Civilization Scale Investment

This brief is about capital formation at civilizational scale: who funds a programme whose payback is measured in generations, on what return horizon, through what instrument, and what the historical record of such mobilisations actually shows. The framing under test is that societies can mobilise capital for projects on civilizational timescales.

Two things are deliberately excluded, and naming them is the fastest way to say what is left. It is not about delivery. Whether money, once raised, is spent well belongs entirely to Megaproject Governance, and the relationship is one-directional and compressible to a single sentence: that brief's finding that promised benefits are systematically overstated is a reason to haircut every return assumption on this page, and this brief does not compute the haircut because that one already owns the question. And it is not about whether long-horizon commitments bind. Statutory duties toward future generations, appraisal guidance and whether a sovereign fund's fiscal rule survives the next government belong to Long-Term Institutions. This brief takes sovereign wealth funds as a stock of capital and the discount rate as a pricing parameter; whether either commitment holds is that brief's subject and is cross-referenced rather than restated.

What is left is a real subject with hard numbers, and the answer it returns is not the one the framing expects. The mobilisation question is settled in the affirmative by the record: a 1%-of-GDP programme is below what Marshall aid cost its recipients, an order of magnitude below wartime, and the world already invests about 3% of global GDP in infrastructure against an estimated 3.7% requirement. The gap is under a percentage point of global product.

The binding constraint is elsewhere, and it is a tenor finding rather than a quantity finding. No capital in the world carries an obligation longer than the institution holding it. Pension liabilities are already longer than pension assets and are converting to a form that cannot hold a fifty-year asset; the flagship hundred-year sovereign bond has cost buyers at issuance more than two-thirds of their capital value; and the largest recent industrial-capital programme was legislated in 2022 and substantially repealed in 2025 — a policy tenor shorter than the construction period of the assets it financed. This brief is about instruments longer than any instrument's tenor; the instruments and intermediaries themselves belong to Future Capital Markets. The seam is duration.

Established Start with the denominators, because the two analogies everybody reaches for are quantitatively wrong in the same direction. US defence spending exceeded 37% of GDP in 1945; federal spending reached 46.59% of GDP in 1943 and the deficit 27.5% of GDP in the same year. The total war cost was $304 billion, about $136.8bn from taxes and $167.2bn borrowed, and debt more than doubled to 105% of GDP. On the household side, more than $185 billion of war bonds were bought by individual Americans through 1946 alongside $24 billion by commercial institutions, while the number of income taxpayers rose from 4 million in 1939 to 43 million in 1945 and federal revenue went from roughly 8% to over 20% of GDP. Established An independent central-bank note puts the same period at over 40% of GDP devoted to national defence in 1943–44, with every later war small by comparison.

Established Against that, the Marshall Plan was small and Apollo was smaller. Marshall aid totalled $13.2 billion over 1948–51 — the United Kingdom $3.2bn, France $2.7bn, Italy $1.5bn, West Germany $1.4bn — which is less than 3% a year of recipients' combined national income. Apollo cost $25.8 billion nominal over 1960–73, about $309 billion in 2025 dollars on NASA's aerospace-specific New Start Index, or $28bn / $338bn including Gemini and the robotic lunar programme; NASA has averaged 0.71% of US federal spending since the 1970s and was 0.36% in 2025. Frontier The Apollo inflation adjustment is method-dependent — the New Start Index is a sector deflator and returns a larger figure than a consumer price index would — and the source is a space advocacy organisation drawing on NASA's historical collection, so the arithmetic is citable and the framing is interested.

Established The Interstate Highway System is the useful domestic comparator because it was delivered over thirty-five years without emergency framing. The 1956 congressional estimate was $27 billion; the final 1991 estimate was $128.9 billion total and $114.3 billion federal, against authorisations of $119 billion, for 46,876 miles on a 90% federal / 10% state cost share rising to 95% in public-land states. The source is the delivering agency, so treat the cost framing accordingly. Established And the only contemporary mobilisation at wartime magnitude in peacetime is China's: gross capital formation at 41.13% of GDP in 2023 on World Bank series, sustained for decades. Frontier The largest current proposal is Draghi's €800 billion a year, almost 5% of EU GDP, against a Marshall Plan that was 1–2% of GDP at the time, with the public share rising toward 50% from a traditional 20% and an operative requirement that Europe's cost of capital fall by 2.5 percentage points — which the analysis reporting it judges unlikely given fragmented capital markets.

