A Toll Set by Another’s Parliament

Its companion piece described the Carbon Border Adjustment Mechanism from the inside, as the elegant instrument by which Europe exports its carbon price — charge imports the same price domestic producers pay, deduct any carbon price paid abroad, and watch trading partners choose to price their own carbon rather than hand the revenue to Brussels. From the European side, it is a self-funding, self-exporting piece of climate policy. From the other side of the border, it is something else entirely.

For a producer in Delhi or Johannesburg or Maputo, CBAM is not a domestic policy their citizens debated and chose. It is a carbon price legislated in a foreign parliament, calibrated to European industry, and applied to their exports whether they consent or not. The mechanism is identical; the meaning is inverted. What is a price signal at home is, abroad, a trade barrier — and the countries it falls on hardest are precisely those with the least say in setting it and the least capacity to escape it. That is the border adjustment problem, and it is one of the cleanest inversions in the series.

The Incidence Falls on Those Who Had No Vote

The numbers make the incidence concrete. The World Bank estimates CBAM could affect around $16 billion a year of developing-country exports; UNCTAD puts the potential losses for India, Brazil, South Africa and Indonesia alone at up to $5.6 billion a year. The exposure is heaviest in iron and steel — by far the largest covered sector — then aluminium, cement and fertiliser, the carbon-intensive commodities that developing exporters disproportionately sell.

Note who bears it. A 0.5% hit to African GDP is roughly four times the EU’s gain from a major trade agreement — a transfer of that scale, moving in the wrong direction on development grounds, from a policy those countries did not write. And the burden is concentrated exactly where export dependence is highest: least-developed economies such as Mozambique and Zimbabwe, with a large share of their trade in covered goods and the least technical and financial capacity to change how those goods are made. The people most exposed to the price are the ones furthest from the room where it was set.

It Is Regressive in Development Terms

The deeper problem is not merely that CBAM costs developing exporters money; it is that it charges them most precisely for the conditions their development stage imposes. CBAM prices the embedded carbon of a good, and embedded carbon is largely a function of grid intensity — how much coal sits behind the electricity that made the steel. A developing economy running a young coal fleet has high grid intensity not by choice but by circumstance, so its exports carry high embedded carbon and attract the highest border charge. The mechanism penalises the carbon intensity that a coal-dependent development path produces — and then charges the countries least able to finance the alternative.

The “Price Your Own Carbon” Offer Is a Trap

CBAM’s designers present an escape: price your own carbon, and the domestic price is deducted from the border charge, so the revenue stays home. From the European side this is the mechanism’s cleverness — it induces others to adopt carbon pricing. From the developing side it can be a trap, because it forces a domestic policy the country may not want, at a fiscal and political cost it did not choose, to satisfy a foreign rule.

The tension is visible in India’s stance: it has built the architecture of an emissions trading scheme, yet has rejected European proposals to raise its domestic carbon price specifically to blunt CBAM — because doing so on Brussels’ timetable is a surrender of policy sovereignty, and because the revenue gained may not offset the competitiveness lost. For a capital-scarce economy, imposing a domestic carbon price is not a free lever; it raises industrial costs across the board, and the promise that CBAM revenue will fund developing-country decarbonisation remains largely that — a promise, with most of the money flowing to the EU budget.

The Responses Reshape Trade

Faced with a toll they cannot vote on, exposed countries are choosing among a menu of responses — and each one reshapes trade flows and carbon policy in a different way. None is clearly dominant, which is itself the point: CBAM does not produce a tidy convergence on carbon pricing so much as a scramble of partial, self-interested adaptations.

And the mechanism is propagating. The UK, Australia, Turkey and Canada are developing or considering their own border levies, which means a carbon-intensive exporter increasingly faces not one such wall but several. That widens the reach of carbon pricing — the EU’s intended outcome — but it does so by exporting a policy rather than negotiating one, and it steadily closes the trade-diversion escape route that many developing exporters are currently relying on.

Positioning: Read the Incidence, Not the Intent

The OECD piece said to read CBAM as a durable, self-funding price and position for its spread. The inversion here is to read its incidence — who actually pays, and how they respond — because that is what moves the trade flows and the asset values.

Reading It Through the Frameworks

Where the conclusion inverts. The carbon-pricing framework is the same on both sides — a price on embedded carbon, deductible against a domestic price — but the vantage flips it. From the EU, CBAM is a policy instrument that funds and exports itself; from the Global South, it is an externally-imposed cost with regressive development incidence, charging the highest toll to those least able to pay or to change. The investable content is not the mechanism, which the mirror piece covered, but the incidence and the response: who clears the wall, who pays, and who reroutes.

