The budget room has two spreadsheets open before anyone starts talking about ideology. One is the visible IPv6 plan: access equipment, routing upgrades, customer-premises devices, monitoring changes, training, public-service readiness and procurement language for a network that should not be trapped in the old address economy. The other is the IPv4 continuity plan: public endpoints, address leases, transfer diligence, reverse DNS, scarce-address inventory, security logs, customer allowlists, abuse evidence, bank and government exceptions, help-desk scripts, reputation management and the legal comfort required to keep address-dependent contracts usable.
The two spreadsheets do not cancel each other. They add. That is the economic fact hidden inside the phrase "dual stack." A network can be serious about IPv6 and still pay heavily for IPv4 certainty. It can deploy modern equipment and still need a scarce public address for a bank gateway. It can train engineers on IPv6 and still keep old support paths for customers whose counterparties whitelist IPv4. It can publish IPv6 services and still keep translation, logging and compliance evidence for applications that do not behave cleanly in a mixed environment.
Dual stack is therefore not only an engineering state. It is a cost-incidence problem. During the overlap, the operator pays for future capacity before the old dependency releases its claim on cash, labour and risk reserves. Some parties can pass that burden through as a premium service. Some can spread it across large customer bases. Some can convert it into vendor revenue, consulting work, platform fees or scarcity value. Others absorb it in margins, delayed investment, higher support costs or weaker competitive position.
AFRINIC makes the problem unusually concrete because the IPv4 side of the ledger is not a quiet background utility. The African Network Information Centre is the Regional Internet Registry serving Africa and the Indian Ocean region, and its own exhaustion material records that it entered IPv4 Exhaustion Soft Landing Phase 2 on 13 January 2020, with Phase 2 requests constrained between a /24 minimum and a /22 maximum. That is a rationing environment. New IPv4 supply is small by design, while the customers, public agencies, banks, data centres and international counterparties that still require IPv4 do not disappear on the same date.
The registry layer has also carried its own premium. Public reporting has described alleged African IPv4 address misappropriation, the Cloud Innovation dispute, frozen AFRINIC bank accounts in 2021, court proceedings in Mauritius, a court-appointed receivership in 2023, election disputes in 2025, later board recovery reporting in 2026, ICANN intervention in a winding-up context and continuing litigation. Those reports are contested in places and should not be treated as final findings on every allegation. For operators, the narrower point is enough: if registry records, reverse DNS, resource certification, transfer recognition, member standing or dispute notation become uncertain, the IPv4 side of the dual-stack budget becomes more expensive.
The central question is not whether African networks should deploy IPv6. They should, and many already do in parts of their networks. The question is who pays for the period in which IPv6 deployment does not retire IPv4 dependence. The answer differs by operator type, customer mix, address inventory, bargaining power, public-sector requirements and the boringness of the registry ledger. A serious AFRINIC-region policy has to ask where the duplicate bill lands, which parts are avoidable, and what reforms would make coexistence cheaper.
The cost begins with a second ledger
The first ledger is easy to discuss in public. It contains the visible IPv6 transition items: compatible routing equipment, software upgrades, address planning, customer-device support, staff training, test labs, monitoring updates and national-adoption commitments. Those costs can be presented as investment. They fit the language of future readiness. They make a good slide for a regulator, a vendor seminar or a board strategy day.
The second ledger is harder to present because it looks like failure. It contains the costs that remain because IPv4 cannot be retired: public endpoints for banks and enterprise customers, IPv4 pools for access networks, translation equipment, reverse-DNS maintenance, abuse handling, address-reputation work, fraud and lawful-request logs, customer allowlists, manual troubleshooting, scarce-address rent, transfer diligence, inventory reviews and contingency plans for address disputes. These are not glamorous costs. They are the costs of compatibility.
The dual-stack economy begins when both ledgers are true at once. An operator can make real progress on IPv6 and still need to spend more on IPv4 certainty. A data-centre operator can offer IPv6 to tenants while still losing customers if it cannot provide stable IPv4. A mobile operator can move handset traffic toward IPv6 while still needing IPv4 attribution and customer support for applications that break behind translation. A public-service network can publish IPv6 while still keeping IPv4 endpoints because citizens, vendors and regional partners are unevenly upgraded.
That simultaneity is the incidence problem. If the old system were simply replaced, costs would move from one account to another. If the old system were irrelevant, its remaining cost would be small. In the real transition, the new system adds cost before the old system releases cost. The operator must pay for IPv6 readiness without being allowed by customers, banks, procurement rules or legacy systems to stop paying for IPv4 reachability.
The burden is not merely a capital burden. It is managerial attention. Every extra system creates meetings, change windows, training plans, escalation paths and risk registers. Every extra policy set creates exceptions. Every exception creates a support memory. The CFO can depreciate hardware, but it is harder to depreciate the mental cost of asking every product team whether a service is IPv4-only, IPv6-ready, dual-stack, translated, customer-specific or dependent on a third-party allowlist that no one can easily change.
For AFRINIC-region operators, the second ledger is made heavier by scarcity. AFRINIC's Phase 2 rules show that remaining new IPv4 allocations or assignments are small by design. Larger growth must be met through conservation, transfers, leases, upstream assignments, address sharing, IPv6 substitution or product redesign. Each option carries a cost and a risk profile. Dual stack therefore does not sit on top of abundant legacy inventory. It sits on top of a scarce input whose certainty, reputation and transferability have become more valuable precisely because new supply is constrained.
The right question is not "what does IPv6 cost?" That is too narrow. The right question is "what does coexistence cost, and where does the burden land?" Once phrased that way, the issue changes. It becomes less like a technology programme and more like a regulated-utility tariff case, an infrastructure depreciation schedule or a tax-incidence study. The person who writes the invoice is not always the person who carries the final burden.
