address space. Researchers traced records involving dormant or defunct entities and estimated that the affected addresses exceeded $50 million in market value. AFRINIC's then chief executive said the organisation was aware of the allegations and investigating. Byaruhanga had resigned shortly before the reporting.

For a fee analysis, the important point is not only the alleged misconduct. It is the cost of restoring trust after misconduct at the record layer. A registry's product is not merely a database. It is confidence that the database reflects legitimate control. If historical records can be altered through insider knowledge, shell entities, dormant allocations or weak authority checks, every member faces higher diligence costs. Buyers ask more questions. Sellers provide more documents. Lenders discount. Customers worry about clean title in the ordinary commercial sense, even if the formal registry language avoids property. Staff must audit. Lawyers must review. The registry must explain. Trust repair becomes a cost centre.

Who pays for that repair? In a narrow sense, AFRINIC pays through staff time, audits, legal advice and systems work. In an economic sense, members pay. They pay through fees, slower service, increased documentation, reputational risk and the discount applied to AFRINIC records. The member whose records were never touched still bears part of the cost because the ledger's credibility is collective. Once a registry has had a public record-corruption scandal, all holders must live with a higher verification environment.

The history also helps explain AFRINIC's later enforcement posture. After a scandal involving alleged misappropriation, a registry has a strong incentive to show that it can police misuse. Members and outside observers may demand tougher review. Staff may become more cautious. Management may see aggressive enforcement as proof of reform. That impulse is understandable. It is also dangerous if the registry cannot distinguish fraud repair from economic control.

Fraud repair concerns false authority, stolen or misrepresented records, forged documents, dormant-company abuse, staff conflicts, fabricated need or deliberate deception. Economic control concerns whether a member's later business model, customer geography, leasing arrangement or commercial use fits a contested theory of regional purpose. The remedies should be different. A registry that treats both as the same problem risks imposing the cost of its past control failure on members whose conduct is not fraudulent but commercially inconvenient to a policy view.

This is where regressivity enters again. Large members can hire counsel to distinguish fraud allegations from policy disputes. Smaller members may not. If AFRINIC responds to past corruption with broad documentation burdens, small networks will feel the compliance cost more heavily. If the registry uses fees to fund enforcement systems without transparent classification, members cannot tell whether they are paying for ledger integrity or for an expanded gatekeeping role. If the registry's public language merges cleaning up records with stopping commercialisation, the cost of trust repair becomes a political levy.

A better model would make trust repair narrow and auditable. AFRINIC should be funded to harden staff access controls, authority verification, change logs, dormant-resource review, dual control for high-risk changes, whistleblowing paths, independent audits, member notifications and historical remediation. Those are legitimate ledger costs. They reduce the risk that any member's record can be commandeered or quietly altered. They benefit the whole membership and justify mandatory funding.

By contrast, broad campaigns against business models should not be smuggled into the same budget category. If the registry believes a business model violates policy, it should identify the policy, evidence, process, remedy and appeal path. If it wants new restrictions, it should use the policy process and disclose economic effects. If it wants to litigate, the legal budget should be classified. The member paying for anti-fraud controls should not discover that the money also funds a discretionary theory of market supervision.

The old corruption story also undermines a common defence of equal fees: that every member receives the same trustworthy ledger. After record corruption, that trust must be rebuilt, and the rebuild cost is not distributed equally in economic terms. Members with scarce administrative capacity pay more in time. Members seeking transfers pay more in due diligence. Members in markets already treated as risky pay more through counterparties' scepticism. Members with small holdings but high operational dependence may pay a large risk premium for a scandal that involved much larger blocks.

AFRINIC's lesson is therefore double. A registry must have enough money to repair integrity. But because integrity failures are so costly to unequal members, the repair budget must be visibly about the ledger, not institutional self-protection. Trust repair funded by captive fees is legitimate only when members can see that it reduces the risk they are forced to share.

