Infrastructure Has Never Had Its Penny

In about 2006 I played a lot of World of Warcraft from New Zealand — and the thing that mattered was not download speed. It was ping — the round-trip time to the server, the gap between pressing a key and the world responding. Mine was bad. Bad enough that the game was playable but every part of it that required timing was not.

The advice going around the forums was to route through a VPN in the United States. That sounds backwards. You are adding a hop, adding encryption overhead, adding a middleman. It worked anyway. Not always, but often enough that people paid a monthly fee for it.

The reason is unglamorous. My traffic was being routed on commercial rules — cheapest available path, settled between carriers on terms that had nothing to do with me and were never going to be revisited because I was playing a game. A better path across the Pacific existed. It was carrying other people’s packets. There was no mechanism by which my session could say this one is latency-sensitive, and here is what I would give up to get it — and no mechanism by which the network could have answered.

So I paid a third party a monthly fee to reach infrastructure my own provider was already connected to.

I did not have the language for it then, but that is the argument of this article in one bad connection. The capacity existed. The constraint was commercial, not physical. The allocation rule was static, blind, and set somewhere I could not see — and the only way to improve my outcome was to pay rent to someone who had positioned themselves next to the rule.

It was also layered, which turns out to matter more than the story.

My local ISP knew its own congestion. As did everyone who worked there including me.

The transit and peering carriers in between knew their route costs.

The game provider knew its server load and its tick rate. Each was optimising correctly at its own level. None of them shared state with the level above or below.

My ping was not any single party’s failure. It was the sum of everyone’s local optimisation on a system with no shared record.

Hold that shape. Everything that follows is the same shape, drawn on a power line.

Because it is now happening to the electricity network, at a scale that will decide what the energy transition costs.

Rooftop solar, batteries, EVs and flexible machines are arriving on a distribution grid built to move power one way, from a large plant to a passive customer. When a low-voltage feeder reaches its limit, someone’s export stops. Nobody is told. Nothing is recorded. There is no queue, no turn-taking, and no way to know whether it was the same house last week. The lights stay on, which is the job — but who gave way, when, and whose turn it is next lives nowhere. The physical capacity is usually there. The record is not.

That is not a small operational gap. It decides who pays for the next decade of network investment, who captures the value of capacity already built and paid for, and whether the households least able to adapt end up subsidising the ones best placed to react.

All three questions are downstream of a smaller one, and it sounds trivial beside them: what is the smallest piece of this network that anyone can name, count and hand to someone else? Nothing that big gets settled until something that small exists.

⁂

No one grows up wanting to be a penny.

A penny is the smallest unit of currency: easy to overlook, almost impossible to spend on its own, and yet capable of becoming a scarcity signal when enough of them are missing from your future. What a penny actually does is make value divisible enough to be seen. Before uniform penny post, the cost of sending a letter was a negotiation conducted at the door; afterwards it was a number printed on a stamp that anyone could read.

The penny is also, as of last year, a historical object. The US Mint struck its last circulating cent in Philadelphia in November 2025, 238 years after the first. Canada retired its own in 2012; Australia let the one and two cent coins go in 1992. Japan still issues the 一円玉 (ichi-en dama), a coin that costs more to produce than it is worth, kept alive largely by consumption-tax arithmetic.

The lesson is not that small units stop mattering. It is that a unit of account earns its place only while it stays proportionate to what it prices. The penny did not die of irrelevance; it died of inflation. A unit stops being minted when it becomes too small to matter. Infrastructure has the mirror-image problem: its unit is too small to see. Nothing has ever been minted fine enough to make a feeder’s spare capacity visible — let alone fine enough to be retired.

Infrastructure has never had its penny.

We are told we need more of everything — generation, wires, bandwidth, data centres, roads, water.

And yet, at the same time, enormous amounts of existing capacity sit unused.

Solar energy is curtailed while electricity prices collapse. Transmission lines have spare capacity for much of the day. Telecommunications networks have unused bandwidth. Roads alternate between congestion and empty lanes. Data centres can have computing capacity sitting idle while other users face scarcity.

The problem is not always a shortage of physical resources. It is that we have no unit small enough, local enough or current enough to see what we already have.

Here is what having one would look like:

Feeder 742 is in surplus from 11:20 to 14:40. But a constraint at the substation above it means 38% of the solar its households could produce will never be consumed by anyone.

That sentence is not currently sayable about most of the low-voltage network in any developed country. Not because it is false, but because nobody holds the record that would make it true.

Two recent papers make it sayable, and they arrive at it from opposite ends of the same problem. Peter Kilby is on both, which is not a coincidence — they are two halves of one architecture.

The first, by Kilby with Hamish Bissett, Brad Smith and Brendan Banfield in a recent CIGRE C6 paper, examines model-free Dynamic Operating Envelopes: using network observations to determine how much distributed energy resource capacity the physical network can actually accommodate as conditions change. It asks what the network can do, right now — and answers it without needing a perfect model of every parameter underneath.

The second, Shaun Sweeney’s stateful pricing and allocation manuscript with Kilby et al., asks who should get it, and on what terms — when that same question has to be answered again every five minutes, for a year, on a network that remembers what it did last time.

One paper mints the unit.

The other produces the market that can spend it.

Put together, they do for network capacity roughly what the penny did for postage: take something everyone needed, nobody could price, and everyone therefore argued about — and turn it into a unit small enough to allocate, and legible enough to contest.

And that convergence points at a much bigger question:

What happens when the physical state of infrastructure becomes an explicit economic object?

This article makes one argument in five moves.

One: infrastructure has never had a unit small enough to see itself with, and two recent papers, taken together, produce one.

Two: the capacity we believe we have is not a number but a state — conditional, local, changing hourly, and at low voltage almost entirely unobserved.

Three: once the state is visible, someone has to decide who gets the residual. That is no longer an engineering question, and the networks already ran this experiment once, in 1986, and solved it with something that looks nothing like a market.

Four: whatever mechanism allocates the residual generates a rent, and that rent will be captured by somebody. The four candidates are worth naming before the default wins.

Five: the record is the institution. Whoever holds it holds everything downstream of it — the investment signal, the financing, and the answer in the title.

The finding that disciplines all five sits in the middle, and is worth stating up front rather than as a late caveat:

• making scarcity visible and then pricing it,
• on a system with no memory of who gave way last time,
• produced worse outcomes than doing nothing.

State has to come before price.

Every argument here depends on that order, and any version of this that drops it is pricing evangelism with better instrumentation.

⁂

It may also help with one of development’s most persistent problems: financing infrastructure before its value is visible.

In 2007 I worked in telecommunications through the era of unbundling and network access. The central question was never simply who owned the copper, fibre or switching equipment. It was who could see the network, who could access it, under what conditions, and how the value of shared infrastructure should be allocated. I have watched this argument before, in another industry, with different acronyms and the same shape — and there it was settled in favour of whoever held the operational data.

That experience left a durable lesson: a network can be physically open while remaining economically opaque.

The same lesson appears in a different form in postal infrastructure. The Penny Black — the first adhesive postage stamp, and the beginning of uniform penny post — evokes the original standardised access system, one that made communication dramatically more scalable by making its price legible and universal. Yet the old postal network is now dying in many places, even as prices rise for the people who still use it. As volume leaves the system, fixed infrastructure and obligations are spread across fewer transactions. The network becomes more expensive precisely as its usefulness becomes less visible. Rowland Hill’s case (1837) turned on a point close to this one: the cost of assessing and collecting a variable price exceeded the cost of carriage.

The lesson is not that physical networks inevitably fail. It is that networks without sufficient visibility, adaptability and transparent allocation can become progressively less useful, more expensive and harder to finance.

A contemporary project makes the same point from the opposite direction. Microsoft Research’s Penny Black project explored whether email could be made more accountable by attaching a tiny computational cost to each message — effectively a digital postage stamp. It was not a new idea even then: Cynthia Dwork and Moni Naor proposed computational pricing against junk mail in 1992, and Adam Back’s Hashcash implemented it in 1997.