Fixed costs duplicate before savings arrive
Dual-stack cost is often underestimated because attention goes first to address space. IPv6 addresses are abundant, so the mental shortcut is to assume the constraint has been lifted. But a network does not consist of addresses alone. It consists of equipment, software, security policies, monitoring systems, billing records, customer devices, staff routines, partner interfaces, procurement files and risk controls. Most of those systems do not halve their cost just because IPv6 is enabled.
The most obvious duplicated cost is equipment. Routers, broadband gateways, firewalls, carrier systems, load balancers, DNS systems, intrusion-detection platforms, test gear and customer devices must support IPv6 at scale, not merely in a vendor brochure. Older equipment may need replacement earlier than its physical life would otherwise require. New equipment must be tested under mixed traffic conditions. Feature parity matters. A box that supports IPv6 forwarding but not the required logging, filtering, telemetry, quality-of-service or operational tooling creates hidden cost somewhere else.
The second duplicated cost is configuration. A network that once had a mature IPv4 policy set now needs IPv6 addressing plans, firewall rules, filtering logic, monitoring thresholds, DDoS playbooks, customer templates, peering settings, troubleshooting procedures and audit evidence. Some tasks are conceptually similar across address families, but similarity does not eliminate work. The address format changes, neighbour discovery matters, extension-header treatment changes, customer devices behave differently and some security products expose weaker IPv6 tooling than their IPv4 equivalents.
The third duplicated cost is testing. A service is not truly ready because it answers over IPv6 once in a lab. It must perform under real customer conditions, fail cleanly, appear correctly in monitoring, survive firewall changes, produce useful logs, work with content-delivery paths, maintain geolocation quality, avoid unexpected mail or fraud problems and support customer-service diagnosis. The operator must prove not only that IPv6 works, but that the combined IPv4 and IPv6 service does not create intermittent failures that are harder to reproduce than a clean outage.
The fourth duplicated cost is operational knowledge. Engineers must know both worlds. Help desks must ask better questions. Security teams must understand two evidence trails. Procurement teams must avoid buying products that support one family poorly. Sales teams must avoid promising customers that IPv6 removes the need for scarce IPv4 when the customer's own counterparties still require it. Training is not a one-off seminar. It is a recurring cost because staff change, products change and the old shortcuts remain embedded in customer language.
Variable savings come later and unevenly. Some IPv4 addresses can be conserved. Some private networks can avoid public IPv4. Some new products can be IPv6-first. Some customer classes can be moved behind translation with acceptable experience. But savings appear by segment, not all at once. The operator may save addresses in mass consumer access while spending more on enterprise IPv4 certainty. It may reduce demand in new internal networks while still buying public addresses for data-centre tenants. It may deploy IPv6 in the mobile core while expanding IPv4 logs because shared-address use increases.
This timing matters. Costs are paid now; savings are uncertain and customer-specific. The firms best able to finance the gap are large incumbents, profitable mobile groups, multinational platforms and data-centre operators with pricing power. The firms least able to finance it are small broadband providers, public-service operators, rural networks, university networks, municipal providers and regulated utilities that cannot easily change tariffs. Incidence starts with timing. Whoever has the weakest cash flow during the overlap pays most painfully.
Customer mix determines pass-through
Cost incidence is not determined only by engineering. It is determined by the customers an operator serves. A network with enterprise customers that explicitly buy dedicated IPv4, managed security, static endpoints or premium support may be able to charge for part of the dual-stack burden. A mass-market provider selling low-cost broadband in a competitive city may not. A public-sector operator bound by procurement rules may have to maintain compatibility without a matching budget line. A mobile operator may spread some costs across millions of subscribers, but it also faces huge logging and support volumes.
Enterprise customers can sometimes carry visible charges. If a bank needs stable IPv4 endpoints for partner allowlists, fraud systems and compliance records, the provider can price that service as a premium feature. If a hosting customer needs clean public IPv4 addresses, the data-centre operator can charge an add-on. If a security-conscious client wants separate egress, logs and reverse DNS, the provider can frame it as a managed service. The burden is still real, but at least part of it moves toward the customer that demands the scarce compatibility.
Consumer broadband is different. A household rarely understands why IPv4 scarcity affects price. It notices whether games, messaging, streaming, payments and work applications function. If the provider raises prices to cover duplicated systems, competitors may undercut it. If it overuses address sharing, the help desk may receive complaints about blocked sites, broken authentication, degraded gaming or incorrect location. The cost moves around the provider's accounts, but not necessarily into a clean customer invoice.
Public-service networks face a still harsher problem. Tax portals, customs systems, health platforms, education networks, courts and emergency services may be politically unable to fail, yet financially unable to pay premium rates for every compatibility feature they require. The operator serving them may need stable IPv4, careful logging, conservative change windows, support for old devices and manual coordination with agencies. The public receives continuity, but the cost may sit with the operator or with a ministry budget that was never designed to price scarce number resources.
Data-centre and cloud-edge providers have more pricing tools, but also more competition from global platforms with deeper address inventories. A local African hosting provider may need IPv4 to win tenants who serve customers outside Africa or must satisfy old enterprise controls. A global platform can spread address-management cost across a vast business and charge product-specific fees. The local provider may face the same customer requirement with less inventory, less leverage and more registry uncertainty. The incidence therefore affects competitive structure, not only cost recovery.
