The current case against IPv4 markets rests on a flattering fiction: that the RIR system was designed to equalize scarcity across unequal countries, and that markets later corrupted that mission. It was not, and they did not. The now-familiar argument says that IPv4 “commercialization” harms poorer countries—especially in Africa—by turning a supposedly low-cost public good into something expensive, with the burden then passed down to end users. In that story, markets are predatory, Africa is passive, and the end user is the inevitable casualty of allowing scarcity to show its price. The Wall Street Journal has recently offered a neat version of that morality play.

The claim is rhetorically effective because it begins from a false comparison. The market is not being compared with the system that actually existed. It is being compared with an egalitarian order that never did. The Regional Internet Registry system was not built to redistribute address space from rich countries to poor ones. It was built to coordinate globally unique number resources and administer scarcity through registration, documentation, utilization review, and policy. The NRO describes it as a framework for “Internet address management,” under which organizations generally become members and pay fees for registry services. That description matters because it strips away the moral packaging. Once one starts from the system’s actual function, the equalizer myth begins to collapse.

A need-based allocator operating inside an unequal world does not neutralize inequality. It codifies it. More infrastructure, more customers, more deployed capital, more staff, and more institutional capacity produce more documented “need.” More documented need produces more address space. Richer and larger networks therefore, do not merely thrive outside the system; they are easier for the system to satisfy inside it. That is not a deviation from the rules. It is the rules operating exactly as designed. The system does not stand above inequality and correct it. It reads existing scale and existing capacity, and then turns them into administratively respectable outcomes. I have addressed this structural error at greater length in an earlier note, On Why Saying IPv4 Commercialization Harms Poorer Countries Gets the Structure Wrong. The core point is the same: the anti-commercialization argument flatters itself by pretending that a need-based registry system operating inside an unequal world was somehow designed to produce egalitarian outcomes. It was not. Once that mistaken premise is removed, much of the moral rhetoric surrounding IPv4 markets collapses with it.

It was built to manage scarcity, not to redistribute it

The system’s foundational texts are plain. RFC 2050 describes the registry system as the mechanism for distributing globally unique address space and operating registries under policies and guidelines governing distribution. RFC 7020 describes the Internet Numbers Registry System in terms of finite-pool management, hierarchical allocation to support route aggregation, and registration accuracy to preserve uniqueness and operational usefulness. These are technical objectives: uniqueness, conservation, routability and accuracy. They are not redistributive ones.

The language of “stewardship” does not change that. In registry policy, stewardship does not mean giving poorer countries more IPv4 addresses because they are poorer. It means managing a scarce technical resource according to conservation and operational need. ARIN’s policy manual says this directly: uniqueness requires a common pool, conservation requires efficient distribution to organizations with technical need in support of operational networks, routability concerns scalable routing, and stewardship is the application of those principles. One can defend that design on technical grounds if one wishes. One cannot honestly describe it as a welfare mechanism.

The gatekeeping is not incidental. It is the mechanism. AFRINIC’s policy manual requires an organization to be a member in good standing, to demonstrate efficient utilization, and to pass through “slow start” and utilization rules; it also makes clear that staff determines allocation size. [6] When scarcity later shifted attention from fresh allocations to transfers, the market did not replace bureaucracy. It remained embedded inside it. AFRINIC, APNIC, and ARIN all require transfer recipients to demonstrate need, sign agreements, and—in ARIN’s case, explicitly—pay applicable fees to complete the transfer. Even the “market” critics condemn is, therefore, often a transaction that must first survive a permissioned administrative bottleneck.

That point is decisive. The RIR system was never designed to give poorer countries more IPv4 because they were poorer. It was designed to preserve uniqueness, constrain waste, maintain routability, and allocate scarce resources according to demonstrated operational demand. The argument that commercialization betrayed some original egalitarian mission therefore, fails at the level of design. There was no such mission to betray.

The outcomes are not in doubt

If the registry layer were genuinely equalizing, one would expect decades of operation to have narrowed the concentration of number resources. The first RIR was established in 1992. The model has therefore had more than thirty years to show what it does over time. Its record is not ambiguous.

At the global level, Geoff Huston’s country distribution report shows the United States with 1,608,786,848 allocated IPv4 addresses, representing 43.63% of allocated address space, and China with 342,951,680, or 9.30%. Together, they account for more than half of allocated IPv4. At the regional level, the same dataset reports AFRINIC with 116,013,824 allocated IPv4 addresses out of a global allocated total of 3,687,286,208. Telecom SudParis’ separate compilation of the RIR delegated statistics files reports the same AFRINIC total and calculates AFRINIC’s share of world-delegated IPv4 at 3.15%.