Structural moat or temporary bottleneck? CBAM is a structural, propagating feature of trade, not a temporary friction — and its spread steadily removes the escape routes. That makes low carbon intensity a durable competitive moat at the border and high intensity a durable disadvantage. The discipline is to separate the developing exporter that can decarbonise into an advantage (access to clean power, capital, and the covered value chain) from the one locked into a coal grid it cannot cheaply change, and to treat every new border levy as narrowing the room the latter has left to divert.

In the Present Tense

The companion to this piece treats heat as a design problem — a tail risk creeping into the envelope, a derating to model and price as the OECD grid warms into conditions it was not built for. That framing is correct for a rich, temperate world discovering heat. It is the wrong tense for most of the planet. Across the Global South, heat is not an emerging risk. It is the present-tense operating condition, and it always was.

This is the cleanest inversion in the whole series, and the International Labour Organization draws the line precisely: on its projections, South Asia and Western Africa will lose around 5% of total working hours to heat stress by 2030, while North America and Europe are “not significantly affected.” The same physical phenomenon that the OECD is beginning to model as a future exposure is, across the tropics and subtropics, the everyday backdrop against which every asset already operates and every worker already labours. And because the starting point is different, the binding constraint is different too — and so, therefore, is the investment case.

The Binding Constraint Is Labour, Not Transformers

In a rich, automated, largely indoor economy, heat’s first-order cost is what it does to machines. In a labour-intensive economy with a large outdoor and informal workforce, that ranking flips: the first-order cost is what heat does to people at work, and it dwarfs the asset-derating story. When it is simply too hot to work at normal intensity, output falls — in the fields, on the building site, in the un-airconditioned workshop — and those are the sectors that carry these economies.

The numbers are already enormous. The ILO puts the global loss at the equivalent of 80 million full-time jobs and roughly $2.4 trillion of GDP by 2030 — and it is heavily concentrated in the Global South. In India alone, an estimated 247 billion labour hours were lost to extreme heat in 2024, roughly two-thirds of it in agriculture and a fifth in construction. Bangladesh lost some 250 million workdays to heat in 2024, about 0.4% of GDP. This is not a future modelling exercise; it is a competitiveness issue being paid in lost output today.

Cooling Is a Demand Explosion, Not a Marginal Peak

The second inversion is on the demand side. In the OECD, air-conditioning load is a marginal peak — the thing that pushes an already-built, already-cooled grid a little higher on the hot afternoon. In the Global South it is something else entirely: a demand-growth story off a near-zero base. About 3.5 billion people live in hot climates, and only around 15% of them own an air conditioner. That gap is not a peak to shave. It is a multi-decade wave of new electricity demand waiting to be built for.

This reframes cooling from a reliability nuisance into one of the largest structural demand drivers in the emerging world — more than 80% of the growth in global cooling electricity demand to 2050 is expected to come from emerging and developing economies, and in India space cooling’s share of peak load is projected to climb from about 10% today to 45% by 2050. For an infrastructure investor, that is not a risk to hedge; it is a generation, grid and appliance build-out to finance. The heat that is a supply-side problem in the OECD is, here, primarily a demand-side one — and demand growth, not derating, is what builds infrastructure.

The Envelope Was Never There

The OECD piece turns on a design envelope being breached — ratings set against a historical climate that has since moved. In much of the Global South, that framing does not apply, because the envelope was never there to breach. Infrastructure across the tropics was built, if it was engineered to a formal climate standard at all, for conditions that the OECD would classify as extreme — and it has always operated in them. There is no comfortable historical baseline being left behind; the baseline was already hot.

Adaptation Is Present-Tense Capex

Because heat is the baseline rather than a forecast, adaptation here is not a future line item to be provisioned against — it is capital that needs deploying now. And it is unusually high-return, because it acts on both binding constraints at once: efficient cooling and heat-resilient design protect labour productivity and restrain the demand explosion they would otherwise fuel. The health dimension makes the case starker still — the World Health Organization associates roughly half a million deaths a year with heat, disproportionately in exactly these regions.

The policy layer is already moving in the present tense: national cooling action plans, appliance-efficiency standards, and cool-roof and passive-design mandates are being adopted across the region precisely because the problem is current, not projected. For the investor, the through-line is that the adaptation build-out is not a hedge against a future scenario; it is the response to today’s operating conditions — which is what makes it fundable now.

