Speed Became the Only Priority

Something unusual has happened in infrastructure policy: the two halves of the political spectrum have converged on the same objective from opposite premises. The deregulatory right wants to build faster by removing rules; the “abundance” left wants to build faster by fixing a government it believes has become better at blocking than building. They disagree about almost everything except the verb. The result is that speed — not cost, not returns, not even reliability — has quietly become the organising priority of the entire buildout.

The evidence is not rhetorical. In May 2025 the Supreme Court decided Seven County Infrastructure Coalition v. Eagle County by a unanimous 8-0 margin, holding that environmental review under NEPA is a procedural cross-check, not a substantive roadblock, narrowing the required scope of review and instructing courts to give agencies substantial deference. Lower courts have since extended it and litigants now face materially higher hurdles. Alongside it, the Council on Environmental Quality’s NEPA regulations were rescinded outright (interim rule in early 2025, final rule January 2026); a bipartisan Build America Caucus — nicknamed the “abundance caucus” — formed to push permitting reform; the House advanced a SPEED Act to rewrite NEPA further; and a Democratic governor signed a rollback of California’s signature environmental statute to accelerate housing. The direction of travel is not contested. Only its price is.

Speed Is a Transfer, Not a Technology

Here is the reframe underneath the whole piece. Making a project faster does not, by itself, make it cheaper or better. It reallocates two things — who decides and who bears the residual risk — and those two things are exactly what the project’s financial model does not contain.

None of this is an argument that the transfers are wrong. Several are plainly worth making — a review process captured by opponents to stop projects on their merits is its own failure. The point is analytical, not moral: a faster project is a differently-distributed project, and whoever ends up holding the residual has both a grievance and, increasingly, a vote. That is the mechanism by which speed generates its own opposition.

The Bottleneck Is Governance, and Governance Doesn’t Scale With Capital

The reason speed is so hard to buy is that the binding constraint is not money or technology. It is process. The clearest case is the interconnection queue — the line a power project must wait in to connect to the grid. At the end of 2023 the active queue peaked near 2,600 GW, roughly twice the entire installed US generating fleet. It has since eased to about 2,061 GW in 2025, but the decline is the tell: it came largely from project withdrawals and from two grid operators pausing new intake, not from faster processing.

Follow the completion rate and the queue stops looking like a pipeline at all. Of the projects that entered US queues between 2000 and 2019, only about 19% reached commercial operation; for solar the figure is closer to 14%. More than 90% of applications arrive with deficiencies requiring rework. FERC’s Order 2023 — the largest interconnection overhaul in decades, moving to cluster studies and “first-ready, first-served” with higher deposits — is a real improvement, but it buys speed the same way everything else does: by raising the bar to enter, which screens out the speculative and the under-capitalised alike.

The lesson generalises past the queue. A governance bottleneck — a study process, a permit, a cost-allocation fight — cannot be relieved by pouring capital into it. More money in the queue just makes the line longer. This is why the buildout’s pace is set by institutions, not balance sheets, and why the scarce input in the entire supercycle is administrative and political capacity, not investable dollars.

The Bill Lands on Whoever Didn’t Choose

When speed’s cost is socialised, it lands on people who never voted for the project and capture none of its upside. The starkest example is the collision between data-center demand and the grid. In two years, PJM’s capacity auction — the price paid to keep enough generation available for 67 million people — went from $28.92 to $329.17 per MW-day, an escalation of more than 800%, hitting the FERC price cap and staying there for three consecutive auctions while still falling short of the reliability requirement.

The independent market monitor attributed 63% of that increase to data-center load — roughly $9.3bn in a single year, recovered from ordinary customers who did not build the data centers. The average PJM household faces an estimated $70 per month in higher bills by 2028. Nationally, utilities requested a record ~$31bn in rate increases in 2025, double the prior year, and residential electricity prices rose 7% in a single year. The cost of building fast for one customer is being paid, quietly and at scale, by everyone else on the wire.

Legitimacy Is a Depleting Reserve

Here is why the socialised bill is not just an equity problem but a risk problem. Speed is bought by spending public consent, and consent is finite. Each acceleration — a narrowed review, a fast-tracked load, a rate rider nobody voted for — draws down a reserve of trust. When it runs low, the response is not a strongly-worded letter; it is a moratorium, a rate-class carve-out, a ballot measure, a revived lawsuit. The backlash is the mechanism by which fast projects become slow ones, retroactively.

That reserve is visibly draining. At least 23 states have already legislated on who pays for the data-center buildout. Virginia is creating a separate data-center rate class; Pennsylvania is running a precedent-setting rate case; Ohio has enacted an 85% minimum-bill ratchet; Oregon, Virginia and Pennsylvania have built frameworks with long contract terms, take-or-pay minimums and full collateral, precisely so that speculative load cannot strand ratepayers. In November 2025 PJM stakeholders voted down every major proposal to make data centers carry more of their own cost, pushing the decision to the board — which chose an incentive route: an expedited connection track for large loads that bring their own generation, and curtailment for those that do not.

What the Market Doesn’t Price

An internal rate of return captures the cost of the build and the value of the offtake. It does not capture the durability of the permission to build — and in a regime organised around speed, that permission is the volatile variable. The projects that will actually get built fast are not the ones with the best headline economics. They are the ones that have already paid the political price of speed up front.

Reading It Through the Frameworks

Where does policy become the cash flow? Directly, and on both sides of the ledger. Permitting reform and interconnection reform are pro-speed policy converting into shorter timelines and higher project throughput. Rate-allocation rules, moratoria and rate-class carve-outs are anti-socialisation policy converting into re-priced cash flows for whoever was carrying the transfer. The same regime that accelerates a project can, twelve months later, re-open its economics — and the trigger is political salience, not project performance.

What kind of risk is it? Speed-driven assets carry a distinctive profile: strong structural tailwind, genuine timeline benefit, and a reversal risk tied to legitimacy rather than technology. The failure mode is not that the plant breaks; it is that the permission is withdrawn or re-priced after the capital is committed. That argues for underwriting on the durability of the consent — the alignment of who pays with who benefits — and treating the fastest-looking deal with the thinnest political cover as the most fragile, not the most attractive.

The Number That Means the Least

Every ceasefire produces a number. Within weeks a joint World Bank, UN and EU team publishes a Rapid Damage and Needs Assessment, and the figure leads the coverage: hundreds of billions, some multiple of the country’s entire economy. The number is real, and it is the wrong place to start. It measures what was lost, not what can be built — and treating it as a pipeline is the first and most common error in post-conflict investing.

The assessment figure grows monotonically and never converges, because it is a running tally of accumulated destruction, not an estimate of a finite job. Ukraine is the clearest illustration. The first assessment in September 2022 put reconstruction and recovery at $349bn. Each subsequent edition has ratcheted upward as the war continued — $411bn, then $486bn, then $524bn — and in February 2026 the fifth assessment set it at almost $588bn over the next decade, nearly three times Ukraine’s projected 2025 GDP.

Read the two magnitudes together. Against a need that has grown by $239bn across four assessments, roughly $20bn of urgent repair and early-recovery work has actually been financed — and that during active conflict, when energy and transport damage is still compounding faster than the ledger can absorb it. RDNA5 records direct damage above $195bn, energy-system needs near $91bn, transport above $96bn, and demining alone at nearly $28bn. None of those numbers is a project. Each is a category of loss awaiting a financing structure that, for the most part, does not yet exist.

The Sequence That Cannot Be Skipped

Reconstruction capital arrives in a fixed order, and each layer is a precondition for the next. The order is not a preference or a best practice; it is a risk gradient. Grant money tolerates the highest risk and the lowest return, so it goes first; equity demands the opposite, so it comes last. Skip a rung and the capital above it does not stick — it disburses, fails to find a stable operating environment, and either stalls or is written down.

