The first obstacle facing a small internet service provider in an African secondary city is usually described in visible engineering terms. The founder needs backhaul, mast space or fibre access, power resilience, routers, customer equipment, field staff, software, vehicles, spares, and enough early subscribers to keep the network alive through its first bad months. Those costs are real. In many towns they are decisive. But they are not the whole entry problem.
The less visible input is credible number-resource evidence. Before the founder has a stable subscriber base, before the support desk has learned its recurring complaints, and before the business has a reputation with banks or upstream carriers, the network must still prove that it can identify customers, announce reachable prefixes, satisfy routing filters, publish usable contacts, answer abuse mail, pass procurement checks, and explain where its IPv4 capacity will come from. It must do this in a region where AFRINIC has operated under IPv4 exhaustion Phase 2, where allocations are small and need-based, where leases and transfers carry their own uncertainty, and where the registry itself has spent years under legal and governance stress.
That changes the economics of entry. A small ISP can start modestly in towers, staff and customer count. It cannot start indefinitely with a vague address plan. Private IPv4 sharing can postpone some demand, and IPv6 can solve part of the future, but many customer applications, enterprise firewalls, payment platforms, public-sector portals, abuse desks, geolocation systems and upstream filters still treat IPv4 identity as part of the service. An entrant that cannot present a credible address story is not merely short of a technical input. It is harder to finance, harder to interconnect, harder to sell to business customers, harder to join tenders, and easier for incumbents to dismiss as a risky reseller rather than a durable access network.
The question is not whether AFRINIC alone decides who may become an ISP. It does not. Markets, regulators, tower owners, energy suppliers, municipalities, incumbents, banks, vendors and customers all matter. The question is narrower and more important: when IPv4 scarcity and registry uncertainty make address evidence expensive, how much extra scale must a small entrant reach before it can be trusted? If the answer keeps rising, the registry layer becomes part of the minimum efficient scale of the access business. That is an entry barrier, even when no rule formally says that small firms may not enter.
Entry starts with credible address evidence
Imagine a founder trying to build a fixed wireless network outside a capital city. The plan is ordinary: rent space on a few sites, buy upstream transit, connect schools and small businesses first, then add households as coverage improves. The spreadsheet begins with radios, poles, batteries, routers and a modest field team. The sales plan assumes that early enterprise customers will subsidise the slower residential build-out. The investor asks about churn, power, competition and the cost of imported equipment. Then the upstream carrier asks about the address plan.
At that moment the business plan changes shape. The founder may have enough private addressing and carrier-grade NAT for residential customers at the beginning. Yet the first business customers may ask for static public IPv4, inbound reachability, clean geolocation, working VPNs, payment-service access, stable reverse DNS and a support desk able to trace abuse reports to the right subscriber. A school may need filtered services to recognise its location. A municipal office may need address evidence that procurement or security staff can document. A local cloud, hosting or managed-services offer will require far more public identity per unit of revenue than a basic home broadband plan. Even when the founder intends to be efficient, the market asks for public evidence before revenue is stable.
The difficulty is not only acquiring addresses. It is proving that the addresses will remain usable. Upstreams want to know whether the prefix is recognised by the relevant registry, whether the origin AS is credible, whether routing-security data exists, whether contact details are current, whether abuse reports will be answered, and whether the address history is likely to create trouble. Banks and investors may not know the details of RDAP, WHOIS, RPKI or routing-registry data, but they understand a more basic question: can the company continue to deliver the service that produces the cash flow? If the answer depends on a contested lease, a delayed allocation, an unclear transfer or a fragile registry relationship, the business looks riskier.
AFRINIC's public materials show why this evidence matters. The institution is a nonprofit, member-based registry registered in Mauritius and serving Africa and the Indian Ocean region. It distributes and manages IPv4, IPv6 and ASNs and provides registry services that networks rely on. Its exhaustion materials record that Phase 2 began in January 2020 and that, in this phase, IPv4 requests are evaluated under policy, handled through tickets, and limited to a minimum of /24 and a maximum of /22 per allocation or assignment. Its policy manual requires documented need, efficient use and registration accuracy. These are not mere formalities for an entrant. They define the paperwork through which a small network becomes legible to the rest of the market.
The larger incumbent starts with a different credibility stock. It already has addresses, staff, records, customer history, legal files, routing relationships and established procurement language. It can promise customers stability because it has spare capacity and institutional memory. The entrant must build credibility while also building the network. Address evidence becomes the bridge between engineering ambition and market trust. If that bridge is narrow, the first customers become harder to win, the first loan becomes more expensive, and the first upstream negotiation begins from weakness.
Scarcity makes documentation a scale threshold
Economists usually describe entry barriers as costs that incumbents do not have to bear in the same way as entrants. Scarce IPv4 turns registry documentation into exactly that kind of cost. In a world of abundant free-pool addresses, a new ISP could expect to justify need, receive enough capacity for early growth, and treat the registry as background administration. In the AFRINIC region after exhaustion, the free-pool path is small, staged and conditional. The entrant may receive a modest block if it qualifies, but the plan must already anticipate shortage.
AFRINIC's Phase 2 limits reset the arithmetic. A /24 is 256 IPv4 addresses. A /22 is 1,024. For a small rural access network, those numbers can be meaningful; for a business that hopes to grow across several towns, support business customers, offer hosting, maintain management infrastructure and avoid overloading shared-address systems, they are not much. The limit does not prevent entry outright. It forces the entrant to design around scarcity from the first day.