Positioning: Own the Demand, Not the Derating

The OECD piece says to price the derated megawatt. The inversion here is direct: own the demand and the adaptation, not the derating. The opportunity is not in hedging a supply-side capacity loss at the margin; it is in financing the largest cooling-and-resilience build-out in the emerging world, and in doing so on the efficient side, so that the demand wave does not simply overwhelm the grid it lands on.

Reading It Through the Frameworks

Where the conclusion inverts. The OECD framework reads heat as a mispriced physical risk to a largely-built system — a derating to underwrite. The Global South framework reads the identical physics through a different economy and reaches the opposite investment conclusion: heat is a present cost to labour and a structural driver of new demand, so the money is in the build-out and the adaptation, not in the hedge. This is the whole point of reading the mirror pair together — the same framework, applied honestly to a different base, produces a different answer.

Structural moat or temporary bottleneck? The cooling-demand wave is as structural as demography, and the adaptation deficit is the bottleneck — but a fundable one, because it pays back in protected productivity and restrained peak load today. The discipline is to back the efficient version of the build, since an inefficient cooling boom simply exports the problem to the grid, and to capture the leapfrog by financing heat-resilient design where the stock is still on the drawing board rather than the retrofit where it is not.

Where the Co-Benefit Becomes the Whole Case

Its companion piece made the case that clean air is the most under-counted line in the decarbonisation ledger — a health dividend, local and immediate, that often exceeds the climate benefit and that markets rarely price. In the OECD, that dividend is a powerful supplement: it strengthens a climate case that was already being made. Move to the Global South and the relationship inverts. The health burden is so enormous that clean air stops being a co-benefit and becomes the entire case.

The distinction matters because it changes who has to be convinced and with what argument. In a country where air pollution is among the leading causes of death and a measurable drag on GDP, decarbonising the power sector, the kitchen and the tailpipe does not need a climate justification to clear an investment hurdle. It clears on domestic public health alone. The avoided deaths, the recovered productivity, the lower health-system cost — these pay for the intervention before a single tonne of avoided CO₂ is counted. The climate benefit becomes the free extra on top of a case that already closed.

The Burden Is the World’s Largest

Consider India, the clearest case. Anthropogenic PM2.5 was associated with roughly 1.72 million deaths in 2022 — up 38% since 2010 — and the monetised value of that premature mortality has been put at around $339 billion, on the order of 9.5% of GDP. Twenty-one of the world’s thirty most polluted cities sit in India; national average PM2.5 runs near ten times the WHO guideline. In sheer aggregate terms, India carries the largest air-pollution health burden of any country on earth, by a wide margin.

That last point is the crux of the investment argument. Whether one uses the high VSL-based figure (~9.5% of GDP) or a conservative output-loss estimate (~1.4%), the annual health cost of dirty air runs to tens or hundreds of billions of dollars — a number against which the cost of cleaner power, cleaner cooking and cleaner transport is small and, crucially, one-time capital against a recurring loss. The math is not close. This is why the co-benefit becomes the case: the case was closing on health before anyone mentioned the climate.

It Reframes Decarbonisation as Domestic Policy

Once the case closes on health, the politics of decarbonisation change entirely — and this is the most investable consequence. The standard developing-country objection to climate action is a fairness argument: why should we bear the cost of a warming the rich world caused? It is a strong objection, and it stalls capital. The health frame dissolves it, because cleaning the air is not a sacrifice made for the world’s sake. It is a domestic public-health investment a government would want to make even if climate change did not exist.

This is why the health frame mobilises capital the climate frame cannot. A coal-plant retirement or a clean-cooking programme justified as public health draws on domestic health budgets, development-bank health envelopes and a domestic political constituency — sources that a “climate” label, with its overtones of foreign obligation, often cannot reach. The same physical project has a wider, deeper funding base when it is costed as a health intervention that happens to cut carbon.

Two Fronts: Ambient and Household

The burden has two distinct sources, and the health case closes on each. The ambient front is the outdoor air — coal power, industry, vehicles, and seasonal crop-residue burning across the Indo-Gangetic plain. The household front is quieter and, per person exposed, often worse: more than 100 million Indian households still cook and heat with solid fuels — wood, dung, crop residue, coal — producing indoor PM2.5 that routinely exceeds 500 µg/m³, a hundred times the WHO guideline.