Customer mix also affects the speed at which IPv6 produces relief. A university can enable IPv6 across parts of its campus but still maintain IPv4 for research partners, student housing, legacy lab equipment and administrative systems. A rural provider may deploy IPv6 in new access equipment but still depend on public IPv4 for school platforms, clinic systems and local businesses. An island operator may use IPv6 where possible while keeping IPv4 for tourism, ports, banking and government services that require predictable reach across old networks.
The result is an uneven map. The same technical state, dual stack, creates different incidence outcomes depending on whether the network sells to banks, households, mobile subscribers, cloud tenants, ministries, schools, hospitals, farms, exporters or foreign partners. A policy that says "deploy IPv6" without asking who the customers are is not wrong; it is incomplete. It identifies the direction of travel while ignoring the toll booth.
Security, monitoring and support turn coexistence into labour
Security is where the duplicate bill becomes hardest to hide. A dual-stack network has more than two sets of addresses. It has two evidence systems for abuse, two families of firewall rules, two families of access-control mistakes, two versions of customer explanations, two sets of monitoring gaps and two bodies of staff knowledge. If the IPv6 side is weaker, attackers will find it. If the IPv4 side remains scarce and shared, investigators and customer-service teams will struggle with attribution.
Logging is the most tangible labour cost. Operators must keep records that allow them to answer abuse reports, customer disputes, enterprise security reviews and lawful requests. Shared IPv4 already makes this harder because many customers may sit behind one public address at different ports and times. Adding IPv6 does not remove the IPv4 logging problem; it adds another path that must be recorded, queried, retained and explained. The operator must know which customer used which address family, through which gateway, at what time and under which service plan.
Security policies also become more complex. A mature IPv4 firewall standard cannot simply be copied into IPv6 without thought. IPv6 traffic may follow different paths, expose different device behaviour and require different filtering assumptions. Customer-premises equipment may support IPv6 inconsistently. Some monitoring and incident-response tools still present IPv4 evidence more cleanly. Security teams must check whether alerts cover both families, whether dashboards distinguish them and whether playbooks tell staff what to do when only one family is affected.
The labour cost is not only in the security operations centre. It reaches procurement. Buying an IPv6-capable firewall is not enough if its logging export, rule management, forensic search, training material or managed-service partner support lags behind. It reaches legal and compliance teams. Retention policies must explain both address families and any translation layer. It reaches customer support. A household complaining that a service fails may not know whether the problem is IPv4 sharing, IPv6 preference, DNS behaviour, application design, geolocation or a remote allowlist.
Address sharing and translation can turn scarcity into a specific support, logging and attribution burden, but they are not the whole story here. The broader dual-stack problem is the budget container around those mechanisms. Even where translation is not the main cost driver, coexistence creates security labour because the operator must keep the old compatibility layer safe while making the new layer safe enough to trust.
AFRINIC's institutional setting matters because security evidence often depends on stable registry facts. Abuse contacts, reverse DNS, RDAP or Whois data, resource certification status and routing records help outsiders decide whom to contact and what to trust. If those records are delayed, disputed or reputationally weak, the operator's security cost rises. Staff must explain, document, reassure and sometimes work around a record layer that should have reduced friction.
Security labour is rarely recovered directly from the party that caused it. A vendor that sells the IPv6 upgrade does not answer every abuse ticket. A governance body that urges adoption does not staff the operator's night shift. A customer that insists on old IPv4 allowlists may not pay for the extra forensic work created by shared addresses and dual paths. The burden lands where incidents land: the operator's desk.
IPv4 certainty turns the duplicate bill into a balance-sheet problem
Dual-stack cost would be easier to absorb if the IPv4 side were merely a declining utility. It is not. IPv4 has become a scarce, priced, financeable and contested input. Operators may own, lease, transfer, conserve or rent access to public addresses. Some carry address holdings as strategic assets even where accounting treatment remains uncertain. Others pay recurring fees or leases for scarce public identity. The dual-stack bill therefore includes both the cost of building IPv6 and the cost of preserving IPv4 optionality.
The scarcity is objective enough to shape behaviour. AFRINIC's Phase 2 rules limit new IPv4 allocations or assignments to small blocks. The Register reported in February 2026 that an AFRINIC executive told APRICOT the registry had 773,376 unallocated IPv4 addresses and looked forward to that number reaching zero so the conversation could shift more fully to IPv6. Whether one hears that remark as optimism or impatience, it is a useful exhibit. The remaining pool is finite, administratively rationed and too small to absorb the region's long-term growth.
Once IPv4 has market value, address certainty becomes a form of capital preservation. A provider with clean, stable, transferable IPv4 can support enterprise products, hosting, premium access, public-sector continuity and future financing options. A provider without it must rely on upstream assignments, sharing, leases or delayed product plans. IPv6 deployment may reduce future demand, but it does not erase the capital value of scarce addresses during coexistence.
This is where the dual-stack burden differs from ordinary technical modernisation. A company replacing a router retires an old asset and gains a new one. A company deploying IPv6 while maintaining IPv4 keeps paying to protect a scarce old asset and to build a less scarce new capability. The two do not cancel each other. If anything, the new capability can make the old asset more strategically concentrated. IPv4 is reserved for the customers and services that still need public compatibility most.
Balance-sheet incidence differs by operator type. A large incumbent with historical address holdings may experience IPv4 scarcity as option value. It can allocate scarce addresses to profitable customers, lease unused inventory, negotiate from strength or delay purchases. A new entrant experiences the same scarcity as a working-capital drain. It must obtain address certainty before some customers will sign, yet it lacks the customer base that would make the cost easy to spread. A public-service operator experiences scarcity as continuity risk because the services it supports may be politically important but commercially underpriced.