Designing around scarcity is not free. It means more planning, more justification, more address-sharing equipment, more monitoring, more customer segmentation, more careful sales promises, more documentation of current and projected use, and earlier engagement with leases or transfers. The fixed cost of this work does not rise in proportion to subscribers. A large operator spreads it across hundreds of thousands or millions of customers. A small entrant spreads it across the first few hundred or few thousand. The same institutional requirement therefore consumes a much larger share of early revenue.
This is why the old distinction between technical policy and market economics is no longer tenable. The policy manual speaks in the language of conservation, aggregation, registration requirements, documentation and utilisation. Those aims may be technically coherent. But once the scarce input becomes valuable, each requirement also acts as a market filter. An entrant must be large enough to justify resources, organised enough to document the case, patient enough to wait, and capitalised enough to survive if the request is smaller or slower than expected. The threshold for being taken seriously rises.
The danger is not that conservation is illegitimate. IPv4 is finite; waste matters; fraud matters; stale records matter. The danger is that scarcity management can accidentally protect incumbency. A rule that appears neutral because it applies to all members may still burden entrants more heavily because they lack archive depth, legal staff, spare inventory and bargaining power. The incumbent can show historical utilisation. The entrant can show projected demand, signed contracts, equipment orders and upstream arrangements, but the proof is necessarily more uncertain because the customers are not yet all connected.
Minimum efficient scale then shifts upward. The founder cannot simply test a town with a lean network and grow organically. To be credible, the founder may need a larger address plan, a better compliance file, a more expensive upstream arrangement, a consultant, a lawyer, a leasing contract, a CGNAT platform, and enough working capital to handle delays. Each addition may be rational on its own. Together they make entry less experimental and more capital-intensive. The region loses some of the competitive energy that comes from small firms being able to try, fail cheaply, learn quickly and expand where demand proves real.
Proof arrives before revenue
The timing of proof is the hard part. A mature operator can document existing use. A new entrant often must document intended use. That sounds like a small distinction until one remembers how early-stage ISPs are financed. Customers may not sign binding multi-year contracts until they know the network can launch. Investors may not release funds until they see a credible launch plan. Upstreams may not finalise terms until they see address and routing evidence. The registry may ask for a realistic plan before issuing resources. Each party wants someone else to move first.
This circularity is not unique to AFRINIC, but scarcity makes it sharper. The policy logic of documented need works best when need is already visible. Entry is the moment when need is becoming visible. A founder can show market research, letters of interest, radio coverage plans, a network diagram, tower leases and signed wholesale terms. Yet the most reliable proof would come after deployment, while deployment depends on the resource evidence that proof is meant to secure. The result is a pre-revenue documentation burden.
AFRINIC's policy manual recognises documentation as a core part of allocation and assignment. It refers to network engineering plans, subnetting plans, topology, routing plans, immediate use, efficient utilisation and recordkeeping. It requires local internet registries to maintain documents about assignments and sub-allocations, including requests, supporting materials, correspondence and reasons for unusual decisions. For a large operator, this is internal administration. For an entrant, it is a second product line: the production of institutional evidence.
Evidence production has a real cost. The founder must translate business demand into the vocabulary of address policy. A school, a clinic, a cash-and-carry warehouse, a small hotel, an agricultural processor and a municipal office do not arrive as neat policy categories. They arrive as customers with messy requirements. Some need static addresses. Some only think they do. Some have legacy security vendors who insist on IPv4 allowlisting. Some need remote access that could be redesigned but not quickly. Some need public identity because a procurement checklist says so. The ISP must convert these demands into a plan that is technically honest, commercially practical and acceptable to counterparties.
The proof burden also creates an asymmetry between clean entrants and ambiguous incumbents. An incumbent may have inefficient historical allocation, but it can rely on installed base and inertia. A new ISP may be more efficient, more local and more competitive, but it must prove its future with less evidence. The system can therefore reward the operator that already has resources over the one that might use marginal resources more productively.
That is a classic entry-barrier pattern. A rule designed to prevent speculative hoarding can still make it harder for genuine entrants because genuine entrants are, by definition, not yet large enough to prove everything with historical data. The answer is not to abandon proof. It is to make proof proportional to entry. A small network should be able to obtain and retain enough capacity to demonstrate demand under clear milestones without being forced to behave like a mature national carrier before it has its first stable cash flow.
Waiting time is working capital
Delay is often described administratively: a ticket is incomplete, a request is under evaluation, an invoice must be paid, a record must be corrected, a transfer must be reviewed. For an entrant, delay is financial. It is rent on towers before subscribers are connected. It is staff on payroll before revenue arrives. It is a customer installation team waiting for a prefix, a school contract at risk of moving to the incumbent, or a bank facility that cannot close because the lender does not understand the address position.
AFRINIC's exhaustion materials describe a queue based on complete applications, with incomplete applications handled case by case until all information is supplied. Earlier exhaustion practice also showed how approved prefixes could be reserved for a period while payment and the registration agreement were completed, with the prefix returning to available inventory if requirements were not met in time. Even if one treats these as reasonable scarcity procedures, they illustrate the working-capital issue. The entrant must be prepared to move when the registry moves. Money, signatures and documents must be ready. But the entrant also must wait while the case becomes complete and is evaluated.