The household front is where the health case is most overwhelming and the climate case weakest — solid-fuel cooking is a small share of emissions but an enormous share of exposure — which makes it the purest example of the whole thesis. Clean cooking is barely a climate project. As a public-health project it is one of the highest-return interventions available anywhere in the world.

Positioning: Underwrite the Health Return

The OECD play was to count the health dividend as an under-priced supplement to the climate case. The inversion here is to lead with the health return, because it is what closes the investment, mobilises the capital and commands the enforcement.

Reading It Through the Frameworks

Where the conclusion inverts. The air-quality framework is identical on both sides — count the local, immediate health value of cleaner air alongside the diffuse, delayed climate value. But the magnitudes flip the answer. In the OECD, relatively clean air makes health a supplement to a climate case that must still carry the investment. In the Global South, a health burden measured in millions of deaths and high single-digit percentages of GDP makes health the case, and climate the free rider. Same framework; the size of the burden decides which benefit does the work.

Where does policy become the cash flow? Through the air-quality mandate, not the climate pledge. The durable, enforced, locally-owned driver in these markets is the clean-air rule, because it answers a grievance voters feel directly — which means the decarbonisation that actually gets built and funded is the subset that also clears the air. The discipline is to separate the intervention whose health payoff closes the case on its own (coal near cities, clean cooking, urban transport) from the one that still depends on a contested climate argument (a remote plant with few people downwind), and to follow the health mandate, because it has the constituency, the budget line and the enforcement the climate mandate often lacks.

Access Before Compliance

Its companion piece described a very particular kind of water spending: the regulatory capex cliff facing the developed world — replacing lead service lines, removing PFAS “forever chemicals,” meeting an ever-tightening Clean Water Act standard. It is real, large and distinct from any scarcity story. But look at what it assumes. The pipe already runs to the house. The water already flows. The question is only how clean it must now be made. That is a compliance problem on a built-out system — the refinement of a service that already exists.

For much of the Global South, that assumption collapses at the first step, and its collapse reframes the entire asset. In 2024, 2.2 billion people still lacked safely managed drinking water and 3.4 billion lacked safely managed sanitation — 354 million of them still practising open defecation, and 1.8 billion with no water on the premises at all. Here the binding question is not how clean the water must be. It is whether there is a connection in the first place. The spend is not compliance capex on an existing pipe; it is access capex to build the pipe that was never there. Same asset class as the mirror — but a different, earlier, and far more consequential phase of it.

Billions Still Without the First Connection

The scale of the access phase dwarfs the compliance phase it will one day reach. Behind the headline counts sit the details that define the problem: 1.8 billion people with no drinking water on premises, in two-thirds of those households the water fetched by women and girls; 106 million still drinking untreated surface water; 354 million with no sanitation facility at all. And the gap is deeply uneven — rural safely-managed water coverage sits around 60% against 83% in cities, so the unconnected are overwhelmingly rural and peri-urban.

This is not a system to be brought up to code; it is a service to be created for a third of humanity. And the progress rate underlines the phase difference: coverage of safely managed water rose only from 68% to 74% over the decade to 2024, a pace that leaves two billion people still unserved in 2030. The Global South is not behind the OECD in the compliance race. It is running an earlier race entirely — the race to the first connection — and it is running it too slowly.

A Different Phase of the Same Asset

Water is a single asset class, but it moves through phases, and the OECD and the Global South are standing in different ones. Recognising which phase a market is in is the whole discipline, because it dictates what the money should buy — and buying for the wrong phase wastes it.

The mis-sequencing risk is the whole point. Specifying PFAS-grade treatment for a town where half the households have no piped connection is buying the mature phase’s solution for an early-phase problem — a standard nobody can yet benefit from, layered on a service that does not reach them. The return on the first connection — in avoided disease, in the hours of a woman’s day no longer spent walking to a well — dwarfs the return on the marginal contaminant removed from water that already flows. Access is the high-return early phase; compliance is the diminishing-return mature one. Fund the phase the market is actually in.

It Is a Financing Problem, and the Return Is Health

Closing the access gap is, above all, a capital problem — and one far too large for aid to solve. The World Bank puts the cost of universal safely-managed water and sanitation at roughly $114 billion a year to 2030 — and that is construction only, before a cent of operation and maintenance. Against that, official development assistance to the entire water sector runs under $10 billion a year. The gap is not a gap philanthropy can close; as the sector itself puts it, this is an issue of capital, not charity.