Scarcity also affects lenders and investors. A network plan that assumes cheap, certain IPv4 compatibility may look less credible in an AFRINIC Phase 2 environment. A data-centre expansion may require proof of address availability. A mobile product may need a plan for translation and logs. A public tender may require reachable endpoints. The cost of proving those plans is part of dual-stack incidence. It is due diligence labour added to technical cost.
The uncomfortable conclusion is that IPv6 does not remove the finance question quickly enough. It changes long-term architecture, but the medium-term operator still has to finance address certainty. If policy treats IPv4 as shameful legacy baggage, it may obscure a real asset and a real cost. If policy treats IPv4 as untouchable capital and ignores IPv6, it underprepares the future. The dual-stack budget forces both truths into the same room.
AFRINIC uncertainty adds a registry-risk premium
In a stable registry environment, IPv4 scarcity is already expensive. In an uncertain registry environment, it becomes more expensive because every address-dependent decision must include a confidence premium. The operator asks not only whether it can route, lease, transfer or conserve addresses, but whether the registry record will remain stable, whether a dispute will affect routine changes, whether reverse DNS can be maintained, whether resource certification remains coherent, whether member standing is secure and whether courts or governance disputes will delay recognition.
AFRINIC's history gives operators reason to price that premium. The 2019 KrebsOnSecurity report described allegations that a long-serving AFRINIC figure was linked to companies that sold African IPv4 blocks, with researcher Ron Guilmette estimating more than US$50 million in affected address value. That history made registry-record integrity more than an administrative issue. It showed how scarce IPv4 value creates incentives for misappropriation, dormant-record exploitation and reputational damage.
The 2021 Internet Governance Project account added a different kind of risk. AFRINIC's dispute with Cloud Innovation escalated from policy interpretation and regional-use claims into litigation, attempted resource withdrawal, bank-account freezes and wider questions about the viability of the registry. The authors criticised both AFRINIC's aggressive remedy and Cloud Innovation's legal response. For cost-incidence purposes, the important point is not to decide the case. It is that a registry dispute over IPv4 use can generate costs far beyond the parties' original disagreement.
Receivership made continuity visible. In 2023 AFRINIC was placed under a court-appointed receiver in Mauritius. Internet Governance Project described receivership as a mechanism for preserving organisational stability while replacing leadership and moving toward elections. That may be a strength of private-law governance. It is also a signal to operators that registry continuity can become a matter for courts, receivers, injunctions and election mechanics. A network planning dual stack must then ask how routine registry services survive institutional stress.
The board issue did not vanish quickly. Reporting in 2025 described a board election attempted under receivership, legal challenges, ICANN objections, disputes over voting rights, alleged proxy irregularities and an annulled poll. Reporting in 2026 described AFRINIC as rebuilding after electing a board, while also describing continuing litigation and ICANN's intervention in proceedings related to winding up the registry. That is not a settled, boring administrative background. It is an active risk environment around the records that make IPv4 usable in commerce.
The premium appears in small ways. A customer asks for extra warranties. A bank asks whether an address block could be disputed. A public buyer asks who controls reverse DNS. A security team asks whether abuse contacts are correct. A broker or lessor asks for stronger paperwork. A board asks whether the company's scarce-address plan is exposed to registry action. Each question creates labour. Each labour item is part of dual-stack cost incidence because it belongs to the IPv4 side of coexistence.
The policy implication is direct. If a registry wants operators to invest in IPv6, it should reduce the uncertainty premium on IPv4. Clean records, predictable transfers, bounded reviews, reliable reverse DNS, stable resource certification, clear dispute notation and continuity plans lower the cost of the old layer while the new layer grows. Institutional drama does the opposite. It makes operators pay more for the compatibility they cannot yet retire.
The clean-transition story has beneficiaries
The clean-transition story is attractive because it assigns virtue before it assigns cost. IPv6 is modern. IPv4 is legacy. Deployment is progress. Scarcity is an obstacle. The story is not entirely false; IPv6 does provide a larger address space and can reduce future constraints. But as a political economy, the story is incomplete. It does not ask who profits from transition activity, who gains status from the transition narrative and who pays for coexistence when the endpoint keeps moving.
Equipment vendors are obvious beneficiaries. Dual-stack readiness supports hardware refresh, software licences, support contracts, professional services, audits, monitoring upgrades and training. Vendors do not need IPv4 to disappear for the business to be attractive. A long coexistence period can be better. It keeps old capabilities necessary while making new capabilities mandatory. Complexity becomes a revenue stream.
Consultants and managed-service providers also benefit. Address planning, security review, migration testing, procurement advice, training and troubleshooting all become services. Some of that work is valuable and necessary. But it still matters that the people selling transition expertise do not always carry the risk of customer churn, regulatory failure or public-service outage if the coexistence plan underperforms.
Governance institutions benefit in a different way. IPv6 adoption justifies programmes, meetings, policy work, training, measurement and institutional relevance. Again, some of that work is useful. The risk is that a governance body can speak as if transition is a collective moral project while the operator experiences it as a private cost. If the body also controls or influences IPv4 recognition, the narrative can become a way of disciplining scarce-asset holders while preserving institutional centrality.
Large platforms may benefit from scale. They can run sophisticated dual-stack systems, buy or hold IPv4 inventory, charge for scarce public IPv4 exposure, build translation into product architecture and spread operational cost across many customers. Smaller local providers face similar compatibility requirements without the same scale economies. The clean-transition story can then hide a competitive shift: the cost of coexistence may strengthen those already large enough to industrialise it.
Large address holders may also benefit. Scarcity makes inventory valuable. If dual stack lasts longer than expected, high-quality IPv4 holdings remain useful. This does not make holding addresses illegitimate. It means policy should be honest about incidence. The operator with inventory has optionality. The operator without inventory has a bill. The customer may see only a service price, while the underlying scarcity premium moves through the market.