Waiting time has option value for incumbents. A large operator with address inventory can keep selling while a request or transfer is pending. It may even benefit when smaller rivals are delayed. The entrant has less slack. It may lease expensive temporary capacity, accept upstream-assigned addresses that increase dependency, use heavier NAT than planned, or slow customer acquisition to match uncertain supply. None of these choices appears on the registry invoice, but each changes the economics of entry.
There is also a market-timing cost. IPv4 prices, lease rates and counterparty willingness can shift while an entrant waits. A seller may decide not to proceed. A lessor may change terms. An upstream may withdraw a promotional offer. A customer may choose another supplier. A small ISP cannot hedge these risks cheaply. The delay sits between capital expenditure and revenue, where young firms are most fragile.
The most damaging delays are not necessarily the longest. They are the ones without a clear clock. A bank can finance a 60-day wait if the wait is predictable and the risk is priced. A founder can plan around a known review window. What is harder to finance is discretionary silence, repeated clarification, unclear status or a policy question that may become a dispute. Uncertainty raises the cash buffer required to enter. Raising the cash buffer raises the minimum efficient scale.
This is where the governance crisis matters even when a particular ticket is handled properly. AFRINIC's prolonged court disputes, receivership period, board instability and contested elections have made the institution itself a perceived risk variable. A small entrant may not know the legal details, but counterparties read the headlines. If the registry is seen as unstable, routine waiting becomes more expensive because no one knows whether delay reflects ordinary evaluation, resource scarcity, staffing strain, litigation exposure or future policy change. The result is a registry-risk premium that entrants pay before they have enough revenue to absorb it.
Leasing is a bridge, not independence
Leasing responds to a real entry problem. A small ISP that cannot buy a block or obtain enough from the free pool can lease IPv4 capacity and match address cost to customer revenue. Leasing can be especially attractive when the founder needs to launch quickly, serve business customers, test demand or avoid tying scarce capital in a purchase. It can reduce upfront cost and allow a more gradual build-out.
But leasing does not remove the entry barrier. It changes its form. The entrant must now prove the quality of the lease. Upstreams, customers and lenders will ask who holds the block, whether the lessor can authorise routing, whether reverse DNS will work, whether abuse complaints will be handled, whether geolocation can be corrected, whether routing-security data will be maintained, whether the term is long enough for customer contracts, and what happens if the lessor is drawn into a registry or court dispute. A lease with weak evidence may be cheaper at signing and more expensive in operation.
The commercial lesson is simple: a transaction is not over when money moves. The resource must remain manageable when routed, recorded, renewed, reviewed, questioned, maintained and used. An ISP does not need addresses for a spreadsheet. It needs them for customers whose services must keep working. If a leased block becomes hard to route, hard to renew or hard to explain, the entrant suffers even if the contract was inexpensive.
Leasing also affects bargaining. A small entrant leasing from a stronger party may avoid dealing directly with the registry, but it becomes dependent on the lessor's operational discipline and legal position. The lessor may have better expertise, larger inventory and more ability to manage registry contact. That can be valuable. It can also make the entrant a price taker. If demand grows and the customers become attached to the addresses, renewal terms become more important. The entrant may have built a business on capacity it does not control.
This is not a condemnation of leasing. In a post-exhaustion market, leasing may be the most practical route for many small ISPs, particularly where buying would consume too much capital. The problem is that leasing grows partly because direct entry through the registry and transfer channels is costly, uncertain or slow. A healthy market would offer entrants several credible options: modest direct allocation, transparent transfer, clear leasing, upstream assignment with portability expectations, and IPv6-heavy designs where customers can accept them. An unhealthy market makes one workaround carry too much weight.
The policy danger is to treat leasing either as a moral failure or as a complete solution. It is neither. It is a bridge over scarcity and institutional friction. Bridges require inspection. Clear provenance, stable authorisation, visible responsibility and predictable remedies matter more for entrants than for large operators, because entrants cannot easily survive a broken bridge. If AFRINIC wants a competitive small-ISP market, it should make legitimate leasing and transfer evidence easier to distinguish from opaque or risky chains, rather than forcing every small entrant to become a private detective.
Transfers help only when they are bankable
Transfers are often presented as the mature answer to exhaustion. If the registry free pool is nearly gone, the argument runs, addresses can move from holders who value them less to operators who value them more. In principle this is pro-entry. It can bring idle or underused capacity into productive networks. It can let a small ISP buy certainty rather than depend on an upstream or a short lease. It can also give sellers a reason to release space that would otherwise remain dormant.
In practice, transfers help entrants only when the transfer path is financeable. A small ISP must be able to know the approximate price, the evidence standard, the review window, the fees, the risk of refusal, the status of the seller, the routing history and the post-transfer obligations. If any of these are unclear, the purchase becomes hard to fund. A bank or investor will not treat "we may receive a block if the registry accepts the need case and if the seller's position remains clean" the same as a predictable capital purchase.
AFRINIC's transfer policy is bounded by justified need and regional membership. Its fee materials say that in a transfer from an existing resource member to a new organisation, the recipient applies for membership and resources, with the applicable allocation fee and membership fee based on the approved resources. It also requires relevant accounts to be in good standing before a transfer is considered. These are bounded facts, not conclusions. The economic point is that each condition becomes part of the entrant's financing file.