The lineage matters, because it runs directly onward to Bitcoin’s proof-of-work. The primary sources are Dwork and Naor (1992) and Back (1997, 2002); the Microsoft work is Abadi, Birrell, Burrows, Dabek and Wobber (2003).

Their follow-on problem is worth noting: processor speeds varied so widely between senders that a CPU-priced stamp fell unequally, which pushed the same researchers toward memory-bound puzzles (Dwork, Goldberg and Naor, 2003). Heterogeneous capability breaking a uniform price is not a new discovery.

The idea is not to charge users for the value of an email, but to make mass automated sending slightly more expensive, so that spam, abuse and bot-generated volume become economically less attractive. A legitimate person sending a handful of messages barely notices the cost; a system sending millions must internalise a burden that was previously imposed on everyone else.

The implication is important. Even in a network where the marginal cost of transmission is close to zero, governance may require a mechanism that makes scarce resources, externalities and access conditions visible. Proof-of-work does not solve email through more bandwidth alone. It changes the economics of participation by linking access to a measurable state or obligation.

Bitcoin’s proof-of-work descends directly from Hashcash. The white paper says so. That much is lineage rather than analogy, and it is worth stating plainly before drawing any looser comparison. Nakamoto (2008) cites Back’s 2002 Hashcash paper directly, as its sixth reference. Dwork and Naor are not cited there at all; the 1992 origin reaches Bitcoin at one remove, through Back.

The looser comparison is still the useful one. What connects proof-of-work, Dynamic Operating Envelopes and stateful markets is not descent but shape: each answers a coordination problem in which open access, limited resources and imperfect trust generate costs that adding capacity cannot fix.

Proof-of-work makes mass sending carry a small computational burden. Bitcoin’s consensus rule makes the ordering and settlement of transactions costly to manipulate without requiring a central authority to approve every transaction. Dynamic Operating Envelopes make physical network headroom observable. Stateful allocation mechanisms make repeated access decisions responsive to changing constraints. In each case, the system attempts to replace an opaque or centralised assumption with a measurable rule tied to system state, resource use or the cost of imposing externalities.

The differences matter just as much as the similarities. Email proof-of-work addresses abuse; Bitcoin addresses decentralised consensus and double-spending; Dynamic Operating Envelopes address physical feasibility; stateful markets address allocation and settlement. The shared insight is narrower but important:

When resources, attention or trust are scarce, a network may need a credible mechanism that makes participation, priority and externality legible.

That is the same architectural question emerging in energy and infrastructure: when capacity is abundant for some users and scarce for others, how do we make the cost, priority and consequences of access legible? The answer need not always be a literal price. It may be a computational burden, a service obligation, a capacity right, a queue position or a state-contingent payment. But without some mechanism for distinguishing valuable use from socially costly volume, abundance can be overwhelmed by unpriced demand.

The same pattern appears across electricity, telecommunications, transport, water and data infrastructure.

01The capacity we think we have isn’t the capacity we actually have

Infrastructure is usually represented through static quantities.

A transformer is rated at X MVA. A feeder has a thermal limit. A connection has an export limit. A telecommunications link has a bandwidth rating. A road has a nominal vehicle capacity.

These are useful engineering properties. But they aren’t the same as available capacity.

Available capacity is conditional. It depends on what else is happening.

For an electricity network, it can depend on generation; demand; voltage; thermal loading; topology; storage; weather; neighbouring network conditions; and the behaviour of other connected devices.

Consequently:

Available Capacity = f(Physical State, t)

The available capacity of an infrastructure system is therefore not a fixed number. It is a state. The general form of this argument is old. Schweppe, Caramanis, Tabors and Bohn (1988) made price a function of system state, time and location at transmission level. What is new is the extension to low voltage, where feasibility has to be inferred rather than assumed, and where the allocation problem carries memory.

The easiest way to understand this is a road. A road may be physically capable of carrying a certain number of vehicles per hour, but the number it can safely and reliably accommodate depends on weather, accidents, traffic flows from adjoining roads and the behaviour of other drivers. The road’s capacity is not simply a property of the asphalt. It is a property of the road in a particular condition. The economics of that road is Vickrey (1969).

This is the conceptual importance of Dynamic Operating Envelopes.

Kilby et al.’s work demonstrates that network behaviour and visibility can be used to infer operating capability without requiring perfect knowledge of every underlying network parameter. In their simulations the comparison is three-way, and the ordering matters more than any single figure: no envelope at all, a conventional model-based envelope, and a model-free envelope inferred from observed network behaviour. The model-free approach curtailed 0.2% of available energy against 3.9% for the model-based envelope, in a modelled network rather than a deployed one — and both sit far below what a static export limit gives up. The headline is not the 95% improvement over the modelled case. It is that you do not need a complete parameter model of the low-voltage network in order to beat one. What stands between a feeder and its own spare capacity is not missing physics. It is missing observation. For the Australian policy context, see ARENA (2022a, 2022b); the inverter-side requirements are in AS/NZS 4777.2:2020.

The implication is bigger than the particular result. Capacity that previously appeared unavailable can become usable when the system can measure its actual state. The value gain is therefore not only additional energy delivered. It is the recovery of economic value from an asset that already exists.

A useful analogy here is an airport. The airport may have a fixed number of runways, but its usable landing capacity changes with weather, aircraft mix, runway configuration and the spacing required between arrivals. A static runway limit is not the same thing as the number of safe landing slots available at 2pm on a particular day. Dynamic Operating Envelopes perform a similar function for the grid: they estimate the safe operating room that exists now, rather than relying only on a permanent worst-case limit.

02But knowing capacity exists isn’t enough

Suppose the network determines that 10 MW of additional capacity is available. There are 15 MW of potential users.

Who gets the 10 MW?

That question has a street address.

Take an ordinary suburban street on a clear spring afternoon. Number 14 has rooftop solar and is exporting. At three o’clock, number 19 plugs in an EV and starts a 7 kW charge. Voltage on the shared low-voltage feeder rises past its statutory ceiling, and number 14’s inverter — doing exactly what it is required to do — throttles its export, then stops.

Nobody did anything wrong. Nobody broke a rule. No message was sent. Number 14 loses an afternoon’s export credit and will never learn why. Number 19 gets a full charge and will never learn that it cost anything. When it happens again next week, nothing in the system knows that it is happening again, nothing knows that it was number 14 last time, and nothing could have offered number 19 a slower charge for a smaller bill.

The same layering runs down that street as ran through my connection to a game server twenty years ago: house, feeder, substation, zone. Every level operating correctly. No level holding a record of who gave way.

This is where engineering ends and economics begins. An operating envelope establishes feasibility. It doesn’t necessarily establish allocation. For the general framing of these choices, see Wilson (2002).

You can allocate equally. You can allocate according to priority. You can allocate according to willingness to pay. You can allocate according to historical service. You can use some combination of these.

But once multiple parties compete for constrained physical capacity, you have created a market problem.

This is the problem addressed by the stateful allocation work described above. The important innovation isn’t simply putting a price on electricity. It is treating repeated constrained allocation as a stateful cyber-physical problem. The mechanism can account for physical feasibility, import scarcity, export congestion, bounded pricing and historical service outcomes.

The market therefore doesn’t exist independently of the network. It is coupled to the network’s state.

And here is the result that should discipline every claim in this article. In that work, adding a price signal to an operating envelope without intertemporal state did not merely fail to help — it made outcomes materially worse than the envelope alone, both in unserved energy and in the fairness of outcomes at the weakest feeders. A price signal on a system with no memory rewards whoever is best positioned to react in this interval, and then again in the next one, and the disadvantage compounds at exactly the locations that were already worst served.

That finding cuts against the intuition this whole article might otherwise be taken to support. Visibility plus price is not automatically an improvement. State has to come before price. Any argument for making infrastructure scarcity legible and priceable has to carry that caveat, or it is simply pricing evangelism with better instrumentation.

03The part of Bitcoin worth borrowing isn’t the ledger

The line from Hashcash to proof-of-work is descent. What follows is analogy, and should be read as such — but it is the analogy that most directly supports the caveat above.