The party least able to benefit from the story is the operator whose customers are not ready, whose tariffs are constrained and whose address certainty is weak. It must keep buying modern equipment and keep defending old compatibility. It must explain to regulators that IPv6 is progressing and explain to customers why IPv4 remains necessary. It must absorb vendor complexity, governance uncertainty and customer conservatism in one operating budget.
This is why the incidence question is sharper than the adoption question. Adoption asks whether IPv6 is spreading. Incidence asks who pays for the period in which adoption does not eliminate IPv4 demand. The first question can produce celebratory charts. The second produces a list of invoices.
Regulators can shift costs without seeing the invoice
Regulators and public authorities often enter the dual-stack debate with good intentions. They want national networks to be future-ready. They want public services to avoid technological stagnation. They want domestic operators to align with global standards. They may worry that slow IPv6 adoption will leave the country dependent on scarce IPv4 or less attractive to investment. These concerns are legitimate. But regulatory language can shift costs in ways that are not measured.
An IPv6 mandate can become an unfunded requirement if it does not distinguish between capability and retirement. Requiring operators to support IPv6 on new networks may be sensible. Treating that requirement as evidence that IPv4 costs should decline quickly may be wrong. The operator may still need IPv4 for banking, payments, government services, enterprise customers, mobile applications, tourism systems, education platforms and cross-border partners. A mandate that sees only the new capability may ignore the old dependency.
Public procurement is a major incidence channel. If government tenders require public IPv4, conservative whitelisting, older security evidence or compatibility with legacy agency systems, the state is not merely observing the dual-stack cost. It is creating part of it. The same state may then ask operators to accelerate IPv6. That is not necessarily hypocrisy; large systems change slowly. But the fiscal truth should be visible. The public sector cannot demand old compatibility and new readiness while pretending that the operator's cost curve is a private inefficiency.
Regulators can also affect pass-through. In competitive or politically sensitive markets, operators may struggle to raise prices even when dual-stack costs are real. A rural broadband provider may face price pressure from customers with low ability to pay. A mobile provider may fear churn. A public-service contractor may be locked into multi-year pricing. If policy mandates technical readiness but prevents cost recovery, the incidence falls on margins, maintenance deferral, lower investment or lower service quality.
A better regulatory posture would begin with evidence. Which public agencies still require IPv4-only integration? Which banks and payment providers require IPv4 allowlists? Which procurement clauses assume IPv4? Which schools, hospitals, customs systems and courts are unable to use IPv6 reliably? Which operators are paying for translation, logs, scarce address leases or manual support because public counterparties have not moved? These questions do not weaken IPv6 policy. They make it financially literate.
Regulators can reduce incidence by changing their own demand. Public services can publish stable dual-stack endpoints. Procurement templates can stop requiring IPv4-only controls where IPv6 evidence is acceptable. National cybersecurity guidance can include IPv6 logging and incident response so operators are not left inventing standards alone. Public funding can recognise scarce-address costs in rural, public-service or small-operator support programmes. Sector regulators can allow transparent recovery for objectively measured dual-stack burdens.
The point is not to socialise every cost. Operators should bear normal business risk. The point is to avoid invisible cost shifting. When a regulator praises IPv6 but leaves public systems IPv4-dependent, it is asking operators to finance the gap between policy language and state capability. In Africa, where many networks already face high equipment, power, backhaul and financing costs, that gap is not trivial.
Smaller and public-service networks face the steepest curve
Dual-stack cost has economies of scale. A large operator can spread training, monitoring, security tooling and address-management systems across many customers. A small operator buys many of the same capabilities for fewer users. A public-service network may need high reliability without commercial pricing freedom. This is why cost incidence is often regressive even when technical requirements are formally equal.
The small ISP's problem is minimum efficient scale. It needs IPv6-capable equipment because new networks should not be built on a dead end. It needs IPv4 reachability because customers and upstream systems still expect it. It needs support staff who can diagnose both. It needs logs credible enough for abuse handling and legal requests. It needs public-address evidence for enterprise customers. It may need leased IPv4 or upstream assignments before revenue is stable. A large incumbent treats these as departments. A small ISP treats them as founder time and working capital.
Public-service operators face a different steep curve. They may support schools, clinics, municipal offices, tax systems, customs platforms or identity services. Their customers value continuity more than novelty. They may not reward elegant IPv6 architecture if a legacy vendor breaks. The operator therefore keeps the old compatibility layer conservative while adding the new layer carefully. It may need duplicated support windows, cautious change management, extra documentation and more manual coordination than a purely commercial provider would tolerate.
Universities and research networks sit between these categories. They often have advanced technical staff and a public mission, but their internal environments are messy. Student networks, administrative systems, laboratories, international partners, identity systems, building controls and grant-funded equipment may move at different speeds. IPv6 can be deployed well in some segments while IPv4 remains essential elsewhere. The cost lands in network teams that must satisfy innovation and continuity at once.
Rural and island networks experience the burden through logistics. A replacement device may require foreign procurement, customs clearance, scarce engineering labour and expensive site visits. A misconfigured dual-stack service may take longer to diagnose when the specialist is remote. A public IPv4 shortage may force address sharing that increases support calls. IPv6 may reduce future pressure, but only after equipment, training and customer environments catch up. Until then, coexistence is a heavier fixed cost per customer.
These networks are also more exposed to registry uncertainty. A large operator can retain counsel, attend governance meetings, maintain dedicated registry staff and build contingency plans. A smaller operator may discover the importance of member standing, contact accuracy, transfer paperwork or reverse-DNS authority only when a customer, upstream, bank or public buyer asks. AFRINIC's years of crisis are therefore not background politics for such networks. They are a reason to spend scarce management time on institutional risk.