The financing file is fragile because a transfer is not just a price per address. It is a sequence. The entrant must identify supply, negotiate terms, pay for diligence, show need, satisfy membership requirements, obtain approval, coordinate registry updates, arrange routing, correct geolocation, prepare customer assignment plans and maintain records. If the entrant is buying only enough space for a modest launch, the fixed cost per address may be high. If it buys more to reduce the fixed cost, it may fail the need case or overextend capital. Scarcity pushes toward larger purchases; need review pushes toward narrower justification. The entrant sits between them.
This creates an incumbent advantage even when transfers are allowed. A large operator can buy a block for strategic reserve, wait through review, use counsel, split transactions across affiliates, and carry the opportunity cost. A small ISP must align the transfer with actual demand. If the transfer closes too late, customers leave. If it closes too early, cash is trapped. If it closes in a smaller size than planned, the network design changes. If it fails, the founder may have spent precious funds on diligence and negotiation without obtaining usable capacity.
Transfer rules therefore need a competition test. Do they make scarce capacity move toward productive new networks, or do they make movement so procedurally expensive that only established firms can use the path confidently? A transfer regime can be anti-hoarding and still pro-entry if it sets clear milestones, explains refusal grounds, gives review clocks and treats small launch cases as legitimate. It becomes incumbent-protective when the facts that matter are broad, the timing uncertain and the remedy for disagreement unclear.
The market does not need transfers to be lawless. It needs them to be bankable. Bankable does not mean guaranteed approval. It means a serious entrant can price the risk, raise the money and know when to choose another path. A transfer market that cannot be financed by small ISPs will not solve the entry problem. It will become another market in which the firms with spare cash and specialist advice buy certainty while smaller rivals rent uncertainty.
Upstreams turn weak evidence into bargaining leverage
Entry into access markets is negotiated through counterparties. The small ISP does not merely announce itself to customers. It negotiates transit, peering, equipment credit, site access, bank facilities, municipal permissions and business contracts. Each counterparty asks a version of the same question: is this network durable enough to rely on? Address evidence is one answer.
An upstream carrier may start with routing questions. Can the entrant announce its prefix? Is the registry record current? Is the AS number properly assigned? Are routing-security materials aligned? Are abuse and technical contacts reachable? Does the prefix have reputation problems? Is there a clean letter of authorisation if the address holder differs from the operating network? In many cases these checks are routine. For an entrant, routine checks are still gates. A missing or ambiguous document can delay launch or weaken the price negotiation.
A bank asks the question differently. It may not understand every registry service, but it understands concentration risk. If the ISP's projected revenue depends on a scarce input obtained under a short lease, the lender may discount the revenue. If the network depends on upstream-provided addresses, the lender may see provider lock-in. If the ISP has applied for resources but has no clear decision timeline, the lender may hold funds back. The address plan becomes part of credit risk.
Public procurement and enterprise buying create another gate. Buyers often ask for operational resilience, security contacts, compliance statements, service levels and evidence of control over critical inputs. A small ISP may be technically capable of serving a school district, clinic network or municipal office, but procurement staff may favour an incumbent whose documents are familiar and whose address position appears settled. Weak number-resource evidence does not have to disqualify the entrant formally. It only has to add enough uncertainty for a risk-averse buyer to choose the familiar supplier.
This is how a registry-layer issue becomes a competition issue. Incumbents benefit from evidence inertia. Their records may not be perfect, but they are known. Their prefixes have history. Their upstreams have accepted them. Their banks have seen them before. Their procurement language is tested. An entrant must convert technical capability into institutional trust from scratch. If the shared registry environment is uncertain, the entrant's burden rises faster than the incumbent's.
The same mechanism affects wholesale bargaining. An entrant with no portable capacity may accept addresses from an upstream. That can speed launch, but it makes switching harder later. Customers may need to renumber if the ISP changes provider. Business clients may resist migration. The upstream knows this. The entrant's weak address position therefore becomes bargaining leverage for the supplier. The market remains formally open, but the cost of moving away from the incumbent wholesale path rises.
Competition policy often focuses on ducts, poles, spectrum and wholesale access. In modern connectivity markets, number-resource credibility deserves similar attention. It is not a substitute for physical infrastructure, but it determines whether the entrant can present itself as an independent network rather than a thin resale layer. A registry that lowers evidence costs helps entrants bargain. A registry that raises or obscures those costs strengthens the parties already in place.
Public tenders reward settled address stories
Public-sector demand can be decisive for a small ISP. A contract to connect schools, clinics, municipal offices or local government buildings can anchor the first phase of a network. It gives the entrant predictable revenue, a reason to build into underserved areas and a reference account for private customers. In many secondary cities, such an anchor is the difference between a viable local ISP and a collection of informal wireless links.
Public buyers, however, are rarely designed to evaluate nuanced number-resource risk. They work through tenders, checklists, compliance files and conservative legal advice. They ask for service levels, security contacts, operational resilience plans, ownership declarations, tax certificates, financial statements, reference customers and proof that critical inputs are under control. Address evidence enters this system indirectly. The bidder may have to show it can provide static IPv4, maintain traceability, handle abuse complaints, support public services, and avoid sudden renumbering. If the entrant's answer depends on an unresolved lease, a pending registry request or upstream-provided capacity that cannot be moved, the bid looks weaker.