Syria is the counter-example that proves the rung. A year and a half after the fall of the Assad government, the country has attracted diplomatic normalisation and a wave of Gulf, Turkish and Western investment pledges — rung-four activity, promised before rungs one through three are secure. Meanwhile the UN’s 2026 humanitarian appeal, the largest regional plan at $2.8bn, sits roughly 20% funded. The bridge is being withdrawn while the far bank is still being advertised. Until the relief-to-recovery handoff is financed and the state can absorb it, the pledges are options on a future nobody has underwritten — not capital in the ground.

Three Wars, Three Bottlenecks

The same ladder governs every case, but each conflict jams on a different rung — and the binding constraint, not the headline total, is what an allocator needs to read. Ukraine’s is time and security: it is rebuilding under fire, so rung three cannot fully open. Gaza’s is governance and access: the assessment is explicit that reconstruction must be Palestinian-led under an agreed political framework, and that condition, not the dollar figure, gates disbursement. Syria’s is the bridge: sanctions relief and pledges without the humanitarian floor and institutional capacity to convert them.

The scale ranking inverts the urgency ranking, which is the point. Gaza is the smallest headline number and among the hardest to deploy against, because the constraint is not capital but the political precondition. Syria is ten times its own annual output in need and rich in pledges, yet stalled at the bridge. Ukraine is the largest and, paradoxically, the most bankable of the three at the margin — because it is the one where a genuine de-risking architecture has actually been built. Which is the subject of Sections 04 and 05.

The Money That Cannot Move

The largest single pool of potential Ukrainian reconstruction capital is not a fund, a facility or a donor. It is roughly €300bn of Russian sovereign assets immobilised across the West since 2022, about two-thirds of it in Europe, with €193bn held at Euroclear in Belgium alone. On paper it is most of a decade of reconstruction sitting in a single depository. And it is the clearest case in modern finance of a headline number that cannot become a flow.

What is actually being used is the yield, not the principal. The G7’s Extraordinary Revenue Acceleration mechanism raised $50bn serviced by the windfall profits those frozen assets throw off — a structure that touches the income and leaves the corpus untouched. In 2025 the Euroclear holdings generated roughly €3.9bn in interest, itself down about a quarter year-on-year as the ECB cut rates. The reconstruction relevance of a €300bn pool that yields single-digit billions, and falls when rates fall, is limited by construction.

The attempt to reach the principal is the “reparations loan” — a proposal to lend Ukraine up to €140bn against the frozen balances, repayable only once Russia pays reparations. It has been debated for over a year and repeatedly deferred over Belgium’s exposure to legal retaliation and over the precedent for sovereign-asset immunity. At the December 2025 European Council the leaders declined to pull that lever and instead agreed roughly €90bn ($106bn) of support for 2026–27 raised on capital markets, with the assets to remain frozen until reparations are paid. The IMF puts Ukraine’s 2026–27 financing need near €137bn; the market-borrowing route covers the state’s survival, not its reconstruction.

The De-Risking Layer: Who Eats the Tail

Private capital does not price a war. It prices whether someone else will eat the war. When conflict erupts, commercial insurers withdraw first and fastest — and without insurance there is no lending, no equity, no rung four. The entire question of private participation reduces to one mechanism: political-risk insurance that transfers the war-and-civil-disturbance tail from the investor onto a public or multilateral balance sheet.

That layer is being built in Ukraine in real time, and the pieces are worth naming because they are the template for every future reconstruction. The World Bank’s guarantee arm, MIGA, is among the very few providers of war-risk cover in the country, operating through its SURE trust fund and writing long-dated policies — for example a €9.1m guarantee on an industrial park near Lviv, covering up to ten years against war and civil disturbance. The US Development Finance Corporation has built a Ukraine political-risk portfolio in the high hundreds of millions. The US–Ukraine Reconstruction Investment Fund, seeded with $75m from each side and targeting critical minerals, energy, ICT and infrastructure, made its first investment in March 2026; in June a MIGA co-insurance layer was added on top of it, extending cover beyond the fund’s own holdings. And in February 2026 Aon and a Ukrainian insurer stood up a $25m DFC-reinsured facility for smaller enterprises — one node in what Aon describes as more than $490m of public and private war-risk capacity assembled for the market.

Notice what each of these is doing. None of them is investing in reconstruction. They are manufacturing the condition under which someone else will. The reconstruction that gets privately financed is not the highest-need slice or the highest-return slice; it is the slice where a public guarantee has already removed the tail. As the World Bank’s president has put it, the private sector will only come in through the right risk-reduction — the insurance is the mechanism, and the mechanism is the market.

Who Builds, and When

Follow the sequence to its conclusion and the shape of the opportunity resolves. First movers are not private investors; they are the multilaterals, DFIs and bilateral donors who finance rungs one through three and write the guarantees. The earliest genuinely commercial capital tends to be strategic rather than financial — contractors, engineering firms, and equipment and materials suppliers whose exposure is contractual and paid on delivery, not equity held through the war-risk window. Insurance and reinsurance names sit alongside them, because the entire edifice runs on their capacity. Pure infrastructure equity — the fund buying the operating asset and holding it for yield — is genuinely last, and enters only against a completed guarantee and an enforceable contract.

This is why sector matters less than structure. The most bankable early reconstruction assets are the ones whose cash flow does not depend on a functioning domestic economy: cross-border energy and transport corridors, export-facing critical-minerals projects, and telecom, where demand is inelastic and revenue can be ring-fenced. The least bankable, regardless of how acute the need, are the assets whose revenue is a claim on a population that has just lost 84% of its output — municipal water, local distribution, social housing — which is precisely why those remain on the public rungs for years.

Reading It Through the Frameworks

Where does policy become the cash flow? Almost entirely, and unusually indirectly. Reconstruction demand is not a market response to a price signal; it is a public and multilateral decision to absorb a risk the market will not. The cash flow that reaches a private balance sheet is manufactured upstream — a guarantee written, a tail reinsured, a concessional tranche laid down first — and it is durable exactly as long as the public commitment behind it is. Guarantee-facility announcements, not damage assessments, are the leading indicator of a deployable pipeline.

What kind of risk is it? Reconstruction investment carries an unusual profile: enormous headline need, strong political backing, and a binding constraint that is political and institutional rather than technical. The failure mode is not that the asset does not work; it is that the sequence stalls — relief withdrawn, a governance precondition unmet, a frozen pool that never thaws. That argues for underwriting these assets on their contracted and insured economics, and treating the strategic narrative as the thing most likely to be revised.

A Different Question From Adequacy

The three preceding pieces asked variations of the same question: is there enough? Enough generating capacity, enough water in the basin, enough refined mineral supply. This piece asks something distinct and often confused with it: who controls the flow, and what can they do with that control?

The distinction is not academic. A resource can be physically abundant and strategically insecure — there is no global shortage of natural gas, but there are a small number of routes and suppliers through which much of it must pass. Conversely, a resource can be genuinely scarce and yet pose little security risk, if what remains is widely distributed among many suppliers. Adequacy is a question about quantity. Security is a question about concentration, and about intent.

Governments have converged on that view. At the International Energy Agency’s 2026 ministerial, ministers stated plainly that energy security is integral to national security, and flagged as emerging priorities the growth in electricity demand, resilient and diversified supply chains, and infrastructure resilience — alongside the continued importance of oil and gas. That combination is the whole story of this piece in a sentence: the security agenda is now simultaneously about the old system and the new one.

The Chokepoint Remains

Whatever the long-run trajectory, the near-term reality is that a very large share of the world’s traded energy still moves through a handful of maritime passages. The Strait of Hormuz alone carries roughly 20% of global oil and LNG flows — a concentration with no equivalent in any other traded commodity.

That vulnerability was tested during a disruption running from late February into April 2026, which produced exactly the pattern the literature predicts: elevated prices, coordinated multilateral response, and importing countries moving to secure supplies for themselves. The behavioural response is worth noting precisely because it is so consistent across episodes — countries hoarded supply rather than allowing markets to clear, which the International Monetary Fund’s managing director characterised as counterproductive, warning against measures that worsen market disequilibrium.