Established Three conclusions follow, and they are the spine of this brief. First: a 1%-of-GDP programme is historically ordinary, not heroic. It is below the Marshall Plan's burden on recipients, far below Draghi's proposal, and an order of magnitude below wartime. The framing's implicit difficulty is answered in the affirmative for the mobilisation part. Frontier Second: the Marshall Plan analogy is backwards. De Long and Eichengreen's assessment is that the Plan mattered but that the direct capital injection was small — the investment channel contributed about 2% of national product by 1951, which they call hardly the dramatic change its champions trumpeted. What did the work was political economy: the Plan sped Western European growth by altering the environment in which economic policy was made, giving governments a cushion to stabilise, decontrol and negotiate a pro-growth social contract. They call it a large and highly successful structural adjustment programme, not a capital transfer. That is a considered scholarly judgement rather than a measured causal estimate — and anyone invoking the Marshall Plan to argue that a big cheque produces growth is invoking the opposite of what the reference study concluded.

Established Third: the wartime case is not a model, because its financing mechanism does not exist in peacetime. The Second World War was financed by a taxpayer base that expanded elevenfold in headcount, a top marginal rate of 94%, and a household bond drive that mobilised more than $185 billion of retail savings under total mobilisation. Every one of those instruments was available because of the war. A peacetime programme that proposes the magnitude without the mechanism is proposing an outcome, not a plan.

Established Now the tenor problem, which is where the framing actually fails, and it fails on arithmetic rather than on will. Pension and insurance liabilities are the only naturally long capital in existence, and they are decades rather than centuries. The IMF puts global pension savings at $63.1 trillion at end-2023, about 98% of OECD members' combined GDP, with over 40% of advanced-country pension assets still in defined benefit and the defined-benefit share falling sharply — US defined benefit from above 40% to below 30%. It states flatly that the duration of liabilities is longer than that of assets: pension funds are structurally short duration relative to what they owe. Illiquid assets are 22% of defined-benefit assets in EU countries, up from 17% in 2021, ranging 2–27% across member states.

Established And the world's largest long-term savings pool is converting into the form that cannot hold a long asset. The Thinking Ahead Institute's 2026 study puts the twenty-two largest pension markets at $68,274 billion at end-2025, up 9.6%, with the seven largest at $61,975bn (91%), an allocation of 48% equities, 31% bonds, 19% other and 3% cash, pension assets at 74% of GDP in those economies — and defined contribution at 63% of the seven largest markets' assets, growing 9.4% a year over a decade against defined benefit's 3.2%. The institute is affiliated with an investment consultancy, so the framing is mildly interested; the totals are the standard reference. Frontier The shift is the tenor story. A defined-benefit scheme has a liability with a duration measured in decades and an institutional obligation to match it; a defined-contribution account has a member with a retirement date and a daily-priced balance. The mechanism is established; the claim that this materially reduces aggregate long-horizon investment capacity is not directly measured, and this brief does not assert it as measured.

Established The nominally longest instrument in existence has behaved catastrophically, and this is the practical refutation of the “just issue century bonds” proposal. Austria's bond maturing in 2120, issued in mid-2020, has cost investors who bought at issuance more than two-thirds of their capital value, with the coupon doing little to change that conclusion. Frontier The mechanism is duration arithmetic and it generalises: a hundred-year instrument at a low coupon has a duration so extreme that a two-point move in rates destroys most of the principal value. Which means no institution with a mark-to-market constraint or a solvency ratio can hold it in size. The instrument exists; the holder does not.