Elections, receivership and the continuity surcharge

Governance failure has a price even when the registry continues to answer tickets. AFRINIC's years without a stable board, its receivership, and the annulled 2025 election created what can be called a continuity surcharge. It is not a line item in the fee schedule. It is the added cost members bear because the institution that records their resources must be kept alive through exceptional arrangements.

Public reporting describes the sequence. AFRINIC had no board from 2022. Courts in Mauritius appointed a receiver to preserve the organisation and arrange elections. A planned June 2025 election was suspended minutes before the end of in-person voting after questions were raised about powers of attorney and voting authority. The receiver annulled the election. Some representatives reportedly found that someone else had already voted on their behalf using a power of attorney they had not provided. ICANN demanded explanations and warned about possible compliance review. A later September 2025 election produced eight directors, seven endorsed by Smart Africa, but litigation and doubts about process did not disappear.

For this fee analysis, the point is not who should have controlled the board. It is who paid for the breakdown. When an election fails, members pay through delay, legal advice, repeated process, risk discount, external intervention and governance distraction. When a receiver is needed, members pay for the fact that ordinary corporate organs have failed. When a new board is elected under contested conditions, members pay for the uncertainty over whether decisions will be challenged. The visible invoice may stay the same; the registry relationship becomes more expensive.

The surcharge falls unevenly. A large organisation can monitor proceedings, speak through industry groups, retain lawyers, and plan for alternative scenarios. A small ISP may simply need to know whether its tickets will be processed and whether its status remains reliable. It has little influence over powers of attorney, nomination committees, Mauritian company-law questions or ICANN letters. Yet it bears the same institutional uncertainty. If counterparties question AFRINIC's stability, the small ISP's contracts may suffer first because it has less market power to reassure customers.

Receivership can be stabilising, and in AFRINIC's case it appears to have preserved continuity that might otherwise have deteriorated further. The NRO welcomed the appointment of a receiver as a path to restoring functional governance, board elections and a chief executive. The Internet Governance Project characterised receivership as evidence that private contract-based governance has remediation mechanisms. Those are useful observations. But receivership is still emergency governance. It is not a normal way for a monopoly ledger to be run.

Emergency governance also raises the incidence of authority uncertainty. Who may approve budgets? Who may classify legal spending? Who may decide which services are essential? Who may speak for the members? Who may change bylaws? Who may accept or reject voting credentials? Who may settle cases? If the answer is unclear or contested, members bear the uncertainty. The cost appears in slower decisions and more cautious staff behaviour. It also appears in legal risk: decisions taken during abnormal governance may be challenged later, which can make transactions and long-term planning more expensive.

The continuity firewall is the institutional answer. Essential registry functions should be protected from corporate-control disputes. AFRINIC's records, reverse DNS, RDAP, WHOIS, IRR, RPKI, support and security operations should continue even when elections are contested. Funds needed for those functions should be ring-fenced. Authority for emergency actions should be documented and limited. Temporary officers should be able to maintain services but not use crisis as an excuse for broad mandate expansion. Members should know which decisions are maintenance and which are policy or strategy.

This firewall matters because continuity rhetoric can be abused. A registry in crisis can ask members to support almost any spending by saying the institution must survive. Some of that spending may be necessary. Some may protect the current leadership, litigation posture or factional preference. Without classification, members cannot tell. The continuity surcharge then becomes a blank cheque.

AFRINIC's 2026 recovery signals show both promise and risk. The Register reported that AFRINIC expected to deliver a budget and action plan and that interim management appointments had improved staff morale. It also reported that AFRINIC still faced litigation and procedural roadblocks. A budget after crisis is necessary. But the key question is not whether a budget exists. It is whether the budget distinguishes ordinary registry cost, continuity repair, legal exposure, election recovery, reserve rebuilding, training, research and discretionary programmes. Without that distinction, the surcharge remains hidden in the general fee base.