The historical evidence on that point is unusually clear. Analysis of the 1970s oil embargoes finds that unilateral export restrictions typically fail to achieve their stated energy security objectives while generating substantial spillover effects onto other nations. Strategic reserve releases show a similar profile: roughly sixty days of price moderation, with diminishing returns beyond ninety days unless accompanied by actual supply restoration or demand reduction.

The Reframe: Flow Security and Stock Security

Here is the structural shift underneath everything else in this piece, and it is the reason energy security is being reconsidered rather than merely reinforced.

A fossil-based energy system creates a flow dependency. Fuel must be imported continuously, forever, and the vulnerability renews with every shipment. Interrupt the flow and the system degrades within weeks. That dependency is permanent by construction: it does not diminish with time, investment or good behaviour.

A renewables-based system creates something different — a stock dependency. The turbines, panels, transformers and batteries must be imported, and the minerals inside them come from concentrated sources. But once installed, the asset runs for twenty-five years on domestic sunlight or wind. The dependency is concentrated at the moment of construction and then substantially discharged.

The Vulnerability Bites During the Escape

The uncomfortable feature of the flow-to-stock trade is its timing. The equipment dependency is at its most acute precisely during the transition — which is to say, during the period when a country is trying to escape the fuel dependency. And the two vulnerabilities are correlated, because the same disruptions that spike fuel prices also raise shipping costs for clean-energy equipment, tighten supply chains and increase perceived risk on large infrastructure projects.

The consequence, documented in the 2026 transition assessments, is that security shocks directly test the bankability of renewable, grid and storage investments at the very moment those investments are most needed. Higher inflation from an energy crisis feeds into higher interest rates, which fall hardest on capital-intensive clean technology — and in much of Southeast Asia, where the cost of capital already runs around twice that of advanced economies, this materially weakens risk-adjusted returns and slows deployment.

Layer on the concentration described in the minerals piece — refining capacity dominated by a small number of suppliers, export controls now extending to processing equipment — and the picture is clear. The market concentration of critical minerals and key energy technologies has itself become the strategic vulnerability, which is why it now sits alongside oil and gas on ministerial agendas rather than in a separate industrial-policy conversation.

Same Shock, Opposite Responses

The most important thing to understand about energy security as an investment driver is that it does not point in a single direction. The same disruption produces opposite policy responses in different places, and the divergence is now visible enough to be modelled explicitly.

For some importing economies, insecurity accelerates the transition: domestic renewables reduce import exposure, so the security case and the climate case align. For others, insecurity does the reverse — it renews interest in domestic coal, which is also a sovereignty play, and a considerably faster one. In Asia, where economies are heavily reliant on LNG imports and acutely exposed to Hormuz transit risk, volatile gas prices and supply uncertainty have already triggered renewed reliance on coal.

The 2026 scenario framing captures the fork: under prolonged disruption, coal’s security role in Asia expands further and the gap between climate targets and actual investment widens materially. Under fragmented regional responses, Western economies accelerate electrification while Asian economies pursue a dual-track model — expanding renewables capacity while simultaneously extending fossil systems as insurance.

Who Can Afford Security

Energy security is usually discussed as though it were a uniform national problem. It is not. Import-dependent economies face asymmetric vulnerability — not only through price exposure but through supply uncertainty and fiscal stress simultaneously — and their capacity to absorb that varies enormously.

Wealthier importing nations can cushion a shock through strategic reserves and fiscal capacity. Lower-income import-dependent economies face a genuinely harder trade-off between energy access, fiscal stability and transition investment, with dollar-denominated energy costs amplifying currency pressure on top. The institutions monitoring this have identified Asia, Sub-Saharan Africa and South Pacific island states as carrying heightened vulnerability, with landlocked economies additionally exposed through transit dependence.

The consequence deserves to be stated bluntly, because it is where security policy and equity intersect. A security shock pushes the countries with the least fiscal space toward the highest-carbon available option, because coal is domestic, cheap to dispatch and requires no foreign exchange. This is simultaneously an equity issue and a security issue, and it determines which countries sustain transition momentum and which are forced into regressive energy choices under pressure.

The Response That Isn’t Happening

One historical comparison is worth drawing, because it undercuts a comfortable assumption. The 1970s oil shocks produced a dramatic increase in public energy research and development budgets, explicitly aimed at reducing oil dependence. That surge is a large part of why the technologies now being deployed exist at all.

The equivalent response is not currently visible. Despite a threat landscape that is broader than the 1970s — now spanning supply chains, critical minerals and grid resilience as well as fuel — public investment has not kept pace. Venture capital investment in energy fell to around $27 billion for the third consecutive year, while close to 30% of all venture funding now flows to artificial intelligence rather than energy technologies.

There is an irony worth naming: the sector absorbing that capital is itself driving much of the electricity demand growth described in the Build thread. Capital has rotated from the technologies that would solve the energy problem toward the technology that is intensifying it.

Reading It Through the Frameworks

Where does policy become the cash flow? More completely here than anywhere else in the Strain thread. Energy security spending is not a market response to a price signal — it is a sovereign decision to pay above market cost for reduced dependence. That creates demand for assets whose economics would not otherwise close, and it is durable exactly as long as the political perception of threat is. Legislative activity on critical mineral supply, domestic production and LNG capacity translates fairly directly into project pipelines.

What kind of risk is it? Security-driven investment carries an unusual profile: strong near-term political support, weak underlying economics, and a reversal risk tied to the threat receding rather than to the technology failing. The LNG lock-in point is the clearest illustration — commitments made during a crisis persist regardless of whether the crisis does, because construction outlasts sentiment. That is a strong argument for underwriting these assets on their contracted economics rather than their strategic rationale.

The Assumption Underneath Everything

There is a single assumption buried in the foundations of nearly all infrastructure engineering, and it has a name: stationarity. It holds that the statistical properties of the climate — the mean, the variance, and crucially the behaviour of the extremes — do not change over time. Under stationarity, a long enough record of the past is a reliable description of the future.

That assumption is what makes conventional design possible. Drainage systems are sized using intensity-duration-frequency curves built from historical rainfall records. Culverts, bridges, storm sewers, embankments, spillways and flood defences are all specified against return-period values derived the same way. In the United States, a national precipitation-frequency atlas has long served as the benchmark; equivalent standards exist across Europe and elsewhere. Every one of them is a statistical summary of what the weather used to do.

Stationarity is no longer true, and this is not a contested point among the people who study it. Warming has reached roughly 1.55°C above the 1850–1900 baseline, sea-level rise is accelerating, and precipitation extremes are intensifying. The consequence is uncomfortable and specific: a large share of the infrastructure now being built is being specified against a climate that has already ceased to exist.

What “One-in-a-Hundred” Actually Means

The phrase does an enormous amount of work in infrastructure finance and insurance, and it is widely misunderstood. A hundred-year flood is not an event that happens once a century. It is an event with a 1% probability of being exceeded in any given year — and that probability is not measured from the river. It is estimated by fitting a statistical distribution to a historical record.

A concrete illustration makes the scale clear. Analysis of stormwater design at US Air Force installations found that at one site, a 24-hour storm carrying a 10-year design value of 14.9 centimetres would, under a high-emissions scenario later this century, recur as frequently as every three to four years. The engineering is unchanged. The specification is unchanged. The protection has fallen by roughly two thirds.

The Scale of the Revision

The physics behind this is unusually well understood. Warmer air holds more moisture — roughly 7% more rainfall intensity per degree of warming — which loads the extreme tail of the precipitation distribution faster than it shifts the mean. Extremes move first and move most, which is precisely the part of the distribution infrastructure is designed against.

Applied globally to transport assets, the result is stark. Under approximately 2°C of warming by mid-century, 43.6% of global transportation assets are expected to see their extreme-rainfall design return period fall by at least a quarter — equivalent to a 33% increase in annual exceedance probability. Under roughly 4°C by late century, that rises to 69.9%. On a broader measure, nearly 88% of global road and rail assets face more frequent extreme precipitation by mid-century.