If policy ignores this steep curve, it may unintentionally favour scale. Requirements that are modest for a large operator can deter entry or weaken smaller competitors. A clean-transition narrative may then produce an unclean market outcome: fewer local providers, more dependence on large platforms, and less competition for public-service and enterprise customers. Cost incidence becomes market structure.
Ledger continuity is a cost control
The phrase "registry continuity" can sound abstract until it appears in a budget. If RDAP or Whois records are trusted, if reverse DNS can be updated, if resource certification remains stable, if routing records are coherent, if transfers and leases are documented, and if disputes are isolated from routine operations, the operator's IPv4-side cost falls. If those functions are uncertain, the operator pays more to prove what the registry should make evident.
AFRINIC's receivership made this clear. A court-appointed receiver can preserve operations, maintain status quo and move toward governance repair. That may prevent collapse. But the very need for receivership shows that registry services should not depend on the uninterrupted health of one board or one corporate structure. Operators running dual stack need the old layer to remain boring while the new layer grows. Boringness is economic value.
Ledger continuity is broader than keeping a website online. It means preserving the chain of authorised changes, maintaining accurate holder information, protecting reverse-DNS delegations, keeping resource certification coherent, marking disputes without contaminating unrelated services and ensuring routine changes do not become hostage to institutional conflict. It means that courts, receivers or governance disputes should not force customers into uncertainty about addresses that remain in live use.
In the AFRINIC setting, this matters because IPv4 is not just an address family. It is evidence. It supports bank allowlists, mail reputation, enterprise contracts, government portals, hosting products, security reviews, abuse handling and customer continuity. If the evidence layer becomes politically or legally unstable, every operator using that layer carries extra cost. The burden is not limited to the party in court. It can spread through customers, counterparties and markets that simply become more cautious.
The 2019 reporting on alleged address misappropriation points to one side of the ledger-continuity problem: integrity and control risk around valuable records. The 2021 dispute points to another: enforcement, legal escalation and proportionality. The 2023 receivership points to a third: institutional continuity under court oversight. The 2025 and 2026 reporting points to a fourth: board legitimacy, election trust, ICANN intervention and continuing litigation. Each type of stress can make IPv4 certainty more expensive.
None of this means AFRINIC should be unable to correct fraud, enforce clear rules or recover from crisis. It means those powers must be designed so that the cost of institutional action does not fall indiscriminately on running networks and downstream customers. Correcting a forged record is different from creating broad uncertainty around commercial use. Marking a dispute is different from breaking service. Maintaining uniqueness is different from claiming open-ended control over business models.
For dual-stack policy, ledger continuity is cost control. It reduces the price of keeping IPv4 alive while IPv6 grows. It lowers the premium customers demand. It reduces legal review. It protects address reputation. It makes transfers and leases more financeable. It allows operators to focus on modernisation rather than explaining registry risk. In a scarce environment, a boring ledger is not administrative luxury. It is infrastructure.
IPv4 cost is more than address rent
When dual-stack burden is discussed, the IPv4 side is often reduced to address rent or transfer price. That misses much of the incidence. Address rent is visible because it is an invoice. Transfer price is visible because it is a transaction. The larger burden often sits in operational tasks required to make scarce IPv4 usable and trusted.
Address conservation is one such task. Operators must decide which services deserve public IPv4, which can use private addressing, which can share, which can move to IPv6, and which require customer-specific treatment. Those decisions require product knowledge, customer negotiation, engineering design and risk review. A scarce-address committee inside an operator is not free simply because it does not buy hardware.
Address reputation is another task. Public IPv4 used by one customer may carry history that affects another. Mail systems, fraud tools, geolocation services, content platforms and security filters may react to address behaviour. Cleaning reputation, documenting use, moving customers and responding to blocks create labour. IPv6 adoption does not eliminate this work if customer-facing IPv4 remains essential for high-value services.
Procurement adds another cost. Vendors may advertise IPv6 support, but the operator must check whether support is adequate across logging, security, telemetry, automation, customer management and troubleshooting. Procurement teams must ask better questions. Engineers must test claims. Legal teams may need service-level language for both address families. Customer contracts must avoid promising a retirement path that the operator cannot control.
Training is recurring. A network can train senior engineers once, but help desks, field teams, security analysts, sales staff and customer-success teams also need enough knowledge to avoid bad diagnosis. When a customer says "the Internet is down", the support path must consider DNS behaviour, IPv6 preference, IPv4 sharing, application limitations, remote filtering and device quirks. That is a higher labour requirement than a single-stack world.
Troubleshooting is more expensive because failures can be partial. A service may work over IPv4 and fail over IPv6. A customer may reach some destinations but not others. A remote partner may prefer one address family unexpectedly. A mobile application may behave differently on Wi-Fi and cellular. A firewall may log one path more clearly than the other. Partial failures consume expert time because they do not trigger simple outage narratives.
Compliance and record retention are similarly complex. The operator must maintain logs that regulators, courts, customers and security teams can understand. It must explain translation where used. It must avoid losing IPv6 evidence because old tools were IPv4-centred. It must protect privacy while retaining enough information for abuse and legal requirements. These tasks are costs even when no address is bought or leased.
A serious incidence study must therefore measure the full IPv4-side burden: address acquisition, address holding, address certainty, reputation, logging, security, customer support, procurement, training, troubleshooting, legal review, registry interaction and opportunity cost. Otherwise policy will undercount the burden and overstate the ease of transition.