Members should not object to paying for continuity. They should object to paying for continuity without a firewall. The African registry function must survive institutional stress. But survival should mean the ledger remains reliable, not that every exceptional cost becomes compulsory and unexamined. If AFRINIC can separate the two, its fees regain legitimacy. If it cannot, continuity itself becomes regressive.

Legal-budget moral hazard

Legal budgets are the least visible and most distributive part of the AFRINIC fee problem. They decide how much member money is spent on defending the registry, pursuing claims, resisting claims, answering courts, managing elections, interpreting bylaws and shaping the boundary between recordkeeping and control. In a crisis, legal advice is unavoidable. But legal spending can also become a second policy process, funded by members who do not share the same interests.

The moral hazard starts with the registry. If AFRINIC can threaten severe remedies while members broadly fund the legal defence and the institution's liability is limited or uncertain, management may underprice the cost of aggressive enforcement. A letter that threatens resource withdrawal can provoke a lawsuit whose cost spreads across the membership. A policy interpretation that seems righteous in a meeting room can become a bank-account freeze, a damages claim, a court order or multi-year paralysis. If the decision-makers do not personally bear the external cost, they may fight too readily.

The opposite moral hazard starts with large members. A well-resourced holder can litigate in ways that impose costs on the registry and therefore on all other members. If litigation can freeze operations, delay elections, consume staff and force external intervention, it becomes a bargaining tool beyond the merits of the underlying claim. Smaller members then subsidise a private dispute through institutional disruption. They may dislike the registry's enforcement posture and still resent being forced to pay for litigation that paralyses their own service provider.

AFRINIC's public record shows both risks. The 2021 dispute began from AFRINIC's attempt to act against Cloud Innovation's resources. Cloud Innovation's response included multiple legal actions and a bank-account freeze. Later, AFRINIC described a continuing web of litigation and procedural roadblocks. Cloud Innovation and associated voices framed the registry as a gatekeeper exercising high-consequence power without matching liability. Each side accuses the other of imposing external costs. Fee incidence asks how to keep those costs from landing indiscriminately on everyone else.

The first discipline is proportional remedy. A registry should have tools for inaccurate records, false documents, fraud, unpaid fees, failure to maintain contacts and policy breaches. But the remedy should match the harm and the evidence. A documentation request is not a revocation. A dispute flag is not termination. A supervised cure period is not business destruction. A restriction on future allocations is not withdrawal of existing operational resources. The more existential the remedy, the stronger the process, authority and liability awareness must be.

The second discipline is independent review. Before a resource dispute becomes a court fight, members should have access to a credible review path that can examine evidence, policy interpretation, proportionality and interim continuity. This does not require the registry to surrender authority. It requires the registry to make the official path safer and cheaper than emergency litigation. If members believe the only effective appeal is court, they will use court. If court is the first effective check, legal costs will remain a tax on the whole membership.

The third discipline is category disclosure. AFRINIC should not have to publish privileged advice, but it should publish enough for members to understand legal incidence. How much is spent on defending core registry services? How much on the Cloud Innovation dispute or related matters? How much on elections? How much on bylaws? How much on corporate-law compliance? How much on external interventions? How much on debt collection? How much on staff or employment matters? These categories tell members whether fees are funding the narrow ledger or the institution's contested ambitions.

The fourth discipline is authority. Legal spending should be tied to documented approval. During boardlessness or receivership, this is especially important. Members should know which body authorised major legal strategies, under what power, and with what budget ceiling. Emergency authority may be necessary, but emergency authority should not become permanent discretion. If the registry lacks clear board approval, the moral hazard grows because no ordinary member-control mechanism restrains the spend.

The fifth discipline is settlement logic. A registry should not settle every claim to avoid cost. Some claims threaten the ledger and must be resisted. But members deserve to know the principles that guide settlement: continuity first, proportionality, preservation of verified records, no private capture, no secret policy changes, and no use of security services as leverage. Without settlement logic, legal spend becomes a reputational contest in which each side fears appearing weak.