Why This Compounds Rather Than Adds

Here is the part most often missed, and it is where the risk becomes materially larger than intuition suggests. Infrastructure is not exposed to a single year’s probability. It is exposed across a service life measured in decades, and exceedance probability compounds over that life.

Take an asset with a fifty-year design life, protected to a genuine hundred-year standard. The probability that it experiences at least one exceedance across its life is not 1% — it is close to 40%. Now suppose warming compresses that hundred-year event into a thirty-year event. The lifetime probability of at least one exceedance rises to roughly 82%.

Two Moving Targets, Not One

The secular warming trend is the more discussed problem, but it is not the only thing moving. Superimposed on it is natural climate variability — the large-scale oscillations, of which the El Niño–Southern Oscillation is the most consequential, that modulate rainfall, temperature and storm behaviour on multi-year cycles across much of the world.

This matters for infrastructure in two distinct ways. First, the oscillations dominate year-to-year outcomes even where the trend dominates decade-to-decade ones, which means a single season tells you almost nothing about a structural shift. The 2025 Indian monsoon is a clean example — an early arrival cut cooling demand sharply and produced an anomalously weak year for electricity growth without anything structural having changed.

Second, and more troubling, is that the behaviour of the oscillations themselves may be shifting as the system warms. If the amplitude, frequency or teleconnection patterns of these cycles change, then even a perfectly climate-adjusted design standard built on the recent past would be estimating from a variability regime that is itself in motion.

Why the Standards Have Not Caught Up

If the flaw is this well documented, the obvious question is why design codes still rest on it. The answer is not negligence, and understanding it explains why the gap will persist for years.

• Deep uncertainty. Adjusting a standard requires agreeing on a specific probability distribution for future extremes at a specific location. Experts and decision-makers frequently cannot agree on that distribution, or on which scenario to specify against. Stationarity, whatever its faults, produces a single defensible number.
• Liability and defensibility. An engineer who designs to the published standard is professionally protected. One who departs from it — even in the direction of greater safety — is exposed if the additional cost is challenged. Codes change slowly for reasons that are institutional rather than technical.
• The absence of an adopted alternative. Governing authorities have largely not adopted climate-informed methodologies for estimating rainfall recurrence. The most consequential gap is not that the old standard is wrong; it is that no agreed replacement exists to design against instead.
• Compounding non-climatic change. Urbanisation increases impervious surface and pushes development into floodplains, raising runoff and concentrating assets in hazardous places. That amplifies effective non-stationarity independently of the climate.

The emerging response is pragmatic rather than elegant: apply a climate adaptation safety factor — design to the historical standard, then add an explicit margin for the shift. It sidesteps the unresolved argument about distributions by treating the uncertainty as something to buffer rather than something to resolve. It is imperfect, and it is considerably better than pretending the mark on the wall still means what it did.

Reading It Through the Frameworks

Where is the physical risk mispriced? This is arguably the largest and most systematic mispricing in the entire section, precisely because it is invisible. An asset carrying a documented hundred-year protection level may in reality carry thirty-year protection, and nothing in its documentation would reveal that. The risk has not been assessed and rejected; it has been recorded at a value that was accurate when it was written. Every asset specified before roughly the last decade carries some version of this.

What does it do to valuation? Three things, all of which run through the cash flow rather than the engineering. It raises expected maintenance and repair over the life. It raises insurance cost, and eventually raises the question of whether cover remains available at all. And it introduces early-obsolescence risk — the possibility that an asset requires substantial retrofit well before the end of its accounting life, which is a capital call that no depreciation schedule anticipated.

The Wrong Mental Model

The phrase “critical minerals” invites a scarcity story — a picture of finite deposits depleting toward exhaustion, with nations racing to claim what remains. That framing is almost entirely wrong, and holding it leads to systematically bad conclusions about where the risk actually sits.

The minerals underpinning the energy transition are, with few exceptions, geologically abundant and widely distributed. Reserves are not the binding constraint. What is scarce is the industrial capacity to turn ore into a material a manufacturer can actually use — battery-grade graphite, separated heavy rare earths, refined lithium chemicals. That capacity is concentrated to a degree that has no parallel in other commodity systems, and it is concentrated in a small number of places.

This is the third piece in the adequacy sequence, and it inverts the pattern of the first two. Power adequacy is about whether enough capacity exists. Water adequacy is about a genuine physical limit in a specific basin. Mineral adequacy is about neither — it is about a processing bottleneck that is entirely man-made, and therefore, in principle, entirely fixable. The interesting question is why, after years of concerted effort and enormous capital, it is not being fixed.

The Chokepoint Is Downstream

Concentration exists at every stage of the mineral chain, but it is far more acute in the higher-value stages. Mining is meaningfully concentrated: in 2025 the Democratic Republic of the Congo accounted for roughly 74% of global cobalt mine production, Indonesia for about 67% of nickel, and China for around 69% of rare earth mining. Those are large shares, but mines can be opened elsewhere, and are being.

Refining is a different order of problem. Excluding rare earths, the average share held by the top refining country reached 72% in 2025, up from 70% in 2023. That direction of travel is the single most important fact in this piece. Through a period of unprecedented policy attention, subsidy and private investment aimed squarely at diversification, dependence on the dominant refiner increased.

The mechanism behind that is worth stating precisely. Over the past two years, the top refining countries — Indonesia for nickel, China for most other key energy minerals — accounted for more than three quarters of all growth in refined supply. In several markets, including manganese, nickel and graphite, virtually all supply growth came from the dominant supplier. Demand grew, and the incumbent met it, because the incumbent was the only participant able to add capacity at speed.

The Risk Map, Mineral by Mineral

The exposures are not uniform, and treating “critical minerals” as a single asset class obscures risks that differ by an order of magnitude.

Cobalt deserves particular attention, because it is the cleanest illustration of the theme. The Democratic Republic of the Congo holds the world’s largest cobalt reserves. The deficit forecast to emerge by 2035 exists not because the ore is exhausted but because an export quota has restricted access to it. That is a supply constraint created by an administrative decision, which means it can be reversed by another one — and it should be modelled as policy risk rather than resource risk.

The Second Chokepoint: Equipment and Know-How

If refining capacity is the bottleneck, the obvious response is to build refineries. Here the problem acquires a second layer that receives far too little attention.

Outside the dominant supplier, equipment providers are limited, lead times are long, and the accumulated technical know-how takes years to develop. Only a handful of suppliers worldwide provide key battery-grade graphite processing equipment. Ultra-high-purity gallium refining sits with a similarly narrow group. Building a refinery is not merely a matter of capital and permits; it requires specialised plant that few can manufacture and operating expertise that takes a decade to accumulate.

That vulnerability has become an explicit instrument. New export controls have been extended to cover equipment for processing rare earths — not the minerals themselves, but the machines needed to refine them and to produce permanent magnets. The effect is to constrain emerging projects precisely at the point where they would otherwise begin displacing incumbent supply.

The Economics Almost Nobody States

Here is the finding most likely to change a reader’s view, because it contradicts the assumption underneath most commentary on this subject — that diversified supply chains would be prohibitively expensive.

Critical minerals account for a remarkably small share of the price of the products that depend on them. Minerals make up roughly a quarter of battery cell costs but only about 3% of the price of an average electric vehicle. Rare earths represent around 40% of the cost of a permanent magnet but less than 1% of a vehicle’s value.

The One That Is Working

A piece in this thread should be careful to report the counter-evidence, and in this case it is genuinely encouraging. Rare earth refining — historically the most concentrated link in the entire chain — is the exception to the worsening trend.

The mechanism matters more than the numbers. This did not happen through market forces — it happened through targeted policy and sustained investment support aimed at a specific stage of a specific chain. Diversification is demonstrably achievable; it simply requires deliberate, sustained and narrowly-focused intervention rather than general encouragement. That is a replicable template, and graphite is the obvious candidate for it.

The qualification is that the projected path to 70% depends on planned projects arriving on schedule — and the export controls on processing equipment described earlier are aimed precisely at making that harder.