Incidence should be measured before policy is declared
The practical reform is to measure dual-stack cost incidence directly. Without measurement, every party can tell a convenient story. Vendors can say readiness is investment. Governance bodies can say adoption is progress. Operators can say coexistence is expensive. Regulators can say customers need protection. Customers can say they simply want working service. A measurement framework would not settle every dispute, but it would make the burden visible.
The first measure should be duplicated capital expenditure. Which equipment was replaced earlier than planned because of IPv6 requirements? Which security systems, monitoring tools, customer devices, management systems and testing environments were bought or upgraded because dual stack had to work properly? Which of those purchases produced new revenue, and which merely preserved compatibility? The distinction matters because preserved compatibility is often a public benefit with private financing.
The second measure should be operating labour. Staff hours spent on IPv6 training, dual-stack configuration, customer support, security monitoring, logging, procurement review, public-sector coordination, bank allowlist support, incident response, address conservation and registry interaction should be estimated. These hours are not secondary. In many networks, labour is where coexistence becomes most expensive.
The third measure should be IPv4 certainty cost. Address leases, transfer diligence, legal review, broker fees, registry fees, reverse-DNS maintenance, resource certification work, reputation cleaning and documentation for customers should be tracked separately. A network that already holds stable IPv4 should report the opportunity cost of using scarce addresses internally rather than leasing or selling them. A network without inventory should report external acquisition or rent.
The fourth measure should be customer pass-through. Which customer segments pay explicit IPv4 or dual-stack charges? Which receive compatibility bundled into ordinary tariffs? Which are protected by regulation or long-term contracts? Which cause support burden without paying the marginal cost? This measure would show whether the burden lands on enterprise customers, mass subscribers, public budgets, operator margins or future investment.
The fifth measure should be registry-risk premium. How long do routine record changes take? How often are transfers delayed? How often do customers ask for extra registry evidence? How many disputes affect commercial plans? How reliable are RDAP, Whois, reverse DNS, resource certification and routing-related records? How much legal or management time is spent understanding AFRINIC governance risk? These questions convert institutional drama into cost data.
The sixth measure should be avoided cost from IPv6. Where did IPv6 actually reduce IPv4 demand? Which products launched without public IPv4? Which public agencies accepted IPv6 evidence? Which customer classes no longer require scarce endpoints? Which applications still forced IPv4? This measure prevents cynicism. It recognises real IPv6 value while refusing to count adoption as retirement unless a dependency actually falls.
For AFRINIC, such measurement would make policy more honest. It would show whether Phase 2 scarcity, registry uncertainty and dual-stack operations are hitting small providers, public-service networks, data centres, mobile operators or enterprise providers differently. It would also show where registry improvements would reduce cost fastest. If a quicker reverse-DNS update, clearer transfer path or better dispute notation saves operators measurable labour, ledger reform becomes an economic policy, not an institutional slogan.
Costs should be reduced before they are moralised
Internet governance often moralises scarcity before reducing cost. IPv4 holders are accused of hoarding. IPv6 laggards are accused of resisting the future. Commercial leasing is treated as suspect. Operators complaining about duplicated burden are told that transition is necessary. Some of these criticisms may fit some behaviour. As policy, however, moralisation is a poor substitute for cost reduction.
The first cost reducer is clarity. Operators need to know what IPv4 records mean, how transfers are recognised, how leases should be documented, how reverse DNS is maintained, how resource certification changes are handled, how disputes are marked and which actions are routine. Ambiguity creates meetings, legal review and customer hesitation. Clarity lowers incidence without requiring anyone to abandon IPv6.
The second cost reducer is proportionality. A registry should correct fraud and protect uniqueness, but it should not make every change in customer geography, service use or commercial structure feel like a threat to the underlying address position unless a clear rule supports that action. Broad discretion forces operators to carry a risk reserve. Proportionate review lowers the premium.
The third cost reducer is public-sector alignment. Governments should update their own systems, procurement language and security guidance so they do not require IPv4 unnecessarily. Where IPv4 remains necessary for public continuity, budgets should acknowledge it. A state that wants IPv6 but still buys IPv4-dependent services should be honest about the cost it is imposing.
The fourth cost reducer is customer segmentation. Operators should identify where IPv6 can genuinely reduce IPv4 demand and where it cannot yet. Charging premium customers for scarce compatibility may be fair where the requirement is specific and valuable. Bundling every cost into general tariffs may hide the burden and punish low-margin customers. Segmentation is not perfect, but it is more honest than pretending all customers are equally ready.
The fifth cost reducer is shared tooling and training. Smaller networks should not each invent dual-stack security checklists, logging models, procurement language and customer explanations from scratch. Industry associations, network operator groups and public programmes can reduce duplicated learning costs. The point is not to create another ceremony around adoption. It is to lower the actual labour of coexistence.
The sixth cost reducer is market transparency. Where IPv4 must be leased or transferred, pricing, reputation, chain-of-registration evidence and dispute status should be understandable. Markets do not eliminate scarcity, but opaque markets add a second burden. Operators can finance a price more easily than they can finance arbitrary delay.
AFRINIC's role in this cost reduction should be narrow and testable. It should make records reliable, services continuous, transfers predictable, disputes bounded and routine maintenance boring. It should not use IPv6 ambition as a substitute for IPv4 accountability. The more valuable IPv4 becomes during coexistence, the more disciplined the registry must be.
Moral language may still have a place. Regions should modernise. Operators should not ignore IPv6. Customers should not be trapped in obsolete assumptions forever. But moral language should come after the cost map, not before it. Otherwise the party with the nicest story may avoid the invoice.