This does not mean public buyers are wrong to ask. Public networks can carry sensitive traffic and support essential services. The problem is that their risk aversion tends to reward firms whose evidence is already settled. An incumbent can attach old allocations, previous tender references and familiar operational language. A small ISP may offer better local service and lower deployment cost, but its number-resource position may be more complex because it is still entering the market. The buyer may choose the safer file rather than the better local network.
The effect is cumulative. Without an anchor contract, the entrant has less cash to buy or lease addresses. Without addresses, the anchor contract is harder to win. Without the contract, the registry need case may look less concrete. Without registry evidence, the bank hesitates. The entry barrier is not one locked door. It is a set of doors that each open more easily after another has already opened.
Transparent registry practice can break part of this circle. If a small ISP can show a dated queue position, an approved allocation, a clear lease authorisation, a recognised transfer status, a working abuse contact and stable routing evidence, the public buyer has less reason to treat the address plan as a mystery. The entrant still must compete on price, quality and delivery. But it is not penalised simply because number-resource evidence is hard to explain.
This matters for development policy. Governments often say they want local connectivity firms, rural coverage and competition outside major cities. They may subsidise access projects or include local participation requirements in tenders. Those aims can be undermined if number-resource credibility remains difficult for small entrants to obtain. A procurement rule that favours the most established address story may unintentionally channel public money back to incumbents, even where a capable local network could serve the site better.
The remedy is not to force buyers to accept weak evidence. It is to make strong evidence cheaper for entrants to produce. Registry clarity, standardised letters of authorisation, bounded lease evidence, visible transfer stages and reliable public records all help. They translate the numbering layer into documents a procurement officer can understand. In markets where public anchor demand shapes private build-out, that translation is part of competition infrastructure.
CGNAT saves addresses but raises the operating floor
Carrier-grade NAT is the usual answer to limited IPv4 supply. It lets an ISP serve many customers behind fewer public addresses. For residential broadband, especially at the low end of the market, CGNAT is often unavoidable. It can make entry possible where public IPv4 per subscriber would be uneconomic. But it is not a free substitute for address capacity. It raises the operating floor.
A small entrant using CGNAT must buy or build translation capacity, log enough information to handle abuse and lawful requests, manage port exhaustion, segment customers, monitor application failures, explain limitations to subscribers, and maintain support scripts for problems that look like application faults but originate in shared addressing. Gaming, VPNs, remote cameras, small-business servers, some payment systems and geolocation-sensitive services can all produce complaints. The customer does not call to discuss IPv4 scarcity. The customer calls because something does not work.
For a large operator, CGNAT is a platform. For a small ISP, it can be an early capital expense and an expertise requirement. The equipment may be cheaper than buying a large block, but it still requires money, staff time and operational discipline. Logging infrastructure is not optional if abuse reports or customer disputes must be answered. Poor CGNAT practice can damage reputation, attract upstream scrutiny or make business services harder to sell.
CGNAT also segments the market. An ISP can serve basic residential customers with shared public identity, but business customers often want more. A small hotel may need remote access. A local software firm may need stable endpoints. A clinic may have a vendor that still relies on IPv4 allowlists. A public office may need externally reachable services or audit trails that shared addressing complicates. If the entrant cannot offer public IPv4 where it matters, it may be trapped in lower-margin residential segments while incumbents keep the lucrative enterprise accounts.
IPv6 helps, but it does not erase the entry problem. New networks should deploy IPv6 well from the beginning. It reduces future pressure and improves long-term scalability. Yet many counterparties still require IPv4 reachability, and many customer environments remain dual-stack at best. A small ISP cannot tell a bank, school, supplier or upstream to complete the world's transition before the launch date. The entrant must operate in the market as it exists.
CGNAT therefore changes the capital mix rather than removing the barrier. The founder can buy fewer public addresses, but must buy more operational complexity. The larger the NAT estate, the more the entrant needs monitoring, logs, support discipline and customer segmentation. The smaller the public pool, the more carefully it must be rationed to customers who justify it. This is a management system, not a trick.
The danger is that policy debates treat CGNAT as proof that small ISPs do not need address certainty. The opposite is closer to the truth. CGNAT makes scarce IPv4 usable for mass access, but it also makes the remaining public IPv4 pool more strategic. Those addresses become the premium service layer for enterprise, infrastructure, management, hosting and customers who cannot tolerate shared identity. Entry barriers rise when the entrant cannot obtain that premium layer predictably.
Compliance fixed costs are regressive at small scale
Small ISP entry is not only a question of paying for addresses, equipment and transit. It is also a question of paying for compliance capacity before the business has enough subscribers to spread that cost. Registry membership forms, corporate identity evidence, assignment records, abuse contacts, routing-security materials, letters of authorisation, customer traceability, lease evidence, transfer diligence, tax documents and procurement certificates are each understandable. Together they create a fixed-cost stack.
The stack is regressive because much of it is indivisible. A network either has a competent person who understands registry interaction or it does not. It either has usable assignment records or it does not. It either has a logging architecture that can answer complaints or it does not. It either can produce documents that a bank, upstream or public buyer recognises or it cannot. A large operator may have a legal department, a regulatory team, a network engineering group and a procurement office. A small entrant may have two founders, one senior engineer, a bookkeeper and a field crew.
This changes the point at which entry becomes rational. A lean founder might be willing to start with one town, a few hundred customers and a modest capital base if the only problem were radios and backhaul. Once the evidence stack is included, that same plan may look too small. The business may need more customers simply to justify a compliance person, more financing simply to carry longer review periods, more inventory to satisfy business customers, and more legal advice to handle leases or transfers. The minimum efficient scale rises before the first customer has chosen the service.