Reading It Through the Frameworks

Where is the physical risk mispriced? In the assumption that mine ownership confers supply security. A portfolio can hold diversified mining exposure and remain entirely concentrated at the refining step — a single point of failure that does not appear in any geographic breakdown of assets. That gap between apparent and actual diversification is the mispricing this piece exists to name.

Where does policy become the cash flow? More directly here than almost anywhere. Roughly a hundred new export measures since 2020, quotas that manufacture deficits, and controls extending to processing equipment mean that mineral availability is now substantially a function of trade policy. Prices rebounded through 2025 and early 2026 as supply tightened, amplified by those restrictions. Meanwhile 58 bilateral partnerships signed since 2022 are actively redrawing the map — and the risk that trade fragments into competing blocs is a live one, with price bifurcation in heavy rare earths already visible.

Is this a moat or a bottleneck? Both, in the primer’s sense — and unusually clearly. For an incumbent refiner, decades of accumulated process expertise and equipment supply constitute a moat that capital alone cannot cross quickly. For everyone else it is a bottleneck. The durable value sits with whoever holds qualified processing capability, which is a far narrower group than those holding reserves.

The Other Water Problem

There are two water stories running through this section, and they are almost always conflated. The first is compliance — lead service lines, PFAS limits, a century-old network reaching the end of its life. That is a regulatory capex cliff with a deadline attached, and it has its own piece. This is the second story: scarcity. Not whether the pipes meet the standard, but whether there is enough water in the basin to run what has been built beside it.

The distinction matters because the drivers are unrelated. Compliance risk is created by a rule and resolved by spending. Scarcity risk is created by hydrology and cannot be resolved by spending at all — only relocated, rationed, or engineered around at a cost in energy. A regulator can extend a deadline. A basin cannot extend its runoff.

Two Systems That Are Actually One

The reason water belongs in an infrastructure section at all is that the power system and the water system are not two systems. They are one system, coupled in both directions, and neither sits upstream of the other.

Power needs water. Thermal generation — coal, gas, nuclear — runs on the steam cycle, and steam must be condensed, which requires cooling. A global assessment of nearly 14,000 thermal plant units totalling some 4,182 GW found freshwater cooling demand to be substantial enough to constitute a first-order sustainability constraint on the fleet. Hydropower is water by definition.

Water needs power. Moving, lifting, treating and pressurising water is energy-intensive, and increasingly so as cities reach further for supply. Every megalitre delivered to a city on a plateau, or desalinated, or recycled to potable standard, arrives as a converted quantity of electricity.

Averages Lie. Basins Are the Unit.

The single most common mistake in water analysis is to reason at national scale. Water is not fungible across geography in the way capital or even electricity is — moving it any meaningful distance or elevation is prohibitively expensive, which means a country can be comfortably water-rich in aggregate and have an asset standing in a basin that is running out.

The correct unit of analysis is therefore the catchment, not the country — and increasingly, the catchment in a specific season and a specific year. That is the discipline this piece exists to establish, and it applies identically to a power plant, a data centre, a steel mill or a semiconductor fab.

The consequence for the build-out is direct: roughly 40% of the world’s data centres are estimated to be operating in regions already facing high water stress. That is not a projection about a future risk. It is a description of the current siting map, assembled over a decade when water was not among the variables the industry optimised for.

Paper Rights and Wet Water

The clearest illustration of what water scarcity actually does to infrastructure is the Colorado River, and it is worth walking through carefully because the mechanism generalises.

The river supplies water and hydroelectricity to some 35 to 40 million people across seven US states and Mexico, governed by a body of compacts and agreements collectively known as the Law of the River. That framework allocates 15 million acre-feet annually to the recipient states, plus a further 1.5 million to Mexico. The allocation was agreed in 1922.

Here is the defect at the centre of the system. The 1922 negotiators set those volumes using flow records from what turned out to be the wettest sustained period in the basin since roughly 1520. The entitlements were calibrated to an exceptional river, and written in perpetuity. Everything since has been a slow collision between that parchment and the hydrology.

The 2000–2022 megadrought made the gap unavoidable: reservoir levels fell to their lowest since the dams were built, and the federal government declared the basin’s first-ever water shortage. Rising temperatures alone reduced streamflow by roughly 10% during that period, independent of precipitation. And the operating rules governing how the major reservoirs are managed are expiring now, in 2026, with successor arrangements still unresolved — which means the allocation framework for one of the most economically important basins in North America is, at this moment, an open question.

The New Claimant

Into an already-contested allocation now arrives a new and unusually concentrated demand. Direct water consumption by US data centres more than tripled between 2014 and 2023, reaching roughly 17.5 billion gallons — some 66 billion litres by one national laboratory’s estimate — and is projected to potentially double again by 2028.

Three features make this demand more disruptive than its absolute volume suggests. It is geographically concentrated, landing in specific counties rather than spread across a state. It is seasonally peaked — daily water demand for evaporative cooling can spike six to ten times average levels during peak summer, which is exactly when the basin is tightest. And most of what it withdraws is genuinely consumed rather than returned: in evaporative systems roughly 80% of the water evaporates, and in an interior basin very little of that returns as local precipitation.

The competition this creates is being negotiated in public. In the Colorado basin, federal programmes have paid water users — largely farmers — to fund conservation projects that free up supply, while state regulators have begun setting out negotiating principles for utilities dealing with data-centre developers. One state’s conservation board estimates it must find hundreds of millions of gallons per day in additional supply by 2050 before accounting for new data centres at all.

The Substitution Trap

The obvious response to water scarcity is to stop using water — and the technology exists. Dry and air-cooled systems can cut water consumption dramatically. But the coupling described earlier means this is not a free move.

Air cooling requires significantly more electricity than evaporative cooling to reject the same heat. So an operator conserving water raises power consumption — and that additional power, if generated thermally, consumes water somewhere else in the basin or the next one over. The saving is real but partial, and it converts a water cost into an energy cost. Under water restrictions, systems forced to run hotter also lose efficiency, which compounds the effect.

The magnitude is not marginal. An assessment of India’s power sector found that an ambitious shift toward renewables combined with improved cooling could cut water withdrawal intensity by as much as 84% by 2030 and consumption intensity by around 25% against a 2014 baseline. Globally, the thermal-fleet study cited earlier concluded that under a best-policies pathway, power-sector water consumption could fall roughly 98% and withdrawal around 95% by 2050. The water problem in the power sector is, to an unusual degree, solvable — by changing what generates the electricity.

Reading It Through the Frameworks

Where is the physical risk mispriced? This is the primer’s first question at its sharpest. Water risk is local, non-linear and legally structured — and it is routinely assessed at national resolution, if at all. An asset’s water position is a function of its basin, its seniority and its cooling technology, none of which appear on a balance sheet. The gap between a documented basin stress level and its expression in an asset’s valuation is precisely the kind of mispricing this section exists to find.

Where does policy become the cash flow? Water allocation is a legal regime, not a market outcome. Seniority systems, curtailment rules and permitting decide who keeps operating in a drought. And the permitting gate is now binding on new construction — in stressed basins, water availability determines whether a project is approved at all, which makes it a siting constraint on par with the interconnection queue.

Where is the moat? In senior water rights in a stressed basin, which are finite, legally protected and cannot be manufactured — the closest thing to an interconnection right in the water system. And in water-free operation, which converts a permitting liability into a siting advantage.

Two True Statements That Sound Contradictory

In its 2026 summer assessment, the North American Electric Reliability Corporation reported that all assessed regions are expected to have adequate resources under normal peak conditions, helped by more than 58 GW of new capacity added since the previous summer. Two months earlier, its long-term assessment found that thirteen of twenty-three assessment areas face elevated or high resource adequacy risk over the next five years.

Neither figure is spin. They answer different questions, and the gap between them is the single most important thing to understand about this topic. NERC’s own director of reliability assessment put it plainly when the summer numbers landed — the improved conditions should not be read as overall reliability risk declining.