The Water Version of Commercial Loss

But access capex only earns if the utility building it can recover its costs — and here the water sector meets the exact problem the power sector had one piece earlier. Its name is non-revenue water: the share of water a utility puts into its network that it never bills or is never paid for, lost to physical leakage, illegal connections and broken metering. Across developing-country utilities it routinely runs from a third to a half of all water produced — the precise water analogue of the electricity sector’s commercial AT&C loss.

The consequence is identical, and so is the sequencing lesson. A utility that loses half its water cannot fund the network extension that would connect the next household; the leak starves the build. So the access phase depends on the same unglamorous fixes the distribution-loss piece prescribed for power — metering, leak reduction, tariff reform and collection — to make the utility solvent enough to extend service at all. Non-revenue water is to the access build what commercial loss is to the grid: the revenue leak that decides whether the capacity plan can be funded. Fix the leak, and the first connection becomes financeable; ignore it, and the access gap stays exactly where it is.

Positioning: Fund the First Connection

The OECD play was the compliance upgrade — lead replacement, PFAS treatment on a built-out system. The inversion here is to fund the first connection: the access build, and the utility viability that determines whether it can be financed.

Reading It Through the Frameworks

Where the conclusion inverts. The water-capex framework is the same on both sides — spend to deliver safe water — but the phase flips what the spend buys. In the OECD, service exists, so the capex is compliance: replacing lead, removing PFAS, meeting the standard. In the Global South, service does not exist for billions, so the capex is access: the first network, the first source, the first toilet. Same asset class; the beginning of it rather than the refinement of it — and the return on the first connection dwarfs the return on the marginal contaminant.

Structural moat or temporary bottleneck? The access gap is a vast, decades-long build, not a temporary shortfall — but the bottleneck that gates it is capital and utility viability, not technology or even, ultimately, cost. That makes it a structural, blended-finance-driven market rather than a market-clearing one, and it makes non-revenue water the hinge: the leak decides whether the build is fundable. The discipline is to separate the access spend that is the right phase’s solution (coverage, sources, sanitation, the leak-and-tariff fix that funds them) from the compliance spend imported too early, and to read a utility’s non-revenue water before its extension plan — because, exactly as with the grid, the revenue leak decides whether the pipe ever reaches the next house.

The Loss Is Commercial, Not Physical

Its companion piece was about squeezing more out of the wires: undergrounding against fire and storm, hardening against weather, and grid-enhancing technologies that recover capacity lost to physics. It is a story about physical loss — the electrons that dissipate as heat, the lines that sag, the storms that snap them — and about the capex that fights it. In a rich grid, that is the right story, because the loss that matters is technical, and it is small: distribution losses across the OECD typically run around 5–8%.

Across much of the Global South, the number is two to four times higher, and — this is the whole point — most of the extra is not physical at all. It is theft, unmetered connections, un-billed consumption and un-paid bills. In India the standard measure is AT&C loss — Aggregate Technical and Commercial loss — and the very name carries the inversion: the technical part is the smaller, and the commercial part is the binding one. A utility here does not mainly lose power in its copper. It loses the money for power it actually delivered. That changes the fix, the sequence, and the entire investment case.

One in Five Units, Given Away

The scale is easiest to feel through the definition. AT&C loss measures the gap between the energy a utility puts into its network and the revenue it actually realises. A distribution company running 20% AT&C loss is, in effect, giving away one unit of electricity in every five it buys — some stolen, some delivered to a broken or bypassed meter, some billed but never paid. India has driven this down impressively, from over 27% in 2008–09 to a record ~15% in 2024–25, but even that leaves it at roughly two-and-a-half times the OECD’s technical loss, and parts of the Global South remain far higher.

A Revenue Problem, Not an Engineering One

Because the loss is commercial, it decomposes not into resistances and reactances but into two revenue failures. Billing efficiency is the share of energy that actually gets onto a bill — what theft, tampering and unmetered supply erode. Collection efficiency is the share of billed energy that actually gets paid — what non-payment, especially by government departments and protected consumers, erodes. AT&C loss is the two multiplied against the technical baseline. Neither is an engineering quantity; both are governance quantities. And that is exactly why the OECD’s toolkit is the wrong one to reach for first.

The bottom row is the operative one. Spending on undergrounding and grid-enhancing technology in a network that loses a quarter of its revenue to theft and non-payment is capex poured into a leaking cistern — every rupee of new capacity serves load that may never be metered or paid for. Fix the revenue first, and the capex that follows earns a return; skip it, and the capex simply enlarges the leak. This is why, across the Global South, metering and collection are not one item on the modernisation list. They are the precondition for the rest of it.