Established Sovereign wealth funds are large, nominally permanent, and doing something else. Global SWF's sixth annual review, dated 1 January 2026, puts total state-owned investors at US$60 trillion across roughly 800 funds — sovereign funds, public pension funds and central banks — with sovereign wealth funds alone at US$15.2 trillion, a historic high, MENA funds at US$6 trillion and the top fifteen holding 81% of total assets. In 2025 sovereign funds invested $180.3 billion across 324 deals and public pension funds $97.8 billion across 238, at an average ticket of $0.49 billion, with $132 billion — 47% — going to the United States; Norway's fund became the first past US$2 trillion. The publisher is a commercial data platform serving this industry, so it is interested in the sector's importance; the totals are the standard reference. Frontier Two things follow. $15.2 trillion is enough — a 1%-of-global-GDP annual programme is roughly $1.1 trillion and sovereign funds deploy about a sixth of that into deals each year. And an average ticket of half a billion with 47% of capital going to one country describes an asset-management operation, not a civilizational-investment vehicle. The mandate that would be required — accept a negative expected return over the holder's own horizon in exchange for a positive return to a successor society — is not the mandate any of these funds has.

Established The framing says societies can mobilise. The recent record says they can announce. The EU's Recovery and Resilience Facility, at €650 billion, was 58% disbursed — about €377bn — with nearly €270bn unspent against an end-2026 deadline; Greece, Croatia, Italy and Portugal absorbed best, and Spain declined €67 billion of its €83 billion in loans. The projected growth impact was +0.4% a year over 2020–2030, which the analysis expects to come in lower because of absorption delays, attributing the shortfall to administrative bottlenecks, limited implementation capacity and changing political contexts. The reporting party is a trade credit insurer, interested in European corporate conditions though not obviously in this finding.

Established And the largest recent industrial-capital programme reversed within three years of enactment. An advocacy organisation's 2025 tally records $34.8 billion of announced clean-energy investment cancelled and 38,031 jobs lost against $12.3 billion of new investment announced — roughly three dollars abandoned for every dollar announced, and the first year since 2022 in which departing investment exceeded new. Manufacturing reversals accounted for $30.2bn, with electric-vehicle and battery sectors each losing over $21bn. Note the direction of interest: this is a clean-energy advocacy organisation reporting that clean energy is contracting, which raises the weight substantially. Established The policy cause is documented and dated. A 2025 reconciliation act repealed or curtailed essentially the entire consumer clean-energy credit suite — the 30D, 25E and 45W vehicle credits after 30 September 2025, 25C and 25D after 31 December 2025, 30C, 45L and 179D after 30 June 2026 — and restricted 45Y and 48E so that wind and solar are ineligible unless placed in service before 31 December 2027 or under construction within twelve months, with the 45V hydrogen credit repealed for construction beginning after 31 December 2027. The ten-year scored revenue effect is $484.49 billion: $267.13bn consumer, $257.27bn business, less $39.91bn of expansions. The reporting organisation's standing view favours lower and simpler taxes, so the statutory dates and the score are the citable content and the framing is not. Frontier The lesson for capital formation is specific and unflattering: a capital programme whose returns accrue over twenty to thirty years was legislated in 2022 and substantially repealed in 2025, inside a single investment cycle and shorter than the construction period of the assets it was meant to finance.

Frontier The same pattern appears in the largest overseas lending programme in history. AidData's reconstruction documents $1.34 trillion across 20,985 projects in 165 low- and middle-income countries over 2000–2021, with outstanding debt of at least $1.1 trillion and possibly $1.5 trillion, annual commitments now around $80 billion, infrastructure lending down from 65% of annual commitments in 2014 to 31% in 2021 while emergency rescue lending rose from 13% to 58%, and an estimate that 80% of the overseas lending portfolio is currently supporting countries in financial distress. These are estimates from a reconstructed dataset rather than official statistics. Within seven years, the dominant activity of the largest cross-border capital mobilisation in the modern record became refinancing its own earlier loans.

Frontier Finally, the requirement, from three independent estimates, and it is smaller than the rhetoric. The Global Infrastructure Outlook, commissioned to a commercial forecaster under a G20 initiative, gives $94 trillion of investment need to 2040 against $79 trillion forecast on current trends — a $15 trillion gap; adding UN access targets gives $97 trillion of need and an $18 trillion gap, 19% of forecast need. Current spending is about 3% of global GDP against 3.7% required. The definition is the part nobody quotes: “need” here means matching best-performing peers, which is a benchmark choice rather than a physical requirement. Established On energy, the IEA puts 2025 global investment at $3.3 trillion — $2.2tn clean and $1.1tn fossil — with solar photovoltaics at $450bn, grids at $400bn a year, nuclear at about $75bn and battery storage at $65bn; its stated gap is structural rather than aggregate, namely that grid investment must reach parity with generation spending by the early 2030s to maintain electricity security and is not on track, and that Africa receives 2% of global clean energy investment with 20% of world population. Established Put together: the world already invests $3.3 trillion a year in energy alone and about 3% of GDP in infrastructure, and the gap against the most-quoted requirement is under a percentage point of global product. Against the historical magnitudes above, that is not a mobilisation problem. It is an allocation-and-durability problem.