Resource adequacy is not a yes-or-no property of a power system. It is a probability distribution, and almost all of the interesting risk lives in its tail. A grid can be comfortably adequate for the summer it usually gets and dangerously short for the summer it gets once a decade — and since the second kind of summer is the one that causes blackouts, headlines and political consequences, the average tells you very little.

What “Adequate” Actually Means

The industry standard is older and stranger than most people assume, and it is worth knowing precisely, because every adequacy claim in the press is built on it.

The benchmark is “one day in ten years” — formally, a loss-of-load expectation of 0.1 events per year. A system is deemed adequate if, on a probabilistic basis, it would be expected to fall short of demand roughly once per decade. That is the criterion behind every reserve-margin target you will see quoted: PJM’s installed reserve margin near 17.7%, MISO’s reference level around 18.6%, and so on. Those percentages are not judgements about prudence — they are the arithmetic output of hitting a one-in-ten standard given a particular fleet.

Three metrics do the real work underneath, and the distinction between them is where the modern debate sits:

Modern assessments are run as large Monte Carlo exercises — thousands of simulations per hour across decades of historical weather, layering load-forecast uncertainty, generator forced outages, and variable renewable output. The sophistication is real. But every one of these models rests on a distribution of assumptions, and the headline that reaches the public is a single word: adequate.

The Number That Broke the Models

What has changed is not the methodology but the input. In a single annual revision, NERC’s ten-year demand outlook moved by an amount without precedent in the three decades it has been tracking.

A 69% upward revision to a ten-year demand forecast, in one year, is not a refinement. It is an admission that the planning baseline was wrong — and it happened because the load additions arriving now are unlike anything the planning process was built to handle. A single data-centre campus can request more power than a mid-sized city, on a timeline shorter than any generation project can be built.

From a Capacity Problem to an Energy Problem

Underneath the demand shock sits a quieter structural change that matters more for how adequacy is assessed. For most of the twentieth century, adequacy was a capacity question: did the system own enough megawatts of dispatchable plant to cover peak demand plus a margin? Thermal plants were available when called, so counting nameplate capacity was a reasonable approximation.

That approximation is breaking down. The 58 GW added over the past year was predominantly solar and battery — resources whose contribution depends entirely on when the system needs them. Solar has substantial capacity value at a mid-afternoon peak and almost none at 8pm. A four-hour battery can cover a short evening ramp and not a three-day cold snap. The relevant question is no longer how many megawatts exist, but whether energy is available in the specific hours the system is short.

Grid operators have adapted the arithmetic — capacity credit and effective load-carrying capability discount variable resources to their reliability contribution rather than their nameplate. And the risk window itself has moved: in solar-heavy systems the tightest hours are no longer the afternoon peak but the evening, as solar output falls while demand remains high. Battery storage directly addresses that short-duration gap, which is why ERCOT expects to have close to 19 GW of it by summer 2026.

Where the Risk Sits, and Why

The regional picture is not uniform, and the causes differ enough to matter. The long-term assessment places MISO, PJM, ERCOT, the WECC Northwest and Basin areas, and SERC-Central in the highest risk category, where planned resources would leave energy shortfalls beyond adequacy targets. The drivers cluster into four recognisable patterns.

• Load growth outrunning additions. The dominant cause, and the one directly traceable to data centres and electrification. ERCOT’s near-term metrics have actually improved, but continued load growth outpaces resource additions in later years.
• Retirements arriving faster than replacements. In the WECC Basin area, summer demand is forecast to rise by more than 1.7 GW over a decade while existing capacity declines by nearly 2.3 GW through retirements — a gap opening from both ends at once.
• Dependence on imports that may not be there. New England’s elevated-risk designation stems substantially from declining firm import commitments and greater reliance on non-firm supply at peak. Neighbours are a resource only when they are not short simultaneously.
• Weather-dependent supply. The Pacific Northwest faces drought-driven hydropower reductions tied to below-normal snowpack — the region’s traditional reliability anchor eroding precisely as summer heat peaks.

That last one deserves emphasis because it generalises. As more of the fleet becomes weather-dependent, supply and demand become correlated through the same variable. A heat dome raises cooling demand, lowers thermal plant efficiency, reduces hydro availability and can coincide with low wind. These are not independent risks that diversify away; they are a single event expressing itself on both sides of the balance.

The Counter-Argument, Taken Seriously

A piece in the Strain thread should be careful not to assume the pessimistic reading is automatically the correct one. There is a substantive critique of these assessments, and it is worth stating.

Independent analysis has argued that NERC’s seasonal assessments are conservative — that supplementing them with interconnection-queue data on resources likely to come online suggests the flagged regions are adequate even under extreme conditions. The mechanism of the critique is straightforward: assessments count resources meeting strict inclusion criteria, and can therefore understate capacity that is genuinely arriving. There is also a structural incentive worth acknowledging: a reliability body faces asymmetric consequences, since being wrong about a shortage is far more damaging than being wrong about a surplus.

Two things temper that. First, the same queue data underpinning the optimistic case has a completion rate closer to one in five, so counting queued capacity as arriving capacity requires care. Second, the conservatism cuts both ways — the assessments also assume normal peak conditions for their headline verdict, which is an optimistic assumption in an era of correlated weather extremes.

Reading It Through the Frameworks

Where is the physical risk mispriced? This is the primer’s first question in its natural habitat. Adequacy risk is quantified in reliability studies and expressed in engineering units, but it becomes a market signal only indirectly — through capacity prices, scarcity pricing and, increasingly, the political response to a shortfall. The gap between a documented tail risk and its market expression is where the opportunity has repeatedly appeared.

Where does policy become the cash flow? Adequacy is the reason capacity markets exist at all: a payment for being available rather than for producing. When adequacy tightens, capacity prices rise sharply — and that revenue stream is a regulatory construct that can be redesigned. Texas has taken a different route, granting operators authority to curtail large new loads in emergencies and funding programmes to expedite reliability resources. Demand-side curtailment as a reliability tool is a significant development, because it makes the load itself part of the adequacy solution rather than only part of the problem.

The Materials Everything Else Is Made Of

Every asset in this section — the reactors, the transmission towers, the data centres, the pipes beneath the street — is built from concrete and steel. That makes these two materials the substrate of the entire infrastructure decade, and it makes their emissions a peculiar kind of problem: the more infrastructure the world builds to decarbonise, the more of these materials it consumes.

The scale is substantial. Steel production alone accounts for roughly 7 to 9% of global CO₂ emissions, with cement contributing a similar order of magnitude. Together they sit near a seventh of the global total — not as a by-product of energy use that a cleaner grid will eventually fix, but as an intrinsic feature of making the physical world.

They are almost always discussed jointly, under the banner of “hard-to-abate” industry. That grouping is understandable and analytically misleading, because the two sectors face problems with entirely different structures — and therefore entirely different investment characteristics, timelines and probabilities of success.

Two Opposite Problems Wearing One Label

The distinction comes down to where the carbon dioxide comes from, and it is worth stating precisely because everything else follows from it.

In steelmaking, the emissions arise from the energy and the reductant. Coal does two jobs in a blast furnace: it provides heat, and it chemically strips oxygen from iron ore. Both jobs can, in principle, be done by something else — hydrogen as the reductant, electricity as the heat. The chemistry permits a substitute.

In cement, roughly 60% of emissions come from the calcination reaction itself — heating limestone drives off carbon dioxide as a matter of chemistry, converting calcium carbonate to calcium oxide. That CO₂ is released regardless of what fires the kiln. A cement plant running entirely on renewable electricity would still emit the majority of its carbon. There is no substitute reaction, because the reaction is the product.

Steel: The Answer Exists and Costs 20–40% More

The technical pathway for steel is well established. Replace the blast furnace and basic oxygen furnace route with hydrogen-based direct reduced iron feeding an electric arc furnace: hydrogen strips the oxygen from the ore instead of coal, and renewable electricity melts the result. The emissions reduction is dramatic.

Real plants are being built on this basis — conversions and greenfield projects across Sweden, Germany, Spain and elsewhere are in construction or commissioning. The technology is not speculative.