The Root of the Doom Loop

Commercial loss is not just a leak; it is the engine of the distribution utility’s slow death, and it wires this piece to the two around it. A discom that fails to bill and collect runs a structural deficit — India’s public discoms carry accumulated losses of roughly ₹6.47 lakh crore (~$78bn), a figure that grew by half over the decade to FY24 despite repeated bailouts. A discom bleeding revenue cannot invest in its network; the network degrades; reliability falls; and the customers who can afford to — the commercial and industrial load that paid the bills — defect to captive power, taking their revenue with them.

Meter and Collect Before You Harden

The fix, then, is not primarily steel and copper. It is measurement and enforcement: smart and prepaid meters that make theft visible and bill automatically, feeder-level metering that localises where the loss occurs, and payment rules with teeth. India’s Revamped Distribution Sector Scheme is the largest such bet in the world — a target of some 250 million prepaid smart meters, plus feeder and transformer metering, aimed squarely at pulling loss down to 12–15% and closing the cost-revenue gap. Prepaid metering in particular flips the cash flow: the consumer pays before consuming, receivables collapse, and the discom’s working capital heals.

But the honest caveat is that this is a governance fix wearing an engineering costume, and the hardware does not guarantee the outcome. As one analyst put it, if it works India gets monthly energy accounting and fast tamper detection; if it does not, the country gets a very large electronics deployment with the same old dispute layer underneath — the political tolerance of theft, the free agricultural power, the unpaid government dues. The meter measures the loss; it does not, by itself, collect. The investable distinction is between the deployment and the enforcement that has to sit behind it.

Positioning: Own the Meter, Not the Wire

The OECD play was the physical grid upgrade. The inversion here is to own the meter and the collection layer — the revenue fix that has to come before, and that determines whether, any capacity capex ever earns.

Reading It Through the Frameworks

Where the conclusion inverts. The loss-reduction framework is the same on both sides — minimise the gap between energy delivered and value recovered — but the nature of the loss flips the fix. In the OECD the gap is physical, so the answer is the wire: undergrounding, hardening, grid-enhancing tech. In the Global South the gap is commercial, so the answer is the meter and the collection rule, and the wire comes second. Same objective; a governance fix where the mirror had an engineering one.

Where does policy become the cash flow? Directly, and this is the crux. In distribution, the revenue is not set by a market clearing but by whether power delivered is metered, billed and paid — each of which is a policy-and-enforcement outcome. So the cash flow is manufactured by governance, and the mispricing is treating a governance problem as an engineering one, or an engineering deployment as if it were the governance fix. The discipline is to separate the metering rollout that is backed by real enforcement (the loss actually falls) from the one that is a large electronics purchase over an unchanged dispute layer (the loss does not), and to read commercial loss as the first domino that decides whether everything downstream — capacity, reliability, solvency — is fixable at all.

The Bypass Was Always Here

Its companion piece treated colocation — siting your own generation right at the load and skipping the grid entirely — as one of the defining innovations of the AI build-out: a clever, of-the-moment answer to a multi-year interconnection queue. It is a good story, and it is genuinely reshaping the OECD power market. But it rests on a quiet assumption of novelty that does not survive a plane ticket. Behind-the-meter generation is not a 2020s invention. Across most of the Global South it has been the ordinary way to keep the lights on for forty years.

The difference is what drives it, and that difference is the whole piece. The OECD data centre bypasses the grid because it works but is full — the connection exists in principle, but the queue to obtain it runs years. The Global South factory, hospital, mall and telecom tower bypasses the grid because it does not work — the connection exists on paper, but the power behind it fails for hours every day. One is queue-skipping; the other is reliability replacement. And because the Global South has been running that experiment at national scale for decades, it is not the laggard in the colocation story. It is the precedent — the natural experiment that already shows where mass grid defection leads.

Captive Power Already Exceeds the Grid

Take the starkest case. In Nigeria, decentralised private generation is estimated at 15–20 GW of installed capacity — against a national grid that supplies less than 5 GW to a population of over 220 million. Roughly 86% of companies own or share a generator, and self-generation meets close to half their electricity demand. More than 80% of businesses run their own power. This is not a backup layer sitting on top of the grid; for the productive economy, captive power is the grid, and the public one is the backup.

Nigeria is extreme but not unique. The six largest users of back-up generation worldwide — Nigeria, India, Iraq, Pakistan, Venezuela and Bangladesh — are all developing economies, and the pattern repeats through Egypt’s capacity shortfalls, Kenya’s diesel balancing, and the copper and cobalt miners of the DRC and Zambia unplugging from the grid to secure round-the-clock processing power. Wherever the grid is unreliable, the load that can afford to leaves it — quietly, decades ago, and at enormous scale.