Proof-of-work is the part everyone quotes. The part that actually holds the system together is the difficulty adjustment. Every 2016 blocks — roughly a fortnight — the network compares how long those blocks actually took against how long they were meant to take, and re-prices access accordingly. Blocks arriving too fast, the required work goes up. Too slow, it comes down. The target it defends is physical: an average interval of ten minutes, held for seventeen years while the computational effort behind it grew by something like fourteen orders of magnitude.

Note what that mechanism is not. It is not a price discovered by bidding. It is a price derived from measured system state, over a defined look-back window, applied on a lag and bounded in how far it may move in a single step. Memory is not incidental to the design; it is the design. A version of Bitcoin that retargeted on the last block — the instantaneous signal, no state — would oscillate violently and be trivially gamed by whoever could swing hashrate fastest within one interval.

That is the same failure described above at the weakest feeders, arriving from a different discipline and a different decade. A price signal without memory rewards whoever is best positioned to react now, and then rewards them again. It is worth knowing that the most durable open-access mechanism of the last two decades reached that conclusion and built for it from the first block.

The closer precedent is older and less glamorous. In October 1986, the link between Lawrence Berkeley Laboratory and UC Berkeley collapsed from 32 kbit/s to 40 bit/s — a loss of throughput by a factor of eight hundred on a network that was physically intact. The problem was not capacity. It was that every sender, acting reasonably on its own information, kept transmitting into a queue that could not clear.

Van Jacobson’s answer, published in 1988, has descendants running on the device you are reading this on. Each sender keeps a congestion window: a per-flow state variable holding how much it may have in flight. The window grows additively while packets arrive and halves multiplicatively on loss. Additive increase, multiplicative decrease. No auctioneer, no central allocator, no money, and no sender knowing anything about any other sender. The participants hold the state, and approximate fairness emerges from the rule.

That is a stronger claim than this article has so far made. TCP does not price congestion. It allocates a hard-constrained shared resource, at scale, between parties with no relationship and no trust, using nothing but per-participant memory and a feedback rule.

Set the four mechanisms out and an axis appears.

Proof-of-work: cost, adjusted by state.

TCP congestion control: state, and no price at all.

Stateful allocation of network capacity: state, with price.

An operating envelope carrying a spot price and no memory: price, and no state — the configuration that made outcomes worse than the envelope alone.

Read in that order, the finding above stops being a caveat and becomes the middle of a range. Statelessness is the failure, not pricing. And the member of the family that has run continuously for nearly forty years is the one that dispensed with price entirely.

There is a detail here that should be uncomfortable for anyone working on grid market design. The fairness metric used in the stateful allocation work — Jain’s index — was defined at Digital Equipment Corporation in 1984 (Jain, Chiu and Hawe, 1984), for congestion avoidance in computer networks. Chiu and Jain (1989) later published the convergence analysis for AIMD itself. The same researchers gave networking both the mechanism and the ruler. A fairness index reported on an electricity feeder is an electricity result scored with a networking instrument, on a problem networking treated as solved before most of today’s distribution engineers started work. The grid is not pioneering here. It is arriving late, with borrowed tools, at a problem that already has a literature.

The borrowing needs one correction, or a networking reader will supply it. TCP’s fairness is not clean. AIMD converges to equal shares only under roughly equal round-trip times and synchronised loss; in practice, flows with shorter round-trip times take systematically more. Position in the topology determines share, independently of merit. That is not a reason to drop the analogy — it is the analogy. Round-trip-time bias is the internet’s version of the weakest-feeder problem, and it was never solved. It was mitigated, argued about, and lived with.

Which brings the useful objection, and it has a name.

Saltzer, Reed and Clark’s end-to-end argument — the 1984 paper that shaped the internet’s architecture — holds that function belongs at the edges, and the network in the middle should stay as dumb as it can. It was right, and the internet is largely its vindication. Applied here it says the opposite of everything above: keep the grid a dumb pipe, put the intelligence in the inverters and the batteries, and stop trying to build state into the network.

Why it does not transfer is the load-bearing difference between the two systems. The internet’s constraint is congestion, and congestion is soft. A dropped packet is recoverable — the endpoints retransmit, and the cost of being wrong is latency. Electricity’s constraints are hard, local and physical. Voltage rises at a particular point on a particular feeder. Thermal limits bind on a particular asset. There is no retransmission, and the cost of exceeding the limit is not delay but damage or disconnection.

An endpoint therefore cannot compute its own feasible action. A household cannot infer, from anything it observes at its own connection, whether its export is the one that pushes the voltage at the pole outside a neighbour’s house past the limit. Feasibility is a property of the level above it, and of the level above that. So state has to live in the network, and at every level, because every level has its own binding constraint.

The second lesson is less comfortable, and it concerns what happens after deployment.

Bitcoin’s block size limit was one constant in one line of code. Once hardware, capital and revenue depended on it, changing the number directly proved politically impossible. The argument ran for years and was resolved by a technical workaround and a chain split rather than by amending the parameter. Nothing about the change was hard to implement. It had simply become load-bearing for too many balance sheets to move.

Every parameter in a stateful allocation mechanism is a candidate for the same fate. The fairness weights. The price bounds. The look-back window. The priority ordering at a constrained substation. These read as engineering choices during design and become distributional politics the moment revenue attaches to them — which is to say immediately after deployment, and permanently thereafter.

The implication for anyone proposing to make infrastructure scarcity priceable is direct. The amendment procedure is part of the mechanism, not an administrative matter to be settled later. Who may change the fairness parameters, on what evidence, with what notice, and with what standing for parties who invested against the old rule — those questions are cheap to answer now and expensive to answer once the mechanism is carrying money.

A simple analogy for what the stateful version does instead is a hotel with a changing number of usable rooms. The hotel may have 100 rooms in total, but a burst pipe, a power outage or a maintenance problem may reduce the number available tonight. If 120 guests want rooms, the hotel needs more than a total-room count. It needs a current inventory, rules for priority, a record of who was turned away last time, and a way to update prices and reservations as conditions change. A stateful market is closer to that live inventory system than to a printed brochure listing the hotel’s maximum capacity.

That gives us a fundamentally different loop:

Physical State → Feasible Capacity → Allocation → Price → Physical Response → New State

The market becomes part of the physical feedback system.

The second paper’s value gain is therefore different from the first. The Dynamic Operating Envelope recovers physical headroom by improving observability and inference. The stateful mechanism recovers allocative value by ensuring that repeated access decisions respond to changing constraints rather than relying on a static rule.

One paper asks how to make more of the network usable. The other asks how to make use of that capacity more economically coherent.

Together, they suggest that the value of digitalisation is multiplicative rather than merely additive. Stated carefully: the economic value released depends jointly on how much of the physical state can be observed, how well the residual capacity is allocated, and how credibly the resulting obligations settle — and a weakness in any one of the three discounts the other two. If the state is invisible, good allocation rules have nothing to act on. If allocation is arbitrary, visibility only reveals unfairness. If settlement is not credible, neither visibility nor allocation will attract capital.

04Why forward curves break when capacity is stateful

This is also where a central instrument of conventional energy markets begins to fail: the forward curve.

Forward curves work best when the commodity being priced is sufficiently standardised and the future delivery obligation can be described in advance. A forward contract can specify a quantity, a location, a delivery period and a price.

But DER capacity is not a static commodity. Its availability depends on weather; local voltage; feeder topology; coincident demand; the behaviour of other DER; storage state; network constraints; and the operating decisions made by other participants.

The capacity available tomorrow at noon may not be the capacity available tomorrow at 6pm. The capacity available at one connection point may not be substitutable for capacity at another. A forecast of future capacity can therefore be materially different from the capacity that is physically deliverable when the contract settles.

Under the current market design, these differences are often compressed into static connection limits, standardised products, ex ante forecasts and imbalance or curtailment rules.

That creates several problems.

First — the price is not the deliverable. The forward price can represent an expected energy value without representing the probability that the underlying DER can actually deliver at the relevant location and time.