Legal-budget moral hazard is regressive because legal intensity is not evenly available. The strongest parties can create or survive it. The weakest pay for it indirectly. A small ISP cannot meaningfully influence a lawsuit over millions of addresses, but it can lose training, support quality or fee stability because the registry is fighting. A university cannot decide AFRINIC's court posture, but it can suffer from a registry whose management time is consumed by litigation. A public-sector network cannot force a settlement, but it can become part of the regional risk story.

If AFRINIC wants its fees to be seen as legitimate after crisis, it must make legal spending boring. Not small, necessarily. Boring: classified, authorised, proportionate, tied to continuity and reported in a form members can use. The opposite of legal-budget moral hazard is not legal weakness. It is legal accountability.

Mandate firewall and the limits of compulsory money

AFRINIC's mission language is broad. It speaks of serving the African internet community, supporting development, fostering self-governance, distributing and managing resources, improving services, training, stakeholder engagement and financial sustainability. Many of these activities are valuable. Africa's internet ecosystem does need training, routing security, local expertise, internet exchange growth, public-sector capacity and community coordination. The question is not whether those goals are good. It is which goals may be funded compulsorily through resource-recognition fees.

The mandate firewall separates the answer. Mandatory fees should fund the functions members cannot avoid and that no one else can perform in the same authoritative way: unique number-resource records, verified holder authority, public registration data, reverse DNS, RDAP and WHOIS, IRR, RPKI, transfer recording, dispute marking, security operations, member support, audits, backups and continuity. These functions make AFRINIC a settlement utility. They justify captive funding because all recognised holders depend on them.

Other activities may be useful but should be separately justified. Training can be valuable, but it does not have the same compulsory character as maintaining the registry record. Research can be useful, but a small network should not have to fund every research ambition to keep its resources recognised. Meetings can support governance, but travel-heavy institutional life should not be bundled invisibly into the same fee as reverse DNS. Advocacy and public positioning may sometimes protect the institution, but they should not be allowed to consume compulsory revenue without clear limits. Development programmes can benefit the region, but their costs and beneficiaries should be transparent.

This distinction is not an argument for an impoverished AFRINIC. A thin registry can still be well funded where funding is tied to critical services. The point is to prevent compulsory money from becoming mandate fuel. When an organisation controls a captive ledger, broad mission language can turn into a blank cheque. Every programme becomes community, every lawsuit becomes continuity, every enforcement theory becomes integrity, every external alliance becomes regional development. The invoice loses meaning because members cannot tell which function they are paying for.

AFRINIC's crisis makes the firewall more urgent. A stable registry may get away with some budgetary fuzziness because trust is high and costs are ordinary. A registry emerging from litigation, receivership and election controversy cannot. Members are entitled to ask whether fees are paying for core operations, legal recovery, reserve rebuilding, capacity programmes, governance repair, communications, or discretionary policy fights. The answer should not require forensic investigation.

The mandate firewall also protects against regressivity. Core ledger services benefit small and large members alike, though in different ways. Non-core programmes may have uneven benefits. A large incumbent may benefit from policy meetings and continental initiatives. A small ISP may need only reliable records and support. A government network may benefit from training. A leasing-oriented holder may benefit mainly from transfer finality. A university may value community programmes but not litigation. A compulsory fee that mixes all of these without classification forces members to subsidise programmes according to institutional preference rather than demonstrated common need.

Fee discipline would therefore separate budgets by purpose. A core registry budget would cover the narrow ledger. A continuity reserve would be defined in months of core operating cost. A legal reserve would be separately governed and reported by category. Capacity-building would have its own budget, outcomes and possible sponsors. Meetings and community programmes would disclose cost and purpose. Extraordinary litigation or institutional recovery would be reported as extraordinary, not hidden inside ordinary operations. Members could then debate cross-subsidy openly.