Reliability, Not the Queue

Line the two bypasses up side by side and the inversion is exact. They share an architecture and almost nothing else — and the differences are precisely what make the Global South version the more instructive of the two.

The last row is the one that matters most for the OECD, and the one the colocation story tends to skip. When the load that can pay leaves the grid, it takes its cross-subsidy with it — the commercial and industrial revenue that helped keep residential tariffs affordable and the utility solvent. That thins the grid’s revenue, which worsens its reliability, which drives more of the remaining load to defect. The Global South has been living that doom loop for decades. The OECD, waving its data centres off the grid, is stepping onto the first turn of the same spiral.

A Hidden Tax, Paid in Diesel

For all those decades the coping mechanism has been brutally expensive, and that cost is the reason it was endured rather than chosen. Captive diesel runs around 30 cents per kWh in fuel alone — roughly double the price of grid power — and the all-in cost climbs from there, to a dollar or more per kWh in remote locations. In Nigeria, businesses spend up to 40% of their operating costs on fuel for power, and the country burns $30–50 billion a year keeping its own lights on.

The Flip: From Diesel to Solar-Plus-Storage

Here is the inflection that turns a decades-old symptom into a live investment thesis. The economics of captive power have quietly inverted: commercial-and-industrial solar plus battery storage is now cheaper than captive diesel — and, increasingly, cheaper and more reliable than the public grid it would replace. The coping mechanism is becoming the superior choice. A manufacturer that once ran diesel because it had no alternative now runs solar-plus-storage because it is the lowest-cost, most reliable option on the table, full stop.

The consequence runs in two directions at once. It is a large and fast-growing build-out — C&I solar, batteries, and solar-diesel hybrids across industry, telecom towers, mining and, increasingly, the Global South’s own data centres. And it deepens the grid’s doom loop, permanently: where diesel defection was at least expensive enough to keep some load tethered to the grid, clean captive power is cheap enough to make the defection final. The flip that saves the factory money is the same flip that removes its bill from the utility for good.

Positioning: Own the Defection

The OECD play was to treat colocation as a novel response to a queue. The inversion here is to treat captive power as a mature, mass-scale market at its cost inflection — and to own the defection as it flips from diesel to clean.

Reading It Through the Frameworks

Where the conclusion inverts. The behind-the-meter framework is identical on both sides — generate at the load, bypass the grid — but the driver flips the meaning. In the OECD it is a fresh, queue-driven innovation whose second-order effects are still hypothetical. In the Global South it is a forty-year-old, failure-driven norm whose second-order effects — cross-subsidy collapse, utility death spiral, a two-tier power system — have already happened, and whose economics are now flipping from diesel to clean. The precedent is the point: the Global South is the completed experiment the OECD is only starting.

Structural moat or temporary bottleneck? Captive power is neither temporary nor a bottleneck — it is a permanent structural feature wherever the grid is weak, now compounding as clean economics make defection cheaper and final. That makes the diesel-to-clean re-equipping a durable, multi-decade market, and the exposed distribution utility a structurally impaired one. The discipline is to separate the captive build that is now the lowest-cost, cleanest option (C&I solar-plus-storage, hybrids) from the legacy diesel it replaces, and to read the defection not as enterprise triumphing over a bad grid, but as a bad grid losing the load that funded it — the same warning the OECD should read off this mirror before it waves its own best customers away.

The Emissions Are Ahead, Not Behind

Its companion piece argued that a fossil asset commits its emissions the day it is financed, and that stranding is that commitment coming due. Follow that logic to the Global South and it inverts the entire Western intuition. The intuition — the one that lets a rich country talk about phasing out coal almost casually — rests on a hidden fact: the OECD fossil fleet is old. Its committed emissions are mostly already spent, and retiring a nearly-depreciated plant strands very little.

The Global South fleet is the opposite in the one dimension that matters. The average coal plant in developing Asia is about thirteen years old, three decades younger than in the United States or Europe, with more than half of it built in the past two decades. Its committed emissions — and its unrecovered capital — lie almost entirely ahead. So the same committed-emissions framework, applied honestly to a young fleet, produces the opposite conclusion: here, stranding is not nearly-costless housekeeping. It means writing off plants fifteen to twenty years before the end of their economic life, at a scale measured in trillions.