The internal African picture is no longer flattering. AFRINIC’s delegation table, as of 23 March 2026, shows South Africa with 27,142,656 addresses, Egypt with 24,149,504, and Morocco with 12,273,920. Those three alone hold 63,566,080 addresses, roughly 55% of the sovereign-state total once Réunion and Mayotte are excluded. By contrast, the bottom thirty sovereign states on that same table together hold only 3,166,464 IPv4 addresses, or about 2.74% of the sovereign-state total.

That is not an equalizer malfunctioning. It is an equalizer myth meeting arithmetic. A system that leaves the bottom thirty sovereign states with 2.74% while the top three hold roughly 55% is not equalizing scarcity. It is administering concentration.

The corporate comparison makes the point harder still to evade. NetworksDB’s ranking of organizations by IPv4 addresses operated lists, among the largest holders, Amazon.com at 119,302,643, Microsoft at 97,174,985, AT&T Enterprises at 95,444,917, Verizon Business at 92,504,365, Comcast at 92,310,946, Amazon Technologies at 78,933,200, Charter at 64,129,162, and Deutsche Telekom at 56,150,336, setting aside the special cases of the U.S. Department of Defense who still have the largest holding on earth, roughly 6% of entire internet’s address, Even the smallest of those major commercial holders, Deutsche Telekom, individually exceeds the bottom thirty African sovereign states combined by about 17.7 times.

After decades of registry “stewardship,” the pattern is therefore plain. Globally, concentration persists. Regionally, it is reproduced. Within Africa, it remains sharp. The system has not neutralized inequality. It has translated unequal scale into administrative outcomes and then asked to be praised for the translation.

Why the burden falls hardest on the poor

Once that is clear, the relevant question changes. The issue is no longer whether commercialization sounds unattractive. The issue is what the registry layer does to the cost of connectivity for weaker operators and poorer end users.

It adds cost. More precisely, it adds fixed cost, delay, uncertainty, and procedural dependence. Those burdens enter upstream and then move downward.

The NRO states plainly that members pay fees to an RIR based on the services they require, and that each RIR sets its own fee structure. RIPE NCC’s billing procedures describe annual service fees, sign-up fees, and additional charges for certain resources. Transfer processes add their own conditions and costs: ARIN requires a signed agreement and payment of applicable fees; APNIC and AFRINIC require demonstrated need; RIPE requires a contractual relationship and restricts transfers of scarce resources. None of this disappears. It is absorbed by operators, priced into plans, and passed through where it can be.

The economic point is simple. A fixed institutional overhead is not neutral in a thin market. It is regressive. Large operators in rich markets can spread legal review, policy compliance, membership fees, administrative delay, and uncertainty across a broad customer base. Small operators in poorer markets cannot. The same bureaucratic layer therefore, becomes more punitive the smaller and poorer the operator underneath it. What is manageable friction in a rich market becomes a meaningful tax in a poor one.

That is why “poverty penalty” is the right term. The burden is not just the explicit fee. It is the fee plus the compliance cost, plus the delay, plus the uncertainty, plus the dependence on discretion. For weaker operators, those hidden costs can be more damaging than a visible price because they are harder to compare, harder to finance, and harder to negotiate around.

This is where the anti-commercialization critique gets the causal story backwards. It treats markets as the source of inequality, when needs-based gatekeeping was already advantaging richer incumbents long before transfer markets became visible. Richer operators already had more staff, better documentation, greater familiarity with the rules, stronger legal capacity, and more ability to survive delay. The market did not create those asymmetries. The registry process was already reading them as “need” and rewarding them accordingly.

Price, whatever else may be said about it, can be compared, budgeted, financed, and negotiated. Discretion cannot. Discretion turns access into a political relationship. A market, even an imperfect one, turns it into a transaction. That is not a moral point. It is an institutional one. The poor do not benefit when the cost of access is hidden inside procedure and then presented as a virtue.

Markets do not create the problem; they expose it

Critics of commercialization often write as if markets created IPv4 scarcity. They did not. Scarcity is a design fact and a historical fact: IPv4 uses a 32-bit address space, with 4,294,967,296 possible values, many reserved for special use. The question is not whether scarcity exists. It does. The question is how scarce resources should move once the free pool is exhausted.