Second — attractive and infeasible are not exclusive. A forward position can be economically attractive while being physically infeasible once local network conditions change.

Third — the risk lands on whoever balances. The risk of non-delivery is often pushed into penalties, curtailment, uplift payments, bilateral renegotiation or the network operator’s residual balancing role.

Fourth — the signal is averaged where the constraint is local. The curve can become a poor signal for investment because it prices an averaged or system-wide expectation while the binding constraint is local, dynamic and state-dependent.

In other words:

Forward Price ≠ Deliverable Local Capacity

The problem is not that forward markets are useless. They remain valuable for hedging fuel, energy and some forms of predictable capacity risk. The problem is that a conventional forward curve assumes a degree of fungibility and predictability that DER often does not possess. Transmission-level markets already hold a partial answer in financial transmission rights (Hogan, 1992), which define a financial instrument against network feasibility rather than against a commodity. The open question is what the low-voltage equivalent looks like, where the feasible set is inferred, moves within the day, and is not underwritten by a central dispatch model.

A stateful market addresses this by making the physical state part of the tradable and settlement-relevant object.

Instead of asking only “what is the expected price of electricity in this future period?” the market can also ask: “what capacity is expected to be serviceable at this location, under which network states, with what probability, and subject to which operating constraints?”

The resulting instrument is less like a single static forward price and more like a state-contingent capacity claim, whose value depends on location, time, network state, serviceability requirements, delivery probability, flexibility, and the participant’s contribution to future system conditions.

This does not eliminate uncertainty. It makes uncertainty explicit, measurable and allocable.

A Dynamic Operating Envelope provides the physical estimate of what can be delivered. A stateful allocation mechanism determines who receives that capacity and at what price. Settlement can then be linked to the realised state rather than relying entirely on a static ex ante promise.

The forward curve is no longer asked to perform a task it cannot reliably perform: compressing a changing local physical frontier into one scalar price. Instead, the market can produce a family of state-dependent prices, capacity rights and flexibility obligations.

That is how the architecture fixes the mismatch:

Observed State → Dynamic Envelope → State-Contingent Capacity Claim → Allocation and Settlement

The result is a more honest market design. It distinguishes expected energy value from deliverable network access. It distinguishes a financial hedge from a physical operating right. And it allows the price of scarcity to update as the physical system changes, rather than leaving the network to absorb the difference between the forward promise and the real operating state.

05This is where abundance stops being a quantity

We normally think of abundance as a quantity. More electricity. More bandwidth. More compute. More transport capacity.

But there is another possibility. Abundance is partly a coordination property.

Consider 10 MW of excess solar. Physically, it exists. Economically, however, it may not.

If the network cannot observe it, it is inaccessible. If the network cannot accommodate it, it is curtailed. If the network can accommodate it but cannot allocate access dynamically, it remains underutilised. If the capacity can be observed, made serviceable and allocated, new economic activity can emerge around it.

So there is a progression:

Physical surplus → Observable surplus → Serviceable surplus → Allocatable surplus → Economic abundance

This suggests a provocative proposition:

Economic abundance is not simply a function of how much physical capacity exists. It is also a function of how much of that capacity can be observed, governed, allocated and utilised.

That doesn’t mean software creates energy. It means poor coordination can destroy the economic usefulness of energy that physically exists.

It is worth being precise about what this adds to the current abundance debate, because there is one. Ezra Klein and Derek Thompson’s Abundance (2025) makes a compelling political case: the constraint on housing, energy and infrastructure is not the absence of resources but the accumulated procedural incapacity of the state to build. They treat abundance as a condition — a state of affairs in which there is enough of what people need — and scarcity as its political opposite, to be escaped by restoring the capacity to build.

That framing is right as far as it goes, and it has one gap. In their account, abundance is an aspiration measured after the fact, in aggregate, at national scale. It is not something you can read off an instrument on a Tuesday afternoon.

The argument here is narrower and more operational: abundance can be a measured state variable, local and temporal, defined over available energy, flexibility and network capacity net of demand and constraints, resolved to a node and an interval. Not “the country needs more energy,” but the Feeder 742 sentence at the top of this article. Klein and Thompson describe the destination. What is proposed here is the instrument that tells you where you actually are — and, more usefully, which specific constraint is binding at which specific place.

The same is true of a telecommunications network. Unbundling did not create copper, fibre or spectrum. It changed who could access network capability and under what rules. But without transparent information about congestion, quality and available headroom, unbundling could only go so far. Access can be formally open while practical capacity remains hidden behind engineering assumptions, queues and opaque operating decisions.

Digitalisation matters because it turns a network from a largely static asset into an observable process.

A useful everyday analogy is a supermarket. Food may exist in the warehouse, but it is not economically available to shoppers if the inventory system cannot identify it, the shelves cannot receive it or the checkout cannot allocate it. Physical stock becomes usable abundance only when the system can observe, move and distribute it. Infrastructure works similarly: capacity that cannot be seen or allocated behaves, for practical purposes, like capacity that does not exist.

06Scarcity and abundance as states

This gives us a different way to think about scarcity.

Scarcity doesn’t have to be defined exclusively through price. It can be defined through the physical state of the system. Likewise, abundance can be defined through the existence of residual serviceable capacity.

Scarcity ↔ Balance ↔ Abundance — one axis, continuously measured.

The important point is that the state is local and temporal.

A feeder can be scarce at 6pm and abundant at noon. A telecommunications network can be congested in one cell while having spare capacity in another. A data centre can be power-constrained while having idle compute elsewhere. A road can be congested during the morning peak and effectively abundant at midnight.

A postal network can have universal coverage but declining economic abundance: the routes and sorting infrastructure remain, while the volume needed to support them disappears. The result is a familiar but underappreciated inversion. Prices rise not because each additional letter consumes more physical network, but because fewer users remain to share the cost of maintaining the network.

Abundance is therefore not a property of the infrastructure in the abstract. It is a property of the infrastructure at a particular place and time, under particular service requirements.

07The machinery is a holarchy

Every example so far has the same shape, and the shape is worth naming, because the name does real work.

A household is a complete system in itself — appliances, storage, a roof, its own purposes — and simultaneously a component of the feeder it connects to. The feeder is complete in itself, and a part of its substation. Substation into zone, zone into region, region into system. Arthur Koestler called such a unit a holon: something that is a whole and a part at once, whose behaviour is governed by the tension between asserting its own integrity and integrating with the level above it. A structure of nested holons is a holarchy. Koestler sets out the properties in The Ghost in the Machine (1967), Appendix I. The engineering-language version of the same claim is Simon (1962), on near-decomposable systems.

Infrastructure is a holarchy. So is any market capable of coordinating it. Shaun Sweeney’s thesis makes this explicit: its architecture is holarchic, with scarcity propagating through nested layers from node to cluster to zone to region to system, and nodal and zonal pricing emerging as special cases of a single propagation rule rather than as rival designs.

Naming the shape resolves three things that otherwise read as separate arguments.

First, it explains what a serviceability floor actually is. Not an arbitrary political intrusion into a market, but the ordinary business of a nested system: the level above setting the boundary conditions within which the level below is free to operate. Constraints travel down; state travels up. Every functioning holarchy works this way, and the ones that fail are the ones where the traffic only goes one direction.

Second, it locates the rent. If scarcity is created at a particular level — a congested feeder, a constrained substation — then the value of relieving it is created there too, by the participants connected there. That is a considerably stronger argument for recycling congestion value at the level that generated it than for sweeping it up into a national revenue cap. And it is only computable if each level keeps its own record.

Third, it explains the failure mode with unusual precision. A holarchy fails at whichever level stops reporting its state. Unbundling granted legal access to a network whose live condition stayed private, so competition took place one level above the information it needed. The postal network kept universal coverage while the economics of its lower levels quietly collapsed, and the price rose without anyone deciding that it should. The electricity market prices the top of its own holarchy exquisitely — five-minute dispatch, hundreds of nodes — and the bottom not at all. In each case the structure survived. The reporting did not.