Coexistence discipline is the policy answer
The practical answer is not to choose IPv4 against IPv6 or IPv6 against IPv4. It is coexistence discipline. That means deploying IPv6 where it reduces future constraint, while treating IPv4 stewardship as a live economic responsibility until customers and counterparties actually retire their dependence. It means measuring the duplicate bill rather than hiding it under the word transition.
For operators, coexistence discipline begins with internal accounting. Separate the IPv6-readiness budget from the IPv4-certainty budget. Identify which products still require public IPv4, which customer classes can pay, which support costs are driven by address sharing, which security tools are weak on IPv6, which staff need training and which registry interactions create delay. The goal is not to resist modernisation. It is to stop modernisation from disguising an unmanaged cost centre.
For boards and investors, coexistence discipline means treating IPv4 as neither shame nor magic. It is a scarce service-enabling input with uncertain legal and accounting treatment, real market value and real operational dependence. IPv6 can lower future need, but it does not instantly write down the strategic value of stable IPv4 holdings. Capital plans should recognise both the asset and the cost of preserving it.
For regulators, coexistence discipline means matching mandates with capability and cost recovery. Require IPv6 support where justified, but also remove IPv4-only requirements from public systems when they are no longer needed. Fund or permit recovery for objectively measured public-service compatibility costs. Use procurement to reduce old dependencies. Ask banks, utilities, agencies and large enterprises when they will accept IPv6 evidence and what blockers remain.
For AFRINIC, coexistence discipline means accepting that its IPv4 ledger remains economically critical even if its policy documents and public events emphasise IPv6. The registry's credibility during dual stack will not be measured only by IPv6 training or allocation. It will be measured by how cheaply and reliably operators can maintain the IPv4 records they still need while building the IPv6 future.
For governance bodies outside AFRINIC, coexistence discipline means resisting the temptation to treat institutional preservation as the same thing as continuity. Continuity means that records, reverse DNS, resource certification, RDAP or Whois services, routing evidence and routine changes remain reliable under stress. It does not require every discretionary claim by an incumbent institution to be insulated from challenge. Protecting the ledger is cheaper than protecting every gatekeeping habit.
For customers, coexistence discipline means clarity about requirements. Enterprises, banks and public agencies that require IPv4 should say why, for how long and what evidence would let them reduce that requirement. Customers that demand premium compatibility should expect to pay for it where the cost is specific. Customers that can move should not force operators to maintain expensive exceptions out of habit.
The long-run goal may be a world in which IPv6 carries far more of the burden and IPv4 scarcity matters less. The medium-term task is less elegant. It is to allocate the cost of getting there without weakening the networks that must operate along the way. AFRINIC's crisis shows why that task cannot be left to slogans. Scarcity, institutional uncertainty and customer dependence all appear in the operator's budget.
The shadow budget will outlast the slogan
The opening budget meeting ends with a more honest answer than the one the transition slogan offers. The operator will keep investing in IPv6. It will also keep paying for IPv4 certainty. It will buy equipment that can support the future and maintain endpoints that customers still recognise. It will train staff on new address behaviour and keep old support scripts alive. It will collect logs for two paths. It will conserve addresses, lease where needed, defend reputation, reassure banks, satisfy public buyers and watch registry risk.
That is not technological backwardness. It is the economics of a mixed network. The Internet does not move because a protocol has better arithmetic. It moves when customers, vendors, regulators, security systems, procurement rules, public agencies, devices and business models move together. Until then, the operator finances the gap. The gap is dual-stack cost incidence.
AFRINIC is a useful test case because the gap is visible. IPv4 is scarce under Phase 2. The region needs growth. IPv6 deployment is necessary but uneven in its ability to retire real customer dependence. The registry has suffered alleged address misappropriation, resource disputes, bank freezes, receivership, board legitimacy problems, election controversy and continuing litigation. These facts make the IPv4 side of coexistence more expensive than a purely technical plan would suggest.
The danger is that transition rhetoric lets the wrong parties avoid the question. If vendors sell complexity, governance bodies sell inevitability, platforms sell packaged compatibility and regulators sell future-readiness, the operator may be left to pay for everything not named. That burden then moves into higher prices, weaker margins, delayed rural buildout, reduced local hosting, worse customer support or lower investment. Incidence ignored is incidence displaced.
The better path is neither nostalgia nor futurism. It is a cost map. Measure duplicate capital, duplicate labour, security and logging, address certainty, customer pass-through, public-sector dependence, registry-risk premium and actual IPv6 substitution. Publish the results at a level that helps policy without exposing sensitive customer details. Use the results to decide where costs should sit and which reforms would reduce them.
For AFRINIC, the cheapest reform is institutional boringness. Keep the ledger accurate. Keep routine services reliable. Keep disputes bounded. Keep reverse DNS, resource certification, RDAP or Whois services and routing-related records stable. Keep transfer and lease evidence predictable. Keep board and court turbulence away from running networks wherever lawfully possible. The less drama attached to IPv4 records, the cheaper the old side of dual stack becomes.
For operators, the discipline is to stop apologising for the second ledger. If customers still require IPv4, the cost is real. If regulators require IPv6, the cost is real. If both are required, the overlap is real. The question is not whether the future should be built. It is whether the present will be financed honestly enough for that future to arrive without eroding the operators that must build it.
Dual stack is not merely a technical architecture. It is a long shadow budget. In the AFRINIC region, that budget falls across operators, customers, vendors, platforms, governments, address holders and the registry itself. The burden is uneven because bargaining power is uneven. The policy task is to see the unevenness, reduce the avoidable part and allocate the rest openly.
Until that happens, the most important question in the budget room remains the one that transition slogans skip: who pays, who benefits, and what would make the duplicate bill smaller?