The firewall is especially important under IPv4 scarcity. Scarcity tempts registries to expand their role from recordkeeper to economic planner. A registry may believe it is protecting regional development by limiting out-of-region use or discouraging leasing. It may believe it is preventing arbitrage. It may believe it is protecting future entrants. Those concerns are not frivolous. But if they are pursued through compulsory fees and discretionary enforcement, they become quasi-tax policy. Members who disagree cannot exit. The costs fall unevenly.

A mandate firewall requires AFRINIC to ask a disciplined question before spending: is this necessary to preserve the authoritative ledger, or is it a broader objective that should be separately approved, funded or limited? If the answer is ledger, mandatory fees are appropriate. If the answer is broader objective, the burden of justification rises. Members should not have to fund economic planning merely to maintain resource recognition.

This also clarifies official statements. AFRINIC, ICANN, the NRO and other official bodies may describe continuity, coordination and community stewardship. Their statements can establish facts about roles, dates and institutional positions. They should not substitute for incidence analysis. The economics of the fee depends on the cost, payer, beneficiary, authority and alternatives. A compulsory fee is legitimate not because it is wrapped in community vocabulary, but because it funds a function that captive members reasonably need and can inspect.

The future of AFRINIC's fee legitimacy depends on this distinction. If members see a narrow, auditable, settlement utility, they will tolerate fees because the service is necessary. If they see a broad gatekeeper funding its own discretion, the same fee becomes a regressive levy.

The market price of uncertainty

Not all fee incidence passes through an invoice. Some of it passes through prices that never mention AFRINIC. A buyer discounts an AFRINIC-administered IPv4 block because transfer approval is uncertain. A lender discounts a data-centre operator because address continuity depends on a registry in litigation. A customer demands a stronger indemnity because reverse DNS or registration status could become disputed. A seller accepts a lower price because legacy status changes on transfer. A small ISP delays expansion because it cannot be sure when resources will be available. These are registry costs expressed through market behaviour.

IPv4 scarcity makes those costs visible. When addresses were abundant, administrative friction was annoying but often survivable. A network could apply for more. A delay might slow a project but not change an asset's value dramatically. In the scarcity era, a block's registry status, transferability, dispute history and jurisdictional risk are part of its price. AFRINIC's fee schedule does not capture this, but AFRINIC's institutional quality affects it.

The transfer policy's need-justification requirement is one example. A recipient must justify need before AFRINIC. That may serve conservation, but it also means that a transaction is not complete merely because buyer and seller agree. The registry's evaluation becomes a condition. If evaluation is predictable, the cost is manageable. If it is slow, discretionary or affected by litigation, the cost rises. Sellers may prefer other regions. Buyers may demand discounts. Brokers may price AFRINIC space differently. Small operators that genuinely need space may find the process harder than better-resourced applicants.

The dispute condition is another example. A source must not be involved in dispute over the status of resources. That is sensible if the goal is to avoid transferring contested records. But in a region with high litigation risk, the definition of dispute can become economically powerful. A claim, objection or registry investigation may freeze liquidity. A large party can survive illiquidity. A small holder needing to finance network upgrades by selling unused space may not. The incidence of dispute rules is therefore unequal.

Legacy-status conversion is a third example. AFRINIC's policy states that transferred IPv4 legacy resources will no longer be regarded as legacy resources. This may align transferred resources with current policy. It also changes the economics of a transfer. A holder with legacy status may hesitate to transfer if the transaction reduces status. A buyer may discount if future policy constraints increase. The registry may not charge a transfer fee, but the status change can be a larger economic cost than any fee would be.

Legal and governance uncertainty magnify all three. A transfer rule administered by a stable board, clear staff authority and reliable appeals is one thing. The same rule administered after boardlessness, receivership, annulled elections and continuing litigation is another. The market prices the institution as well as the rule. AFRINIC-administered resources can carry a risk premium not because the addresses route poorly, but because the recognition layer has been contested.

This premium is regressive. Large holders can wait for better conditions, diversify across regions, negotiate legal protections or absorb a lower sale price. Small holders may need liquidity now. New entrants may lack alternatives. Local ISPs may be forced into NAT, customer rationing or expensive third-party arrangements. Customers in lower-income markets ultimately bear the cost through higher prices, poorer service or slower deployment. The registry's uncertainty becomes a network-development tax.