Which reframes the entire architecture in a sentence: the task is not to build a new market on top of infrastructure. It is to give every level of an existing holarchy a state it can report, and a rule for what to do with what it receives.

08Governance decides what abundance is available to the market

This is where the economics cannot be separated from governance.

Not every unit of capacity should simply go to the highest bidder. In the limit the point is obvious — no one proposes that a hospital bid against a data centre for supply during an emergency, and no jurisdiction allocates emergency supply that way.

But the real cases are less dramatic and much more common. On a constrained feeder at 1pm, whose export gets curtailed: the household with rooftop solar and no battery, the neighbour with both, or the commercial site charging a fleet? At 6pm, whose import is limited: the heat pump keeping a house at 18 degrees, or the second EV topping up for a weekend trip? These are not emergencies. They are ordinary Tuesday allocation decisions, they happen thousands of times a day, and at present they are resolved by whichever inverter happens to hit its voltage limit first.

Governance therefore establishes a serviceability floor — the minimum service each participant is entitled to regardless of price. The design principles for this kind of collective allocation are set out in Ostrom (1990), in particular that the rules, the monitoring and the dispute resolution have to be designed together rather than sequentially.

Physics determines what is physically possible at that moment. The market determines how the residual — everything above the floor, within the frontier — is allocated.

This creates a useful division of labour:

Governance defines the floor.

Physics defines the feasible frontier.

Markets allocate the residual.

That separation is important. The market doesn’t need to decide whether hospitals should receive priority. Society decides that. The market’s job is to coordinate everything that remains after those requirements have been satisfied.

And governance needs one thing more, which is far cheaper to specify now than later: a rule for changing its own rules. The serviceability floor is a number. So are the fairness weights, the price bounds, the look-back window and the priority ordering at a constrained substation. Each will be set by someone, on some evidence, at some moment — and each begins accruing beneficiaries the day it is switched on. The block size lesson applies without modification: the difficulty of changing a parameter has almost nothing to do with the difficulty of changing the code, and almost everything to do with how much capital has been committed on the assumption that it will not change. So the design question is not only where the floor sits. It is who may move it, on what evidence, with what notice, and with what standing for participants who invested against the old level. A mechanism that specifies the floor but not the amendment procedure has specified half a rule.

But governance also needs information. A serviceability floor cannot be enforced intelligently if the state of the network is invisible. Regulators and public institutions need to know whether a provider is genuinely constrained, whether capacity is being withheld, whether service quality is deteriorating, and whether scarcity is physical or administrative.

This is why digitalisation and transparency are not merely efficiency tools. They are governance infrastructure.

09Then comes the question nobody likes asking: who owns the rent?

This is where an older economic idea becomes unexpectedly relevant. Geoism.

The Georgist tradition is concerned with economic rent: value arising from control over scarce resources rather than from productive activity itself. George’s argument is in Progress and Poverty (1879); its formal statement is the Henry George theorem (Arnott and Stiglitz, 1979), which shows that under specified conditions aggregate land rent equals optimal expenditure on public goods. That is a direct claim about financing infrastructure out of the rent it creates.

Modern infrastructure creates a remarkably similar problem.

A network operator may legitimately earn a return on capital; maintenance; operations; reliability; engineering; and risk.

But infrastructure can also create scarcity rents. If access to a constrained network is valuable, whoever controls the allocation of that access can capture economic value.

And this creates a subtle problem with natural monopolies. The issue isn’t necessarily “one company owns the wires.” The deeper issue may be “one company owns the wires and controls the information and allocation mechanism determining who can use their scarce capacity.”

Those are different forms of power. The first may be economically efficient. The second can create substantial information asymmetry and rent extraction.

This has already happened once, in a network, within living memory. IPv4 addresses were free. They were allocated on request to anyone who could justify a need, because there was no scarcity worth pricing. When the free pools ran down, a market appeared: addresses now transfer between organisations at real prices, recorded by the same regional registries that had handed them out. Nothing about the addresses changed. What changed is that scarcity became measurable and ownership became registrable — and the registry, which existed only to keep a record, turned out to be the institution that made a market possible.

The uncomfortable part is who received the rent. It went to organisations holding large legacy allocations from the era when addresses were free — universities, early corporations, agencies — because ownership had been settled long before scarcity was priced. Nobody decided that. It was the residue of an administrative convenience.

That is what this section is really about. The rent appears whether or not anyone plans for it. The only question that stays open is who holds title at the moment it becomes visible, and that is usually answered years earlier, for unrelated reasons.

The same issue appears in telecommunications, postal systems and other network industries. Ownership of the physical asset is visible. Control over the data, queue, interface, standard, tariff and operating rule is often less visible, even though it may determine who benefits from the network.

A familiar analogy is a toll bridge. Owning the bridge is one form of control. Controlling the traffic data, deciding which vehicles receive priority and changing the toll without revealing the basis for the decision is another. The second form of control can be just as economically important as ownership of the concrete and steel.

So: who should own the rent? The question in the title deserves an answer rather than an elegant evasion, and there are only four honest candidates.

The network owner may keep it, which is the current default. This is defensible only to the extent the rent reflects genuine risk-bearing and investment. It becomes indefensible when the rent arises from an information asymmetry the owner also controls — because then the party pricing the scarcity is the same party deciding whether the scarcity exists.

The community may capture it, which is the orthodox Georgist answer: the value arises from a shared network and a shared physical constraint, not from any one participant’s effort, so it should be socialised — through the regulated revenue cap, through a rebate, or through the rate base.

The participants who gave way may receive it, which is the answer I find most persuasive and least explored. When a feeder is constrained, someone is curtailed so that someone else can proceed. The scarcity value is created jointly by everyone connected, and realised at the expense of whoever stood down. Recycling the rent to those participants — in credit, in future priority, or in both — turns rationing into something closer to a reciprocal obligation, and it is only computable if the system keeps a record of who gave way and when. Which returns us to state.

Nobody may capture it, in the sense that the correct long-run response to persistent scarcity rent is to invest until it disappears — the rent read as a construction signal rather than an income stream.

These are not mutually exclusive; a mature design would use the floor to protect the first tranche, recycle the residual to those who gave way, and treat persistent rent as an investment trigger. But the choice is a political one, and it should be made deliberately rather than settled by default in favour of whoever happens to hold the meter data.

What is not optional is the precondition. None of these four answers is available while the rent is invisible. You cannot socialise, recycle, tax or compete away a rent that nobody can measure.

10Digitalisation can separate ownership from allocation

This is perhaps the most consequential implication — though it is important to be accurate about what is new here, because the principle itself is old.

Separating ownership of a network from control of access to it is the foundation of thirty years of regulatory economics: the essential facilities doctrine, local loop unbundling, open-access transmission, and the long-running debate about distribution system operator models. The idea that you can introduce competition above an infrastructure layer rather than duplicating the layer is not a discovery. It is orthodoxy.

What is new is that the separation has never before been possible at the timescale and granularity where the constraint actually binds. Unbundling gave competitors a legal right of access to a network whose real-time condition remained the incumbent’s private knowledge. The right was formal; the state was not shared.

So imagine a distribution network remains a regulated monopoly. The DNSP still owns and operates the infrastructure. But suppose the system continuously exposes available capacity; congestion; voltage headroom; reliability; serviceability; scarcity; residual capacity; and the economic value of additional capacity.

Now the infrastructure owner still owns the asset. But the allocation of residual capacity becomes transparent and contestable.

That changes the monopoly problem. You don’t necessarily need to duplicate the infrastructure. You need to unbundle the state, not just the wires.

Infrastructure Ownership ≠ Capacity Allocation

It also creates a practical test for governance. If a network operator claims that capacity is unavailable, the claim should increasingly be supported by observable state data. If a user is denied access, the reason should be legible. If scarcity rents are collected, the basis for them should be auditable. If investment is deferred, the opportunity cost should be visible.

Transparency does not remove the need for regulation. It makes regulation less dependent on trust, lobbying and static engineering declarations.

11Efficiency and investment are not opposites

There is an obvious objection. If we make infrastructure more efficient, don’t we reduce the incentive to invest in more infrastructure?