The obstacle is cost. Hydrogen-based steel currently runs 20 to 40% more expensive than the conventional route at prevailing hydrogen prices, and the reason traces directly back to the previous piece in this thread: green hydrogen at $4–6 per kilogram cannot deliver competitive steel. The threshold usually cited for commercial viability is hydrogen below roughly €2 per kilogram. Analysts project cost parity somewhere in the 2035–2040 window on current trajectories — potentially 2033–2035 in jurisdictions with strong carbon pricing.

There is a second, cheaper route that deserves more attention than it gets: the scrap-fed electric arc furnace. It is deployable today, needs no hydrogen at all, and delivers 0.3–0.5 tCO₂ per tonne. Its constraint is scrap — availability, quality and traceability — not technology. In a world that has been accumulating steel in buildings and vehicles for a century, the recycled stream is a genuinely underrated asset, and the pre-processing and sorting technologies that improve scrap quality are a quieter investment theme than the hydrogen headlines.

Cement: Chipping Away at an Irreducible Core

Cement’s strategy is necessarily different. Since the calcination emissions cannot be eliminated by changing energy inputs, the sector attacks the problem from three directions at once — none of which is complete, and all of which are cumulative.

The strategic implication is that cement decarbonisation is a portfolio of partial measures rather than a single conversion. That makes it less capital-intensive per plant than steel’s route change, more incremental, and considerably less likely to reach zero. It also makes cement the anchor customer for carbon capture — and the reason the CCUS piece and this one are structurally linked: without capture, cement has a floor it cannot go below.

The Arithmetic Changed in January

Both sectors have spent a decade with the same problem: a cleaner product that costs more, sold into markets where buyers compete on price. Voluntary green premiums were never going to move commodity volumes. What changes that is policy — and on 1 January 2026 the European Union’s Carbon Border Adjustment Mechanism moved from a reporting exercise to a financial obligation.

Importers of steel, cement, aluminium, fertilisers, hydrogen and electricity into the EU now accumulate carbon costs on every shipment, settled through certificate purchases from February 2027. At an EU carbon price around €75 per tonne, ordinary Portland cement at roughly 0.83 tCO₂/t carries a gross cost near €62 per tonne of product. For steel, the spread between blast-furnace and low-carbon routes is far wider still, so the cost differential scales accordingly.

Two qualifications keep this honest. First, the mechanism reduces rather than eliminates cost gaps: gas-based production in the Middle East and North Africa retains a meaningful cost advantage even after CBAM is applied — roughly 24% against 35% before it. Second, CBAM is a European instrument, and its incidence falls substantially on exporters elsewhere. That is a live fairness argument, and it is the subject of its own piece in the Global South thread.

Reading It Through the Frameworks

Where does policy become the cash flow? Directly and unusually explicitly. Absent a carbon price, low-carbon steel and cement are simply more expensive versions of commodities. With CBAM and the EU emissions trading system, the carbon intensity of a tonne of product becomes a line item in the buyer’s landed cost. The investment case for a conversion project is, quite literally, a forecast of the carbon price.

What kind of asset is this? Not infrastructure in the primer’s sense — these are industrial manufacturers with commodity price exposure, not toll-takers on a protected flow. The relevance to this section is as a supply chain: the cost and carbon intensity of steel and cement flows into every asset built downstream. A rising carbon cost on materials raises the capital cost of the entire build-out described in the Build thread.

Where is the moat? In three places. Scrap access and quality, which lets an electric arc furnace deliver low-carbon steel today with no hydrogen at all. Proximity to cheap hydrogen and clean power, which determines which conversions ever pencil. And supplementary cementitious material supply — slag and calcined clay are locally-sourced, and access to them is a genuine constraint on the cheapest decarbonisation lever in the cement industry.

The Correction, in Numbers

No corner of the energy transition has produced a wider gap between announcement and delivery than hydrogen. For four years it was described as the missing molecule — the answer to aviation, shipping, steel, heating, heavy trucking and seasonal storage simultaneously. The 2025–26 correction has been correspondingly brutal, and it is worth looking at plainly before assessing what survives.

Start with the headline ratio. Of roughly 520 GW of globally announced electrolysis capacity, only about 4 to 7% has reached a final investment decision. The rest is press releases, memoranda of understanding and feasibility studies. The International Energy Agency’s pipeline of announced low-emissions hydrogen has shrunk to around 27 million tonnes of potential 2030 production — and, more tellingly, the volume of projects either committed or with a realistic chance of operating by 2030 fell from roughly 10 million tonnes to just over 6 million in a single year. Since the prior review, only about 300,000 tonnes a year of genuinely new capacity has cleared FID.

The cancellations behind that shrinkage were not marginal projects. Over 33 GW has been cancelled or deferred across a handful of flagship schemes — including a 26 GW Australian renewable hydrogen hub, a 2.88 GW Queensland project, a 2.5 GW Norwegian scheme and a 1.4 GW US development. Close to sixty major clean hydrogen projects were cancelled during 2025 alone.

The most instructive failure was administrative rather than technical. In the European Hydrogen Bank’s second auction, despite a budget of some €1.2 billion, seven winning projects representing 1.88 GW of the 2.33 GW awarded subsequently withdrew. Developers had bid aggressively — as low as €0.20 to €0.48 per kilogram of subsidy — to win support, and then could not secure offtake agreements at the prices those bids implied. They had won the subsidy and lost the customer.

And Yet: There Is Real Steel in the Ground

A fair assessment has to resist the opposite error. The correction is not an extinction, and the sector that emerges from it is smaller but considerably more real than the one that entered.

Global installed electrolysis capacity doubled during 2025 to surpass 4 GW, with more than 2.5 GW under construction and due online through 2026. More than 500 hydrogen projects worldwide have now passed final investment decision, entered construction, or begun operating, backed by upward of $110 billion of committed capital. That is no longer a speculative pipeline; it is a real, if concentrated, industrial base — heavily weighted toward a handful of very large projects rather than spread evenly across markets.

The Cost Gap Widened Instead of Closing

Every forecast made between 2020 and 2022 rested on a single assumption: that electrolyser costs would fall steadily with scale, in the manner of solar panels and lithium cells. That assumption failed. Electrolyser system costs rose by a median of around 57% since 2022, driven by input inflation, supply-chain constraints and the discovery that balance-of-plant costs at industrial scale were considerably higher than pilot economics implied.

The result is a cost gap against incumbents that is wider today than when the hype began.

The threshold usually cited for competitiveness is around $2 per kilogram. Reaching it requires two conditions to hold simultaneously: electricity below roughly €20/MWh, and electrolyser utilisation above about 5,500 hours a year. That combination exists in parts of the Iberian peninsula, Scandinavia and the Middle East and North Africa corridor. It does not exist across most of Europe, and it is precisely why the map of viable projects looks so different from the map of announced ones.

This is also why blue hydrogen keeps winning the decisions that actually get made. Its cost base is anchored to natural gas, a commodity with decades of liquid futures markets behind it, so a lender can hedge and underwrite it. Green hydrogen’s cost base is renewable electricity and electrolyser capex over a twenty-to-thirty-year life — a forecasting problem with no comparable instruments. When a US ammonia complex reached FID and began construction in 2026 with Japanese offtake attached, it was blue ammonia, and bankability was the reason.

The Financing Mechanics That Actually Killed the Projects

For a markets audience this is the section that matters, because the cancellations were not decided by engineers. They were decided in credit committees, and the mechanism is precise enough to generalise to every other first-of-a-kind technology in this section.

A 100 or 200 MW electrolyser is a genuine first-of-a-kind at the site level in most jurisdictions. That means no established lender track record, no proven engineering contractor willing to wrap completion risk at that scale, no operating history from which to model stack degradation, and an equipment supply chain untested at the required volumes. Lenders priced that accordingly: debt for green hydrogen projects has cost more than three times the equivalent for mature renewable energy.