Fixed compliance costs also affect product design. The entrant may avoid services that require public IPv4 because they complicate documentation. It may decline small business customers who need inbound reachability, even though those customers would improve margins. It may choose an upstream assignment because it is administratively easier, even though it weakens independence. It may delay participation in public tenders because the file is not yet polished. These are not abstract opportunity costs. They are the practical ways in which a small network's growth path narrows.
There is a further reputational burden. A young ISP must explain why its address plan is not a sign of weakness. It must tell customers that CGNAT is normal for residential broadband, that static public IPv4 is available only under defined conditions, that geolocation corrections can take time, that some addresses are leased, or that a transfer is pending. These explanations may be truthful and professionally managed. They still make the entrant sound more complicated than the incumbent, whose historical address position often needs no explanation at all.
The policy lesson is not that compliance should disappear. Fraud, abandoned records and unverifiable authority are real harms. The lesson is that compliance should be scaled and timed to the risks that matter. A small entrant should face high standards where a false claim would harm the routing system or other users. It should not be forced to carry mature-carrier documentation costs for every modest launch case. Proportionality is not leniency. It is a way of preserving competition while keeping the registry reliable.
Incumbents absorb what entrants must explain
Every market has a gap between the cost of compliance and the cost of explaining compliance. Incumbents often pay the first. Entrants pay both. They must comply, then persuade counterparties that their compliance is good enough. In the AFRINIC region, scarcity and institutional volatility widen this gap.
An incumbent with existing allocations can tell a buyer that the address plan is established. It may have legacy inefficiencies, but those inefficiencies are rarely visible to the customer. It can use public addresses for business clients, absorb abuse complaints with specialised staff, maintain long-standing upstream relationships, and spread registry fees and legal review over a broad base. It can also hold inventory for future projects. Inventory becomes an entry weapon when rivals cannot obtain comparable capacity.
The entrant must explain scarcity as a design constraint without making itself sound weak. It must tell business customers which services can receive public IPv4 and which cannot. It must explain why some products are behind CGNAT, why geolocation corrections may take time, why reverse DNS depends on another party, why a lease renewal matters, or why upstream-provided addresses could change if wholesale terms change. Each explanation is honest. Each also creates an opportunity for the incumbent to look safer.
This asymmetry matters most in secondary cities and rural edge markets, where the business case is already thin. The entrant may be the only party willing to serve a neglected area, but it still faces the address and evidence burden of a global internet. A tower in a remote town may require the same routing credibility as a rack in a capital-city data centre. The revenue per customer may be lower, the cost of power higher, and the staff pool thinner. Fixed compliance work therefore becomes even more regressive.
Incumbents also have political and procedural advantages. They know the registry vocabulary, policy history and meeting culture. They may have staff who can follow mailing lists, participate in discussions and anticipate changes. Small entrants are busy deploying. If they join governance only after a rule has hurt them, they are late. If they do not join, others shape the conditions under which they enter. This is not necessarily malicious; it is how low-participation institutions drift toward the organised.
The result is a market in which neutral scarcity rules can have non-neutral competitive effects. A cap on allocation size, a high proof standard, a slow transfer review, an ambiguous leasing posture or an unstable appeals path may apply to all, yet incumbents can absorb it more easily. The entrant experiences the rule at the most fragile point in its life cycle.
This is why entry analysis should not ask only whether AFRINIC treats applicants identically. It should ask whether the total system lets a capable small ISP reach independent operating status at reasonable cost. Formal equality can still preserve market concentration if the cost of satisfying the same rule falls unequally across firm size. Competition is not served by making the first step so demanding that only firms with incumbent-like resources can take it.
Institutional volatility becomes an entry-risk premium
AFRINIC's governance crisis is not the centre of this article, but it is part of the entry environment. A registry can recover operationally and still leave a residue of risk in the market. Entrants face that residue directly because they must persuade counterparties during the recovery, not after history has forgotten the dispute.
The public record includes litigation involving Cloud Innovation, court orders affecting AFRINIC's funds, a period without normal board continuity, receivership, contested election activity, concerns about proxies and member registers, and later attempts to restore governance. Some claims in that history remain contested, and the legal merits should not be simplified into slogans. The economic effect is easier to state: counterparties learned to associate AFRINIC-administered resources with institutional uncertainty. Once learned, that association does not disappear quickly.
Risk premiums can be subtle. A lessor may ask for stricter terms. A seller may hesitate. An upstream may require more documentation. A lender may discount the business plan. A customer may choose the incumbent. A consultant may advise extra caution. None of these decisions requires a final judgment about the litigation. Markets price uncertainty before courts settle it.
The market mechanism is broader than any one dispute. Once IPv4 became scarce, transferable and financeable, registries moved from low-stakes coordination to high-consequence recognition. If a registry record is critical to revenue, but remedies for registry failure are weak or slow, entrants must insure themselves through extra cash, extra documentation and extra caution. Those costs raise the threshold for entry.
Institutional volatility also magnifies policy uncertainty. During stable periods, operators can live with imperfect rules because practice becomes predictable. During unstable periods, the same text can look more discretionary. Will a future board revisit leasing? Will transfer interpretation change? Will regional-use debates become stricter? Will registry services remain funded and staffed? Will court orders affect pending matters? Even if the most likely answer is benign, the entrant must plan for variance.