Sometimes. But there is an opposite effect, and it is the larger one. Better markets make better investment signals.

Imagine a feeder where scarcity is currently hidden inside connection queues, curtailment, manual operating limits and engineering assumptions. It is difficult to answer: “where should we spend the next dollar of capital?”

Now imagine the system continuously measures hours of scarcity; magnitude of constrained capacity; willingness to pay; curtailment; reliability degradation; congestion rents; flexible demand; and the value of additional capacity.

Now the investment question becomes empirical.

Instead of “we think this network might need augmentation,” you can say: “this location experiences persistent scarcity, users repeatedly demonstrate willingness to pay for additional capacity, and the economic value of relieving the constraint exceeds the cost of augmentation.”

The market has become an investment discovery mechanism.

This is one of the clearest value gains from combining the two papers. The first improves the measurement of the physical frontier. The second improves the measurement of demand for access to that frontier. Together, they sharpen the signal sent to investors.

This is also where the proof-of-work idea becomes relevant again. A tiny computational cost on email is not merely a spam filter; it is a way of revealing the social cost of volume that otherwise appears free. In infrastructure, state-dependent prices and access charges perform a related function. They reveal where demand is genuinely valuable, where usage is imposing congestion or reliability costs, and where additional capacity would create more value than the cost of building it.

The mechanism is not identical. Email proof-of-work and DER allocation operate in different physical and institutional environments. But both challenge the assumption that a near-zero marginal transmission cost means access should be economically unstructured. A small, well-designed friction improves the signal without undermining legitimate use.

This creates a feedback loop:

Scarcity → Price/Rent → Investment Signal → New Capacity → Reduced Scarcity

That is a much healthier relationship between markets and infrastructure investment.

12Why this matters for development finance

Development infrastructure often fails to attract finance for a reason that is not simply a lack of social value. The problem is that value is difficult to observe, allocate and capture.

A rural feeder may support productive activity, health services and household welfare, but its future demand is uncertain. A distributed solar project may have abundant generation but face uncertain grid access. A telecommunications network may have long-term development value but weak early cash flows. A water system may create large public benefits while generating limited direct revenue.

In each case, financiers face a familiar combination of problems: uncertain demand; uncertain operating conditions; weak visibility into future cash flows; fragmented users; political and regulatory risk; currency risk; and uncertainty about who can pay for the capacity being created.

This is where the convergence between dynamic operating envelopes and stateful allocation becomes important. Dynamic operating envelopes make the physical service frontier visible. Stateful markets make the economic demand for access visible.

Together, they turn previously opaque infrastructure value into a stream of observable operating data: how much capacity is available; when it is available; who uses it; who is constrained; what users are willing to pay; how much revenue is reliable; and where additional investment would create measurable value.

That does not eliminate development risk. It removes a particularly damaging form of it: not knowing whether the infrastructure can generate and allocate enough value to support repayment at all.

The financing loop begins to look like this:

Observable Physical State → Reliable Allocation → Predictable Cash Flows → Better Risk Assessment → Lower Cost of Capital → More Infrastructure

It is worth being concrete about the mechanism, because the chain from measurement to cost of capital is short, and it does not run through average revenue. A lender does not size debt against the expected case. It sizes against the downside, and specifically against the minimum debt service coverage ratio it is willing to underwrite.

Today a distributed project’s downside is bounded by nothing better than an assumption about how often local constraints will bind, because nobody keeps the record that would bound it. That unmeasured tail gets priced twice: once as lower gearing, once as a higher required equity return. So compressing the tail does more work than lifting the mean. A modest improvement in worst-case deliverable output, if it is evidenced rather than forecast, moves the binding downside DSCR, which moves debt capacity, which moves the blended cost of capital, which moves marginal projects across the investment-decision threshold. The claim is not that measurement makes projects more profitable. It is that it converts an unmeasured tail into a measured one — and the discount capital applies to an unmeasured tail is larger than the discount it applies to a bad one. Whether that is worth tens of basis points or hundreds is an empirical question no deployment has yet answered.

This is also where the amendment problem returns as a credit question rather than a design one. A lender underwriting a project whose downside depends on a fairness weight, a price bound or a serviceability floor is underwriting a parameter — and the party best placed to change that parameter is not the borrower. Unless the amendment procedure is specified, with notice periods, an evidentiary standard and some grandfathering of positions taken under the old rule, the lender is not holding a cashflow with a bounded tail. It is holding a cashflow plus a short regulatory option, written by someone else, struck at an unknown level. That is a worse instrument than the one this architecture set out to create, and the remedy is drafting rather than engineering.

Five financing problems change shape. None of the five has been demonstrated at scale — the architecture is younger than any project it would finance — but each follows from the measurement rather than from optimism about it.

Bankability. A project evidenced by measured capacity, transparent allocation rules and auditable usage data is a different object to underwrite than one evidenced by a forecast. The lender stops pricing the sponsor’s assumptions.

Blended finance. The floor and the residual are separable, so the capital can be too. Public or concessional money funds the serviceability floor; private money finances the capacity above it. That split is not available while both are bundled into one revenue line.

Revenue certainty. Stateful allocation settles against events that happened rather than periods that were forecast. Capacity charges, congestion payments and availability fees become separately observable instead of netted into a single tariff.

Targeted guarantees. A development institution guaranteeing a whole project against demand risk is insuring an unknown. Guaranteeing a defined service floor, and letting market revenue emerge above it, is insuring a number someone can check.

Capital sequencing. Build to measured scarcity rather than to a forecast of it. The expansion trigger stops being a committee’s view of demand in 2032 and becomes how many hours the constraint actually bound last year.

This creates a more adaptive development model:

Minimum Viable Infrastructure → Measured Utilisation → Market Revenue → Targeted Expansion

The key point is not that markets can replace public finance. They cannot. Many development benefits are externalities and will never be fully captured by private users.

The point is that better physical observability and better allocation separate three things that are often bundled together: the public value of basic service; the private value of additional access; and the investment value of relieving persistent constraints. Once separated, each can be financed more intelligently.

A useful analogy is a bridge built in stages. Public money usually has to build the first crossing because the social benefits are too broad and uncertain for private investors to capture. Once traffic patterns become visible, private capital finances additional lanes, logistics facilities or connected services. The first investment creates access; measurement reveals where the next investment has value.

13From rent extraction to rent discovery

This is where the Georgist intuition becomes particularly useful.

The objective isn’t necessarily to eliminate economic rents. Scarcity rents can contain valuable information. A high scarcity value tells us: something is constrained and someone values relieving that constraint.

The problem is when that information is hidden.

A state-based market makes scarcity measurable. That allows us to distinguish between returns to productive investment and rents arising from control over scarce access.

Once measured, those rents can become signals for investment; network augmentation; flexible demand; storage; alternative infrastructure; new market entrants; and policy.

The rent stops being merely something extracted from a captive user. It becomes information about where the system needs to change.

For development finance, this distinction is crucial. A hidden rent is difficult to finance against and easy to appropriate. A measured rent can be audited, shared, taxed, securitised, or reinvested.

That creates the possibility of a more explicit social contract: public institutions protect the serviceability floor; infrastructure owners receive a fair return for providing and maintaining assets; users pay for residual access according to transparent rules; and scarcity rents help reveal where new investment is needed — or, per the earlier answer, are recycled to the participants who actually gave way to create them.

The objective is not to privatise every benefit of infrastructure. It is to make the different sources of value visible enough that public and private capital can be directed toward the roles they are best suited to perform.

14This is where physics and economics actually meet

Again, this isn’t the claim that nobody has ever connected physics and economics. Electricity economics has been doing this for decades. Locational marginal pricing is itself an elegant example: network physics affects economic prices. There is also extensive research in transactive energy, congestion pricing, network economics, mechanism design and cyber-physical systems.

The different proposition is more specific:

Can the evolving physical state of a cyber-physical system become a first-class state variable of a market that repeatedly allocates access to that system?

That is a much stronger and more testable claim. The market isn’t simply informed by physics. It becomes dynamically coupled to the physical state.