The same dynamic affects leasing. Whether one approves of IPv4 leasing or not, it exists because IPv4 demand persists and supply is scarce. AFRINIC and Cloud Innovation have disputed representations around leasing, court orders and commercialisation. For incidence, the important point is that market actors cared about perceived recognition. If the registry's position is unclear, or if litigation is used to settle communication disputes, all market participants face higher diligence costs. Those costs again fall hardest on operators with less legal capacity.

The market price of uncertainty also affects members who never buy, sell or lease addresses. A small ISP's customers may ask whether its address space is stable. An upstream may impose more documentation. A bank may question collateral or revenue assumptions. A public project may face procurement concerns. The fact that the member has no role in the Cloud Innovation dispute does not protect it from the region's risk premium.

This is why fee reform cannot be limited to reducing nominal charges. A lower invoice is useful, but a low fee attached to high uncertainty can still be expensive. AFRINIC's most important fee reduction would be institutional: make registry actions narrow, timely, auditable and legally bounded. In a settlement utility, predictability is a form of price cut. It reduces the hidden tax that markets impose when the ledger is uncertain.

What a fair fee bargain would require

A fair AFRINIC fee bargain would not begin by asking how much money the registry would like to raise. It would begin by identifying the essential, non-substitutable functions captive members must fund. Then it would allocate costs in a way that recognises unequal ability to absorb fixed charges, legal shocks and administrative friction. The goal is not the cheapest possible AFRINIC, but a defensible incidence.

The first requirement is a core registry budget. AFRINIC should define the annual cost of maintaining the narrow ledger: accurate resource records, member authority, RDAP and WHOIS, reverse DNS, IRR, RPKI, transfer recording, dispute marking, support, security operations, audits, backups, data escrow and disaster recovery. These costs are the strongest claim on compulsory fees. They should be disclosed separately from programmes whose benefits are broader or more discretionary.

The second requirement is a continuity reserve tied to that core budget. A reserve target expressed in months of core registry operation would tell members what they are insuring. It should not be a vague pool available for any crisis. Drawdowns for payroll, systems, security and critical services are different from drawdowns for strategic litigation, public campaigns or extraordinary governance initiatives. If the reserve is used for non-core purposes, members should see the category and replenishment plan.

The third requirement is legal-budget classification. AFRINIC's legal costs should be reported by category, not buried in a single line. Members need to know the scale of spending on core continuity, resource disputes, corporate governance, elections, bylaws, regulatory or ICANN matters, employment, debt collection and extraordinary litigation. This would not waive privilege. It would make incidence visible. A member can tolerate legal spending more easily when it can see what risk it is funding.

The fourth requirement is proportional enforcement. Fee arrears should have consequences, but the path should protect live network continuity where possible. Late penalties, payment plans, notice, cure periods, suspension of non-essential services and independent review can maintain discipline without turning billing into existential leverage. Good standing may be necessary for some actions, such as transfers, but the effect should be predictable and not used as a discretionary weapon.

The fifth requirement is small-operator sensitivity. AFRINIC already has discounts for academic and critical infrastructure categories. The same principle should be extended to the design of deadlines, payment options, documentation demands and support. Quarterly billing exists with a service charge; in markets with foreign-exchange stress, payment-plan design may matter more than small nominal discounts. The registry should ask not only whether the fee is low in global terms, but whether the payment mechanism is realistic for local operators.

The sixth requirement is transfer neutrality. Transfer-related charges and procedures should recover reasonable processing cost and preserve record integrity. They should not become a hidden tax on liquidity or a tool for broad economic planning. Need checks, good-standing rules and legacy-status changes should be justified in terms of clear policy goals and measured against their effect on market access for smaller networks. If a rule makes legitimate transfers difficult, the cost should be acknowledged rather than hidden behind conservation language.