This is a particular problem for small ISPs because their first years are already full of variance. Subscriber uptake may be slower than expected. Power costs may rise. Imported equipment may be delayed. A tower landlord may change terms. A competitor may cut prices. Adding registry-risk variance to this stack makes the financing problem harder. Investors do not need to understand the whole RIR system to ask for a higher return or smaller exposure.
The best repair is not better public relations. It is boring predictability. Transparent queues, clear status labels, published service targets, audited decisions, current records, reliable routing-security services and appeal clocks do more to lower the entry-risk premium than speeches about institutional importance. Small entrants do not need a heroic registry. They need one whose ordinary outputs are trusted enough that counterparties stop adding a surcharge.
The gate should remain a ledger
A registry should maintain a ledger of uniqueness, not become a licensing authority for business models. The distinction is crucial for entry. If AFRINIC verifies that a resource is uniquely registered, that the holder is legitimate, that need or transfer criteria are met, that contacts are current and that fraud is addressed, it supports competition. If it begins to judge whether a small ISP's business model, customer mix, leasing arrangement, geography or commercial evolution deserves approval beyond narrow evidence standards, it risks becoming an entry gatekeeper.
The danger is not always intentional. Institutions under scarcity pressure tend to expand the meaning of stewardship. A registry may fear waste, speculation, fraud, out-of-region use, abuse or political criticism. It may respond by asking for more information, reserving more discretion and slowing more cases. Each step can be defensible in isolation. Together they can move the registry from recordkeeper to market supervisor.
For entrants, this is especially costly. Mature firms can hire people to manage a supervisor. Start-ups cannot. A small ISP does not have the capacity to litigate policy philosophy while building towers and acquiring customers. It needs to know what facts are required, how those facts will be assessed, when a decision will be made, what reasons will be given, and how a mistake can be appealed. Without that, discretionary review becomes a business risk no founder can fully control.
The competition risk is that entry becomes permissioned in practice. Not formally, perhaps; no document may say that AFRINIC licenses ISPs. But if credible number-resource evidence becomes indispensable to upstreams, lenders and customers, and if the route to that evidence depends on broad registry discretion, the economic effect resembles licensing. The registry may not control fibre, spectrum or retail authorisation, but it controls a key recognition layer without which independent operation is harder.
This is why the core doctrine should be modest. The registry's authority is strongest when it is narrow: uniqueness, accurate records, legitimate holder identity, clear resource status, fraud correction, security assertions and documented changes. It is weakest when it tries to decide the moral standing of commercial arrangements. Leasing, transfers, sub-delegation, managed services and cross-border customers can all create evidence problems. Evidence problems require evidence rules. They do not require a registry to become a planner of the regional access market.
The founder in a secondary city should not have to persuade a distant institution that the business is socially admirable. The founder should have to show specific facts: who the company is, what network it will operate, what resources it needs, how existing capacity is used, how customers are assigned, who handles abuse, who may authorise routing, how records will remain accurate, and what happens if assumptions change. A narrow evidence standard can ask these questions without turning the registry into a discretionary commercial judge.
When the ledger becomes a gate, incumbents benefit. When the ledger stays a ledger, entrants can compete on service, price, local knowledge and execution. The difference may sound institutional, but it decides whether a small ISP's hardest problem is building a network or satisfying an elastic permission structure.
Failed entry is hard to count
The cost of entry barriers is often invisible because the missing competitor leaves no outage report. A small ISP that never launches does not create a customer complaint, a routing incident or a regulatory case. A founder who postpones a town because address capacity is uncertain does not appear in registry statistics. A bank that declines a facility because the address plan is too hard to underwrite does not publish a technical analysis. The market simply remains more concentrated than it might have been.
This invisibility matters because registry debates often count the risks that are easiest to see. Fraud is visible when discovered. Abuse is visible when reported. Litigation is visible when filed. A disputed transfer is visible when parties fight. The absent entrant is quieter. Yet from a development and competition perspective, the absent entrant may be the larger cost. A region with fewer local ISPs has less pricing pressure, fewer service models, weaker local support and less resilience when incumbents underperform.
The lost experimentation is especially important. Small ISPs often discover demand that larger firms overlook: a cluster of farms needing remote monitoring, a school network outside the standard footprint, a market district with poor indoor coverage, a town where residents will accept fixed wireless before fibre arrives, or a group of small businesses that need practical support rather than a national call centre. These opportunities are not always visible in advance. They emerge because someone can start modestly. If the address and evidence burden requires too much scale before launch, fewer experiments occur.
There is also a talent cost. Local network operators learn by building. A market with many small entrants creates engineers, installers, support staff and managers who understand the specific geography of service. If entry is too formal, talent concentrates in incumbents and large contractors. That may be efficient for large projects, but it weakens the local ecosystem that makes competitive access possible over time.
AFRINIC's role in this lost-entry problem should be stated carefully. It is not responsible for every failed ISP plan. Power, fibre, spectrum, currency risk, taxes, politics and customer affordability may dominate in many cases. But when the registry layer adds avoidable uncertainty to an already difficult launch, it contributes to a market structure in which fewer small firms attempt the first step. The harm is not a dramatic collapse. It is a quieter narrowing of possibility.