Physical state → market → allocation → next physical state

Physics changes the market. The market changes behaviour. Behaviour changes physics. And the loop repeats.

That is not merely an economic model of a physical system. It is a coupled economic-physical dynamical system.

For development, the significance is that this loop can also connect operations to finance. Measured physical performance affects allocation. Allocation affects revenue. Revenue affects creditworthiness. Creditworthiness affects investment. Investment changes the physical system.

Physical state → Allocation → Realised revenue → Financing capacity → Investment → New physical state

The infrastructure market becomes not only a coordination mechanism, but also a potential information layer for development finance.

More broadly, it becomes an information layer for governance. Digitisation can expose where a public agency is underperforming, where a monopoly is withholding capacity, where a market is mispricing risk, where a subsidy is failing to reach its intended recipient, or where a regulatory rule is producing avoidable scarcity.

Transparency does not guarantee good decisions. But without transparency, poor decisions are difficult to distinguish from unavoidable constraints.

15Why electricity may only be the beginning

Once expressed this way, the architecture stops looking uniquely electrical.

In telecommunications the state is bandwidth, latency, congestion, reliability. In transport it is capacity, congestion, travel time, reliability. In data centres it is compute, power, cooling, network. In water it is flow, pressure, storage, treatment capacity.

The specific physics changes. The market architecture doesn’t necessarily have to.

In each case there is a physical system; a continuously changing state; multiple users; different service requirements; constrained capacity; a need for allocation; and economic value attached to access.

That is the definition of a state-based infrastructure market. It is also the definition of a holarchy — which is why the same architecture travels: these are not four analogies for one idea, they are four instances of one structure, each failing at whichever nested level has gone dark. And it is a potential basis for more adaptive financing.

A telecommunications project can demonstrate actual bandwidth utilisation and service quality. A water project can demonstrate delivered volumes, pressure and reliability. A transport project can demonstrate throughput, congestion relief and availability. A data centre can demonstrate compute utilisation, power availability and service performance. A postal network can demonstrate delivery volumes, route utilisation, service quality and the cost of maintaining universal access.

The more these outputs are measured and linked to transparent allocation and payment rules, the less financing depends on a single static forecast.

The telecommunications lesson is the sharpest version of this. Unbundling gave entrants a legal right of access, but the value of that right depended on information the incumbent held: line availability, fault rates, congestion, and the ability to verify that access was not being quietly degraded.

The same principle applies to electricity and other infrastructure. Digitisation without transparency can simply automate opacity. The objective is not merely to collect more data. It is to make the relevant state, constraints, decisions and payments legible to users, regulators, investors and competing service providers.

The comparison with Satoshi should remain similarly disciplined. Bitcoin’s contribution was not simply to attach a price to digital activity. It combined proof-of-work, a public transaction history and a consensus rule to make a shared ledger credible without a central clearing institution. The broader lesson for infrastructure is not “put everything on a blockchain.” It is that allocation and settlement become more trustworthy when the rules, state transitions and costs of manipulation are explicit and independently verifiable.

That lesson may be relevant where infrastructure markets need auditable records of capacity, priority, service delivery and payment. But physical infrastructure still requires trusted measurement, legal authority, operational control and public governance. A cryptographic ledger cannot by itself determine whether a feeder is safe, whether a hospital received adequate supply, or whether a network operator reported its constraints honestly. The useful insight is institutional: credible markets require not only prices, but also verifiable state, transparent transition rules and clear responsibility for the physical system.

The implication is not that every network should impose a universal usage fee. It is that network design must distinguish between capacity, access and externality. A system can be abundant in bandwidth and still scarce in attention, trust, processing capacity or administrative legitimacy. The relevant economic object is whatever the network is actually struggling to allocate.

16The emerging architecture

The convergence between Dynamic Operating Envelopes and stateful allocation suggests a general sequence:

Observe → Infer State → Determine Serviceability → Expose Residual Capacity → Allocate → Price → Respond → Observe Again

This is fundamentally different from the traditional infrastructure cycle:

Forecast → Build → Set Limit → Connect → Constrain

The first is closed-loop. The second is largely asset-centric.

As infrastructure becomes more distributed, variable and digitally observable, the first architecture becomes increasingly attractive.

For development finance, the difference is equally important. The traditional model asks investors to commit large amounts of capital before demand, operating conditions and revenue are fully known. The closed-loop model allows some uncertainty to be resolved through operation. It supports a staged approach:

Build → Measure → Allocate → Learn → Expand

There is a shape for this that people already accept, and naming it makes the ask small.

The internet protocol is the thin middle of an hourglass. Many link technologies below — fibre, copper, radio, satellite. Many applications above. One deliberately minimal protocol between them. The waist is narrow on purpose: almost nothing is standardised there, which is precisely why invention is possible everywhere else.

The regulatory ask is that shape. Standardise the state record and the event format — what the constraint was, where, when, who was curtailed, what was offered and to whom. Nothing above the waist needs specifying: envelopes, auctions, tariffs, retail products, aggregator business models can all compete. Nothing below it needs to be uniform: topology, inverter vendor and metering hardware can differ. Standardise the substrate; keep the mechanism open.

One caution the internet’s version carries. The waist worked partly because nobody in the middle was liable for delivery. Electricity has no such luxury, so the standard has to carry obligation as well as format.

That does not make infrastructure risk-free. It makes risk more observable, more divisible and more manageable.

It also makes accountability more concrete. A regulator can ask whether a constraint was physical, contractual or informational. An investor can ask whether revenue came from genuine service delivery or from opaque scarcity. A user can ask why access was denied. A public institution can ask whether a subsidy is buying additional service or merely compensating for an inefficient allocation system.

One last lesson from the same source, and it is commercial.

TCP/IP created an amount of value that is hard to state without sounding foolish, and captured none of it. The protocol earned nothing. The rent went to the layers above — applications, platforms, marketplaces — and to the layer below, in access and interconnection. That is where my VPN fee went. The waist stayed thin, open and free, which is why it won, and equally why it was worth nothing to whoever maintained it.

If the state layer of infrastructure becomes the waist, the same thing happens unless someone decides otherwise in advance. Whoever builds the record should choose the capture point deliberately — an exchange that clears against it, a member-owned co-operative that governs it, a financial instrument defined on it — rather than discovering afterwards that they built the most useful and least valuable component in the system.

And a darker footnote. There is exactly one place where the internet left its record unauthenticated: BGP, the protocol that decides how traffic reaches a network. Route hijacking has been a live problem for thirty years and the repair is still incomplete. A state record that anyone can assert into is not a state record. It is an attack surface.

17What could go wrong

An argument this tidy deserves its objections stated at full strength, not in a footnote. There are four that I think are serious.

The state can be gamed. Any allocation rule that depends on observable state creates an incentive to manipulate what is observed. A participant who knows that scarcity raises their payment can create the appearance of scarcity; one who knows that historical service affects future priority can manufacture a record of having gone without. This is not hypothetical — it is the ordinary consequence of making a measurement economically consequential. The mitigations are real but partial: bound the prices so manipulation has limited upside, derive state from physical measurements that are expensive to fake, keep the ledger auditable so patterns of gaming become visible after the fact. None of them eliminate the problem. A market coupled to physical state inherits every weakness of the measurement layer beneath it.

It can be regressive. State-contingent pricing rewards flexibility, and flexibility is a purchased good. The household with a battery, an EV and a home energy management system can shift; the household on a fixed income in a poorly insulated rental cannot. If the residual market is allowed to allocate everything, the least flexible participants pay the most and receive the least — and the fairness-state result described earlier suggests this compounds rather than averaging out. This is precisely what the serviceability floor is for, and it is why the floor has to be set politically and generously rather than as a technical afterthought. An architecture that makes scarcity legible also makes it possible to see exactly who bears it, which is either the strongest argument for this design or the most damning, depending entirely on what is then done about it.