The seventh requirement is auditability of record repair. After the reported address-record corruption, members have a legitimate interest in stronger controls. Mandatory fees can fund those controls. But the controls should be specific: access logs, dual approval, dormant-resource review, authority verification, member notification, whistleblowing, independent audit and transparent remediation categories. That is different from funding broad campaigns under the label of cleanup.

The eighth requirement is election-cost containment through better design. Re-running elections, investigating powers of attorney and defending legitimacy in court are expensive. The cheapest election is not the least supervised one; it is the one whose credentials, proxies, voter authority and counting process are robust enough not to collapse. Spending on election integrity is a legitimate continuity cost. Spending repeatedly because prior processes were unclear or vulnerable is a surcharge members should not have to normalise.

The ninth requirement is separation of development programmes. Training, research, fellowships and community engagement may be valuable, but they should have transparent budgets and outcomes. Where possible they should attract sponsorship, grants or opt-in support. If compulsory fees support them, the registry should show why the benefit is common enough to justify the levy. This protects small operators from paying for institutional ambitions that do not reduce their registry risk.

The tenth requirement is a scarcity-era stress test. Every new fee, penalty, transfer condition or enforcement rule should ask whether it falls as a fixed cost on small networks, turns legal uncertainty into a member-funded levy, taxes liquidity without improving record integrity, or encourages the registry to behave like an economic planner rather than a settlement utility. If so, the burden should be explicit.

This bargain would make fees neither purely equal nor purely market-based. It would combine tiering, discounts, transparency, purpose classification and process discipline. It would recognise that a small member's burden is not only the invoice, that a large holder's low fee relative to address value creates its own fairness question, and that the registry should not confiscate scarcity value merely because it records scarce resources.

The essential trade is simple: members pay compulsory fees; AFRINIC keeps compulsory functions narrow, reliable and auditable. That is the bargain a settlement utility owes its captive users.

Scarcity-era discipline for a rebuilt AFRINIC

AFRINIC's recovery cannot be judged only by whether it has a board, a budget or a strategy document. Those are necessary signs of corporate life. The harder test is whether the rebuilt institution lowers the hidden cost of being an AFRINIC member. If members still price the registry as a litigation platform, a discretionary gatekeeper or a fragile corporate shell, nominal fee reform will not solve regressivity.

Scarcity-era discipline begins with vocabulary. AFRINIC should be careful not to describe every institutional interest as community protection. It should describe functions precisely. Record accuracy is record accuracy. RPKI continuity is RPKI continuity. Transfer review is transfer review. Litigation is litigation. Training is training. Election repair is election repair. When categories are named precisely, costs can be allocated more honestly. When everything is community, everything can be charged to everyone.

The next discipline is humility about IPv4. AFRINIC can promote IPv6 and should support the region's long-term transition. But the continued economic value of IPv4 is not abolished by saying addresses are not property in the traditional sense. Many valuable rights are conditional, administrative or licence-like. They still support revenue, contracts and investment. A registry that affects recognition of such resources must behave with due process even if it rejects property language. The fee payer is funding an institution whose actions can affect real economic reliance.

A third discipline is non-confiscatory funding. AFRINIC should not attempt to solve scarcity by extracting market value from holders through annual charges. That would punish networks for holding resources and could turn the registry into the very gatekeeper critics fear. At the same time, it should not underfund the controls needed to protect the ledger. The balance is cost-based core funding plus transparent extraordinary funding where necessary. The registry should recover the cost of reliable administration, not rent from the value of the resources.

A fourth discipline is service continuity as the first claim on funds. AFRINIC's staff appear to have maintained essential services through difficult conditions. That operational resilience should be formalised. Core systems, security keys, publication services, support queues and record integrity should have protected budgets and contingency plans. Legal and governance crises should be built around preserving those services, not allowed to consume them. Members should see service metrics because the value of the fee depends on service quality.