The primer’s point about duration sensitivity applies here with unusual force. Hydrogen projects are long-dated, capital-heavy and front-loaded — exactly the cash-flow profile most punished by a higher discount rate. Hydrogen did not fail a technology test. It failed a discount-rate test, and the same test is being applied to every first-of-a-kind clean-industrial proposal now seeking finance.

The Physics Nobody Put in the Model

Production cost is only half the story, and the neglected half is thermodynamic. Hydrogen is the lightest element in the universe, which makes it extraordinarily awkward to move and store — and every step taken to make it transportable consumes a large fraction of the energy it contains.

• Liquefaction consumes 30–40% of hydrogen’s own energy content — on the order of 10 to 13 kWh per kilogram, spent purely to make it cold enough to ship.
• Conversion to ammonia and back again collapses round-trip efficiency to roughly 11–19%, if the objective is to recover hydrogen at the far end.
• Together, these penalties add something like $2.70–3.20 per kilogram to delivered cost — frequently more than the entire target production price.

Two further physical constraints sit underneath. PEM electrolysers depend on iridium, one of the scarcest elements in commercial use, creating a genuine ceiling on that technology’s deployment rate irrespective of capital availability. And stack degradation over an operating life remains imperfectly characterised, which is precisely the uncertainty that pushes up the cost of debt.

Where It Genuinely Works

The now-standard way to think about this — popularised as a “hydrogen ladder” by the analyst Michael Liebreich — is to rank applications not by how well hydrogen performs, but by how good the alternatives are. Where direct electrification works, it wins decisively on efficiency and cost. Hydrogen earns its place only where nothing else does the job.

The pattern is consistent: hydrogen works where it is a feedstock replacing an identical incumbent feedstock, and struggles wherever it is proposed as an energy carrier competing with electrons. Most of the cancelled capacity was aimed at the second category.

Reading It Through the Frameworks

How does it get paid? This is the sector’s defining weakness. The European auction failure showed developers who had secured a subsidy but no customer — and a subsidy without offtake is not a revenue model. The industrial buyers meant to anchor demand (ammonia producers, refiners, steelmakers) run thin margins and make fuel-switching decisions on total cost of ownership over ten to fifteen year cycles. They are not paying a premium for a molecule that performs identically.

Where is the moat? Not in electrolysers, which are commoditising even as they inflate. It is in geography — sites combining very cheap power with high utilisation are genuinely scarce — and in signed, long-dated offtake, which is the rarest asset in the sector. A twenty-year offtake agreement with a creditworthy industrial buyer is worth more than any technology position.

The Counterweight Argument, Stated Fairly

There is one genuinely serious argument against the infrastructure supercycle, and it deserves to be stated at its strongest before it is qualified. It is that the rich world is ageing, its populations are peaking or shrinking, and the demand assumptions underwriting a decade of construction are therefore too high.

The evidence for it is real. Japan’s electricity consumption has been in structural decline for years. China’s population has passed its peak. And the sharpest data point of all comes from Europe: even with forecast growth of around 2.3% a year to 2030, European Union electricity demand is not expected to return to its 2021 level before 2028. An entire decade will have passed with the bloc consuming less power than it once did — while that same bloc is being asked to fund an unprecedented grid and generation build.

Layer on efficiency, and the case gets stronger still. Mature economies have been decoupling growth from energy use for two decades, so even where population holds, consumption need not rise with it. Anyone underwriting long-duration assets against a demand curve should take this seriously. The mistake is not believing the argument. The mistake is applying it globally.

Where the Argument Stops

The demographic picture that produces the counterweight thesis is an OECD picture. Outside it, the arithmetic runs the other way — and the divergence is now visible in the demand forecasts themselves, not only in the population tables.

In absolute terms the gap is starker than the percentages suggest. India is expected to add more than 570 TWh to annual consumption within five years — having already added nearly 430 TWh between 2021 and 2025. That single increment is comparable to the entire annual consumption of a large European economy, added by one country, in half a decade.

But headcount is the wrong way to hold this, and it is where most commentary goes shallow. Population growth explains only part of it. The larger mechanism is that each person is climbing a consumption ladder from a very low rung, and the rungs above them are ones the rich world reached decades ago.

The Ladder, Not the Headcount

Indian per-capita electricity consumption remains below the global average and a fraction of developed-world levels. That gap is not a deficiency to note in passing; it is the entire investment thesis, because it defines how much demand growth is available without a single additional person being born.

Three forces do the climbing, and they compound rather than substitute. Urbanisation moves people into denser, more electricity-intensive lives. Industrialisation adds load directly — industry is expected to contribute roughly a third of India’s incremental demand as manufacturing, semiconductor and hydrogen programmes scale. And rising household income converts into appliances, one purchase at a time. The household share of Indian electricity consumption rose from 22% in 2012–13 to 25% a decade later, the fastest-growing segment in the system.

Cooling: The Mechanism That Does Most of the Work

Of everything on that ladder, one appliance dominates — and it is the one the temperate world under-weights precisely because it is already saturated there. Air conditioning in India sits at roughly 15% of households, up from 10% in 2020, with forecasts putting it near 35% by 2030. The comparison is instructive: China now averages more than one unit per household, and US penetration is near-universal.

The load implications are extraordinary. India is expected to add 130 to 150 million new room air conditioners between 2025 and 2035. On current efficiency-standard trajectories, room ACs alone could contribute over 180 GW to peak electricity demand by 2035 — close to 30% of the projected national peak. Analysis of household data suggests that if 70% of Indian homes eventually adopt air conditioning, cooling would account for something like 18% of total national electricity demand.

Two features make this demand unusually reliable. First, its geography is inverted in a way that guarantees headroom: several of India’s hottest states have among its lowest penetration — Punjab is near 70% while West Bengal and Bihar sit around 5% — so growth is concentrated exactly where the physical need is greatest. Second, as wet-bulb temperatures rise, cooling shifts from discretionary comfort toward health necessity. Globally, around 3.5 billion people live in high-temperature regions and only about 15% own an air conditioner, with more than 80% of projected cooling electricity demand to 2050 arising in emerging and developing economies.

The 2025 Lesson: This Curve Has Weather In It

A thesis this strong needs its own stress test, and 2025 supplied one. After four consecutive years of growth above 6%, Indian electricity demand rose just 1.4% — despite running at 5.8% through the first four months. An early monsoon arrived in May, bringing cooler temperatures and heavier rain. Cooling degree days fell more than 7% against the prior year, with a 12% drop in June, a month that typically carries around 15% of annual demand. Air conditioners ran less; so did agricultural pumps.

Demand Is Not the Constraint

Here is the turn that separates this from a straightforwardly bullish piece. In the rich world, the risk is that capital gets built into a market whose demand disappoints. In the Global South, demand is the one thing not in doubt. The binding constraints sit entirely on the supply side of capital and contract.

Two dominate. The first is the cost of capital: an identical solar or grid project carries a materially higher financing cost outside the OECD, driven by currency risk, perceived country risk and thin local capital markets. Since infrastructure returns are overwhelmingly set by the discount rate applied to long cash flows, this is not a detail — it is frequently the difference between a project happening and not happening.

The second is offtaker credit. Growing demand only becomes a bankable cash flow if the entity buying the power can reliably pay for it. Where distribution utilities carry structural losses and stretched payables, robust underlying demand does not translate into a financeable contract. A gigawatt of genuine need behind an unbankable buyer is not investable demand.

The Window Is Not Permanent

The final qualification is one that enthusiasts of the demographic story routinely skip. India’s fertility rate has already fallen to around replacement level or below. The population is young today because of births that happened twenty and thirty years ago, not because of births happening now. The demographic dividend therefore has a defined horizon — likely two to three decades — after which India begins ageing too, on a trajectory China has already entered.

This matters for asset duration. An infrastructure asset underwritten on a forty-year life is being underwritten across the end of the window, not merely inside it. The convergence argument holds firmly for that period — per-capita headroom does not close quickly, and much of Africa sits a full generation behind India on the same curve. But the honest statement is that this is a long, powerful, finite trend rather than a permanent condition, and terminal-value assumptions should reflect it.