Visibility creates exposure. Low-voltage state data is, unavoidably, behavioural data about households: when they are home, when they cook, when they charge, when they travel. And a market dynamically coupled to a physical control system is a new class of attack surface — one where manipulating a price signal can move real power flows. The engineering answer is aggregation, minimisation and strict separation between the settlement layer and the control layer. The honest answer is that this is a genuine cost of the architecture, not an implementation detail.

And the incumbent has no reason to volunteer. The entire argument assumes the network operator will expose its state. Why would it? Published scarcity data reveals under-investment where constraints persist, and over-recovery where they never appear. It converts an engineering judgment that currently cannot be second-guessed into an auditable claim. A regulated monopoly earning a return on assets has no commercial incentive to make the case for using its existing assets harder. This is the real obstacle — not technical feasibility, but the absence of any party whose interests are served by transparency except the users, who have no mechanism to demand it. Which means the state layer will not emerge from the market. It has to be required: standardised data access, mandated recording of curtailment and dispatch events, and the right of affected parties to read the record. That is a regulatory ask, and it is a modest one — it is bookkeeping, not market design, and every more ambitious architecture depends on it.

And over-provisioning may simply win. This objection has the best track record of any here, because the internet already ran the experiment. Two serious attempts were made to build quality-of-service guarantees into the network — IntServ in the mid-1990s, DiffServ at the end of them. Neither was deployed end-to-end at scale, and the reason was not technical. Guarantees that cross administrative boundaries require settlement between parties with no reason to trust each other, and nobody could make that work. What won instead was over-provisioning plus local caching: build more capacity than you need, and move the content nearer the user. Content delivery networks beat quality of service.

The energy translation is unflattering and worth writing down. Batteries are the content delivery network of the grid, and “build more and buffer locally” may beat any elegant coordination protocol on exactly the grounds that beat QoS.

Three things make the analogy imperfect, and they are the whole of the answer. A local battery cannot relieve a thermal constraint two levels above it; it can only reshape one household’s profile. Over-provisioning in electricity carries a cost floor that bandwidth never had — copper, easements, planning consent, and a decade of lead time. And CDNs did not eliminate coordination; they relocated it into DNS, anycast and peering agreements, which is a state layer that somebody built, owns and profits from.

18The bigger question

The most interesting implication may therefore have little to do with electricity markets themselves. It may be this:

What happens when physical infrastructure becomes sufficiently observable that its changing capacity can be treated as a continuously traded economic resource?

If that happens, several boundaries begin to move. The boundary between engineering and economics. The boundary between infrastructure ownership and access. The boundary between scarcity and abundance. The boundary between operating decisions and investment decisions. The boundary between public service obligations and private demand.

And perhaps most importantly, the boundary between physical resources and economic resources.

The infrastructure doesn’t necessarily change. What changes is our ability to coordinate it, finance it and learn from its operation.

That may be the real path toward abundance. Not simply building more of everything. But developing markets capable of finding the capacity that already exists, protecting the services society requires, allocating the remainder efficiently, pricing genuine scarcity, generating credible revenue signals, and directing investment toward the constraints that actually matter.

The deeper lesson from telecommunications, postal networks, email, Bitcoin and energy systems is that infrastructure failure is often not a single engineering failure. It is a failure of visibility, governance, credible settlement and economic adaptation.

A network can have spare capacity and still deny access. A postal system can remain universal and still become unaffordable. An email system can have abundant bandwidth and still be degraded by unpriced automated demand. A grid can have abundant generation and still curtail it. A market can have willing buyers and still fail to finance expansion.

In each case, the missing ingredient is not necessarily more physical capacity. It is a trustworthy way to observe the system, distinguish public obligations from residual commercial access, allocate capacity, verify outcomes, and turn operating results into investment signals.

The future of abundance may therefore depend less on eliminating scarcity than on making scarcity — and abundance — legible.

And once physical state becomes legible, it can become measured, governed, allocated, priced, verified, financed, and invested around.

That is where the economics of abundance begins to meet the physics of the real world — and where better coordination may begin to solve not only operational problems, but also the governance and financing problems that keep essential infrastructure from being built.

The timeline is therefore:

Physical Capacity → Network Visibility → Dynamic Operating Envelope → Stateful Allocation → Transparent Scarcity Rent → Credible Revenue → Development Finance → Targeted Investment → Greater Serviceable Abundance

The penny is not the point because it is valuable on its own. It is the point because it makes value, scarcity and participation visible.

The same may be true of infrastructure. The future may not belong to whoever owns the most capacity, but to whoever can make changing capacity legible enough to govern, allocate, finance and expand.

19References

Abadi, M., Birrell, A., Burrows, M., Dabek, F. and Wobber, T. (2003). “Bankable Postage for Network Services.” ASIAN 2003, LNCS 2896.

ARENA (2022a). DEIP Dynamic Operating Envelopes Working Group: Outcomes Report. Australian Renewable Energy Agency, March 2022.

ARENA (2022b). Review of Dynamic Operating Envelope Adoption by DNSPs. Australian Renewable Energy Agency, July 2022.

Arnott, R. and Stiglitz, J. (1979). “Aggregate Land Rents, Expenditure on Public Goods, and Optimal City Size.” Quarterly Journal of Economics 93(4), 471–500.

AS/NZS 4777.2:2020. Grid connection of energy systems via inverters, Part 2: Inverter requirements (Amendment 2:2024).

Back, A. (1997). “Hashcash.” Announcement to the cypherpunks mailing list, March 1997.

Back, A. (2002). “Hashcash — A Denial of Service Counter-Measure.” Technical report, August 2002.

Chiu, D. and Jain, R. (1989). “Analysis of the Increase and Decrease Algorithms for Congestion Avoidance in Computer Networks.” Computer Networks and ISDN Systems 17(1), 1–14.

Dwork, C. and Naor, M. (1992). “Pricing via Processing or Combatting Junk Mail.” CRYPTO ’92, LNCS 740, 139–147.

Dwork, C., Goldberg, A. and Naor, M. (2003). “On Memory-Bound Functions for Fighting Spam.” CRYPTO 2003.

George, H. (1879). Progress and Poverty.

Hill, R. (1837). Post Office Reform: Its Importance and Practicability.

Hogan, W. (1992). “Contract Networks for Electric Power Transmission.” Journal of Regulatory Economics 4(3), 211–242.

Jacobson, V. (1988). “Congestion Avoidance and Control.” ACM SIGCOMM Computer Communication Review 18(4), 314–329.

Jain, R., Chiu, D. and Hawe, W. (1984). “A Quantitative Measure of Fairness and Discrimination for Resource Allocation in Shared Computer Systems.” DEC Research Report TR-301, September 1984.

Kilby, P., Bissett, H., Smith, B. and Banfield, B. (2025). [TITLE TO CONFIRM]. CIGRE Study Committee C6.

Klein, E. and Thompson, D. (2025). Abundance. Avid Reader Press.

Koestler, A. (1967). The Ghost in the Machine.

Nakamoto, S. (2008). “Bitcoin: A Peer-to-Peer Electronic Cash System.”

Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press.

Saltzer, J., Reed, D. and Clark, D. (1984). “End-to-End Arguments in System Design.” ACM Transactions on Computer Systems 2(4), 277–288.

Schweppe, F., Caramanis, M., Tabors, R. and Bohn, R. (1988). Spot Pricing of Electricity. Kluwer Academic Publishers.

Simon, H. (1962). “The Architecture of Complexity.” Proceedings of the American Philosophical Society 106(6), 467–482.

Sweeney, S. ([YEAR TO CONFIRM]). [THESIS TITLE TO CONFIRM]. PhD thesis, [INSTITUTION TO CONFIRM].

Sweeney, S., Kilby, P., Moghaddasi, K., Mudiyanselage, S. et al. ([YEAR TO CONFIRM]). [TITLE TO CONFIRM]. Manuscript under review.

United States Mint (2025). Final circulating one-cent coin struck, Philadelphia, 12 November 2025.

Vickrey, W. (1969). “Congestion Theory and Transport Investment.” American Economic Review 59(2), 251–260.

Wilson, R. (2002). “Architecture of Power Markets.” Econometrica 70(4), 1299–1340.

---