A fifth discipline is faster uncertainty resolution. Transfer requests, resource reviews, billing disputes, authority changes and election challenges should move through defined timelines. Delay is a cost. Under scarcity, delay is a price signal. A member waiting for a decision may lose customers, financing or transaction value. Timelines should be part of fee accountability because members are paying for the administration of a settlement layer.

A sixth discipline is restraint in official communication. AFRINIC's public statements during disputes can affect markets. A statement about leasing, court recognition, member status, election validity or resource misuse can move counterparties' behaviour. The registry should separate factual notices from legal argument and policy preference. It should state uncertainty where uncertainty exists. A settlement utility should not communicate like a campaign organisation.

The final discipline is to treat member fees as constrained money. The phrase member-funded should mean more than available to the institution. It should mean money raised from a captive base for defined purposes. If AFRINIC can rebuild that understanding, members may still disagree about policy, transfers, leasing, IPv6 and governance, but they will be arguing inside a system whose costs are visible and whose core service is protected.

AFRINIC does not need to become small in ambition for Africa's internet. It needs to become narrow where compulsion is involved. Ambition can be funded, sponsored, debated and measured. The ledger must be maintained. Confusing the two is what turns fees into a regressive quasi-tax.

Who really pays

The final question is not whether AFRINIC's fees are high or low in isolation. A $1,400 annual fee may be low by global enterprise standards and high for a fragile local operator. A $38,400 fee may be large as an association charge and small relative to a large IPv4 holding. A legal bill may be necessary to defend the registry and still unfair if its costs are spread without disclosure. A reserve may be prudent and still dangerous if it funds discretion rather than continuity. Incidence depends on the whole institutional bundle.

AFRINIC shows how equal-looking registry fees can become regressive in three ways. First, fixed charges and hard-currency deadlines fall on members with unequal margins, currencies, payment systems and administrative capacity. Second, the hidden costs of compliance, delay, transfer uncertainty and policy discretion are heavier for smaller or less legally sophisticated operators. Third, governance failure converts ordinary member fees into crisis financing, and the resulting burden spreads to members that did not choose the legal or institutional strategy.

This is why fee incidence belongs at the centre of the AFRINIC debate. The governance crisis is not merely about who sits on the board. The Cloud Innovation dispute is not merely about one large holder. The corruption history is not merely an old scandal. The receivership is not merely an emergency procedure. The election problems are not merely procedural embarrassment. Each event changes who pays for the registry, who bears risk from the registry, and who can afford the uncertainty created by the registry.

For the small ISP, the invoice under the generator is part of a larger squeeze. The registry fee competes with fuel, transit, tower maintenance, staff, customer support and currency conversion. If AFRINIC is a narrow, reliable settlement utility, the fee is a tolerable cost of doing business. If AFRINIC becomes a source of legal and policy uncertainty, the fee is only the cover charge to a more expensive system.

For the university or public-sector network, the fee is part of public infrastructure. The burden is not measured by profit but by what else the budget could have done: bandwidth for students, security upgrades, redundancy, research connectivity, rural public services. A registry crisis financed through member funds is ultimately financed by those foregone services.

For the data-centre operator, the fee is embedded in customer confidence. A stable registry lowers sales friction. An unstable registry raises diligence, indemnity and contract costs. Those costs move through the market even if no line item says AFRINIC.

For the large holder, the fee may be small relative to asset value, but registry discretion can be enormous. If the registry can damage a large holding without bearing matching liability, the holder will fight. If the holder can impose costs on the registry without bearing the system-wide externality, other members will pay. The path out is a narrower bargain: charge enough to run a secure, auditable, continuous registry; classify legal and crisis costs; protect essential services from institutional conflict; separate mandate from ambition; and treat scarcity as a reason for discipline, not broader discretion.

If AFRINIC does that, the invoice can become what it should have been: a charge for a shared settlement utility whose value is reliability. If it does not, the invoice remains something else: a regressive levy on unequal networks to finance the risks of a monopoly ledger in crisis.