A Young Fleet, Still Growing

The lock-in is not a settled inheritance to be managed down; it is a choice still being made, at scale, right now. In 2025 the world built more coal even as it used less: China commissioned 78 GW of new coal — a decade high — India added another 10 GW, and the global pipeline of coal under development grew 12% in a single year, from 633 GW to 710 GW. China and India together accounted for roughly 87% of all new coal capacity begun in 2025.

Every one of those plants resets the clock. Analysts note that if no new coal were added, roughly 90% of Asia’s fleet would be past its twenty-year payback and feasibly retirable by 2040 — the lock-in would age its own way out. But the additions keep the average age low and the committed emissions high, which is precisely why the young-fleet problem is not resolving on its own. The single most important variable is not the retirement rate; it is the build rate, because each new unit re-commits decades of emissions the budget cannot absorb.

Stranding Means Retiring Fifteen Years Early

Because the fleet is young, retiring it early is expensive in a way the Western experience simply does not capture. A net-zero-aligned path retires these plants fifteen to twenty years before their economic end — while the capital that built them is still being repaid. The bill reflects it: decommissioning and replacing a single gigawatt runs around $1.9bn, retiring Indonesia’s current and under-construction fleet is estimated above $114bn, Vietnam’s above $57bn, and the combined China-and-India transition at $3–5 trillion. Average compensation per gigawatt can exceed the cost of building a new plant outright.

South Africa is the instructive exception: a Global South country with an old fleet, whose stranding problem looks far more Western — low committed emissions ahead, retirement as much about reliability as carbon. It proves the point that the “young fleet” framing is about age and financial structure, not geography. Where the plant is young and contracted, the emissions and the money both lie ahead.

The Fleet Has a Constituency to Keep Running

A young, unrecovered, contracted asset does not just cost more to strand; it comes with a powerful coalition determined that it should not be. More than $1 trillion of coal capital is still to be earned back across today’s plants, mostly in Asia, and every party with a claim on that return — state owners, independent power producers, lenders — has an interest in the plant running to term. Long-term power-purchase agreements harden that interest into contract: in Vietnam, such agreements govern roughly half the fleet, guaranteeing revenue regardless of dispatch.

It Is a Financing Problem, Not a Technology One

Here is the conclusion that reorders the investment case: unwinding the young fleet is fundamentally a capital problem, not a technology one. The clean alternatives exist and are cheap; what is scarce is the money to pay off a young plant so it can close before its PPA expires. And the dedicated vehicle for that money is running far behind the need. Three years after Indonesia’s $20bn Just Energy Transition Partnership, no plant has been retired, decommissioning is not scheduled to begin until 2035, and the country’s coal capacity is projected to keep rising to 2030 first.

The unwinding toolkit is nonetheless where the action is: the Asian Development Bank’s Energy Transition Mechanism blends concessional and commercial capital to accelerate retirement; refinancing and securitisation structures shorten plant lives; and abate-in-place retrofits — ammonia and biomass co-firing on young supercritical units — offer a middle path that keeps the asset while cutting its emissions, for which a young, modern fleet is actually the better candidate. The scarce input across all of it is not the clean megawatt. It is the capital to retire the dirty one early, in a region where that capital is dearest.

Positioning: Own the Unwinding, Price the Lock-In

The OECD play was to price the derated, near-depreciated asset and let it retire cheaply. The Global South play is the opposite: finance the expensive unwinding of a young fleet, and price the lock-in that every new build deepens.

Reading It Through the Frameworks

Where the conclusion inverts. The committed-emissions framework is identical on both sides of the mirror — emissions and stranding fixed at financing — but the fleet’s age flips the answer. In the OECD it licenses a cheap, near-complete phase-out. In the Global South the same logic reveals a multi-trillion, decades-long unwinding of assets still being built, defended by unrecovered capital, contracts and the need to power growth. The money is not in owning a soon-to-strand asset cheaply; it is in financing the unwind and avoiding the fresh lock-in.

Structural moat or temporary bottleneck? The bottleneck is capital, and it is bindingly real because it sits exactly where the cost of capital is highest. That makes the unwinding a structural, policy-and-blended-finance-driven market rather than a market-clearing one — it will not resolve on price alone. The discipline is to separate the plant that can be gracefully unwound (financeable owner, expiring PPA, retrofit-ready) from the one whose lock-in is rigid (fresh build, long contract, captive use), and to read the build rate, not the retirement rate, as the true measure of whether the problem is getting better or worse.