That is why entry analysis should focus on thresholds rather than only incidents. How much capital must a founder raise before becoming credible? How much documentation must exist before the first customer is connected? How much address capacity must be secured before a public buyer signs? How much uncertainty must a bank price? How many months can a small team wait? Each threshold may look modest to an established operator. In combination, they decide whether the entrant exists.
Proportional proof would lower the entry floor
The safeguards that matter for small ISP entry are not grand constitutional slogans. They are practical limits on uncertainty. They should make the first independent step cheaper without weakening protection against fraud or waste.
The first safeguard is a transparent queue. Entrants should know where a request stands, what makes an application complete, which missing facts matter, what the expected review window is, and when escalation is available. A queue does not create more IPv4, but it converts waiting time into a financeable cost. A founder can plan around a known delay. A founder cannot plan around a status that is informal, shifting or unexplained.
The second safeguard is proportional documentation. A small ISP seeking early capacity should not face the same evidentiary burden as a mature carrier seeking a large block or a complex transfer. The standard should distinguish launch evidence, growth evidence and remedial evidence. Launch evidence might include corporate identity, network design, upstream arrangements, site plans, customer pipeline and an initial assignment plan. Growth evidence might require observed utilisation and customer records. Remedial evidence might apply where records are stale, claims conflict or fraud indicators appear. Different risks need different proof.
The third safeguard is a small-operator correction window. Entrants should be able to correct good-faith errors without immediate existential threat. Missing contact data, imperfect assignment records or a change in customer mix should trigger correction and review, not sudden loss of recognition. Strong remedies should remain for fraud, forged authority, deliberate concealment or abandoned resources. But the default for young networks should be cure before destruction. Entry is messy; policy should distinguish mess from abuse.
The fourth safeguard is a narrow standard for leases and transfers. AFRINIC need not bless every commercial term. It should make clear what evidence is needed for routing authorisation, holder legitimacy, abuse responsibility, renewal expectations and customer protection. If an entrant leases capacity, the market should be able to see enough to trust the chain without exposing confidential customer lists. If an entrant receives a transfer, the facts required for approval should be bounded and timed.
The fifth safeguard is an appeal clock. Disagreement will happen. The entry barrier is not disagreement itself; it is indefinite disagreement. A small ISP can survive a clear refusal more easily than endless uncertainty, because a refusal permits a new plan. Appeals should have deadlines, reasons and a record of what was considered. Silence is costly because it traps capital.
The sixth safeguard is an audit trail. Decisions that affect scarce resources should leave a usable trail: request, evidence category, decision, reason, correction path and review status. The trail need not publish confidential materials. It should be strong enough that future staff, counterparties and reviewers can understand why a decision was made. For entrants, auditability lowers the fear that rules will be reinterpreted after customers are connected.
These safeguards share one aim: lower the amount of scale a small ISP must reach before it can be trusted. They do not ask AFRINIC to ignore scarcity. They ask it to manage scarcity in a way that does not make institutional navigation a larger barrier than engineering competence.
What competition needs from AFRINIC
African access markets need more than stable registries. They need power, backhaul, spectrum access, ducts, local exchange points, fair wholesale terms, sensible taxation, finance, skilled technicians and customers who can afford service. But number-resource credibility sits across all of these. A network that cannot present a trusted address story will struggle even if the radio link is clean and the customers are waiting.
For AFRINIC, the competition test is practical. Does the registry lower the cost of becoming an independent network, or does it raise the threshold until only incumbents and well-funded firms can pass comfortably? Does it make scarce IPv4 more legible and usable, or does it hide price inside delay and discretion? Does it help small entrants prove what needs proving, or does it demand mature-operator evidence before the entrant has had a chance to mature?
The answer will be visible in market behaviour. If small ISPs increasingly rely on upstream-provided addresses because direct paths are too uncertain, upstream bargaining power will rise. If they lease because buying or transferring is too hard, lease quality and renewal risk will become central to competition. If they overuse CGNAT because public IPv4 is inaccessible, enterprise opportunities will tilt toward incumbents. If banks and public buyers keep discounting new networks because number-resource evidence is weak, capital will flow to firms already established. The registry may not intend any of these outcomes, but it will have helped produce them.
A better path is available. Treat the registry as a ledger of uniqueness and evidence, not as a general permission gate. Keep documentation real but proportionate. Keep queues visible. Keep transfer and leasing evidence bounded. Keep appeals timed. Keep correction paths open. Keep strong remedies for fraud and deliberate deception. Keep registry services boring even when governance politics is not. Above all, measure success by whether capable entrants can obtain enough credible number-resource evidence to compete.
Small ISPs are not asking for exemption from scarcity. They cannot be. IPv4 is finite, and the AFRINIC free pool is not large enough to carry the continent's long-term growth. They are asking, in effect, for scarcity not to be made worse by unnecessary uncertainty. They need prices they can compare, rules they can understand, evidence they can produce, and records counterparties can trust.
The founder in the secondary city will still have hard problems after that. Towers will be expensive. Power will fail. Customers will bargain. Incumbents will respond. Radios will break in rain. Backhaul will cost more than the spreadsheet hoped. Those are the ordinary hardships of building access networks. Registry design cannot remove them.
What registry design can do is avoid adding a hidden licence to the pile. It can ensure that the ability to prove number-resource credibility does not become the privilege of firms that already have scale. If AFRINIC succeeds at that narrow task, it will not create competition by itself. It will remove one of the quieter barriers that keeps competition from starting.