Here, the evidence is less flattering to the anti-market story than its defenders would like. CAIDA’s “Lost in Space” study, using active and passive measurement, estimated that only about 37% of usable IPv4 space was used under its methodology, that 3.4 million assigned /24 blocks were not visible in global BGP, and that only 9.5% of legacy /24 blocks were used, with most unused blocks in the United States. That is not a world in which poorer countries simply lost a fair race. It is a world shaped by historical privilege, legacy overhang, and underutilization.

Transfer-market research points in the same direction. CAIDA’s analysis of reported IPv4 transfers found no evidence of systematic hoarding by buyers in its dataset, reported that most transferred space was routed after transfer, and found increasing utilization trends after transfer, with much of the space originating in legacy holdings. APNIC’s own analysis similarly reported that a high share of transferred space was routed after transfer and that much of it either became routed for the first time or returned to routing after a period of being unrouted. That pattern looks less like predation than reallocation: underused historical holdings moving into active deployment.

Markets do not solve everything. Richer actors can still buy more. Abuse risks are real. Research on transfer-market abuse suggests that transferred space can be disproportionately represented in blacklists under some threat categories, and that transfers create windows of uncertainty that malicious actors can exploit. But that is an argument for a cleaner title, stronger verification, better routing security, and more transparent registration. It is not an argument for romanticizing a bureaucracy whose allocation logic already mirrored existing power and whose procedural burdens already fell hardest on the weak.

The serious comparison, then, is not between a dirty market and a benevolent bureaucracy. It is between two imperfect ways of allocating scarcity. One hides cost inside gatekeeping and presents the result as moral stewardship. The other exposes cost in a visible transaction and, at least potentially, reallocates underused resources into productive use. The answer to a distorted market is a cleaner market. It is not a sanctified bottleneck.

The arithmetic of blame

There is another tell in this debate: the distortion in how blame is assigned.

With over 10 million IPv4 addresses against the size of the IPv4 universe—4,294,967,296 possible values—that is about 0.23%. That is too small a share to bear the explanatory weight now being placed upon it. Yet an actor associated with well under one quarter of one percent is cast as a symbol of global inequality, while the governance structure that administered the rest for decades is still allowed to speak in the language of equality.

That is not analysis. It is misdirection.

The point is not whether one approves of me, my companies, or the secondary market. The point is scale. Ten million addresses do not explain the global concentration of IPv4. They do not explain Africa’s small share of world allocations. They do not explain why three African states hold roughly 55% of the sovereign-state total while the bottom thirty hold only 2.74%. They do not explain why one large European or American operator can still outweigh dozens of weaker African states combined. Those outcomes were produced by a much larger structure over a much longer period.

In global terms, 10 million is not the system. It is a small fraction within it. To turn that figure into the emblem of inequality while treating the system that allocated the other 99.77% as morally authoritative is not serious reasoning. It is a refusal to look at where power actually sat.

A serious pro-poor agenda

Once the false comparison is removed, the reform path is not difficult to see.

Keep the shared uniqueness layer as thin as possible: the part that actually ensures uniqueness, basic registry integrity, and accurate records. That is what the system was built for. Do not ask a numbering bureaucracy to perform global social policy by way of friction, delay, and discretionary review. It was never designed for that, and the results show it.

Then remove the tolls that make connectivity more expensive precisely where it is already weakest. Reduce discretionary friction. Clarify title. Make transfer and leasing rules cleaner. Make prices and processes more transparent. Strengthen verification, registration accuracy, and routing security where transfers occur. In other words: keep the coordination function; cut the regressive overhead.

And if governments genuinely want redistribution, they should redistribute. They should subsidize power, backhaul, fibre, devices, education, and a competitive market structure. They should lower the real costs of connectivity directly. What they should not do is bury a regressive cost structure inside the registry procedure and then call the result equality.

Allocation is not justice. Administration is not solidarity. A system does not become pro-poor merely because it dislikes the language of markets. After more than three decades, the RIR model has no credible claim to being the world’s equalizer. In practice, it has done something narrower and harsher: it has formalized existing inequality, imposed a regressive layer of gatekeeping on top of it, and asked the weakest actors in the system to accept that burden as a moral principle.

That is the poverty penalty. The poor do not need moral theatre about why access must remain constrained. They need cheaper, faster, more predictable access.