Frontier The binding conceptual question is what rate prices a two-century benefit, and the field is honestly split. Nordhaus's objection to the Stern Review is that its radical revision arises from an extreme assumption about discounting and that it proposes a social discount rate that is essentially zero — a near-zero pure rate of time preference of 0.1% against his own preference for a rate starting at 3% and declining to about 1% over 300 years, calibrated to observed market interest rates, savings behaviour and capital returns. His reductio: a minor climatic problem causing 0.01% output losses beginning in 2200 and continuing indefinitely would, under the Review's test, justify a payment of 15 per cent of world consumption today — approximately $7 trillion — which he calls completely absurd.

Established The expert distribution is tighter than the dispute suggests, and it has been measured. A survey of 197 experts, 185 giving quantitative responses, returns a mean recommended long-term social discount rate of 2.27%, a median of 2% and a range of 0–10%; a median pure time preference of 0.5% with a mode of zero; and a median and modal elasticity of marginal utility of 1.0. 77% find 2% acceptable, 92% are comfortable with a 1–3% range, 58% include Stern's 1.4% in their acceptable range and 31% include Nordhaus's 4.5%. The authors conclude that the simple Ramsey rule cannot explain the majority of their experts' responses. Frontier Stated in capital terms, which is this brief's use of it: at 2%, a benefit arriving in 200 years is worth about 1.9% of its face value today; at 4.5%, about 0.016%. The difference between the mainstream expert median and the Nordhaus position is a factor of roughly a hundred in whether a two-century project is worth building. Established And no financial instrument resolves it, because the discount rate is not a market price for a two-century horizon — no such market exists. The longest actually traded instrument is a hundred-year sovereign bond that has lost two-thirds of its value. The rate used to appraise a civilization-scale project is therefore an assumption, not an observation, and the profession's own acceptable range spans two orders of magnitude in present value. Whether a government's appraisal guidance should adopt one is a separate question owned by Long-Term Institutions.

Handwave “The capital exists; the constraint is political will.” The mainstream advocacy position across climate, space and infrastructure. The arithmetic supporting it is real — historical mobilisations far exceeded 1% of GDP and the current gap is under a percentage point. Frontier What is not real is “will” as an explanatory variable: nothing would settle it, because nothing measures it, which is precisely why the hypothesis is so durable. The evidence against is that will, once exercised, did not persist through one electoral cycle in the largest recent case.

Frontier “The constraint is tenor, not quantity.” This brief's leading hypothesis. Evidence: pension liabilities longer than pension assets, defined contribution at 63% and rising, a century bond down two-thirds, and a sovereign-fund average ticket of half a billion dollars. What would settle it in either direction is concrete: an instrument with a century-plus tenor achieving institutional take-up at scale would falsify it, and another decade of failed ultra-long issuance would confirm it.

Frontier “The constraint is the discount rate, and the discount rate is an ethical choice masquerading as a parameter.” Held by Stern, the dismal-theorem literature and much of intergenerational ethics. Evidence: a modal pure time preference of zero among surveyed experts while government guidance uses positive rates, and a two-order-of-magnitude spread in the present value of a two-century benefit across the profession's own acceptable range. Against it: the argument that a rate uncalibrated to observed capital returns produces absurd prescriptions, with the $7 trillion reductio. Frontier Nothing empirical would settle this. It is a normative disagreement with an empirical costume, and saying so is the honest position.

Speculative “Sovereign wealth funds are the natural vehicle and merely need a mandate change.” Held in parts of the sovereign-fund and development-finance community. For: $15.2 trillion, nominally permanent horizons, existing governance. Against: 47% of 2025 deployment went to the United States at an average ticket of $0.49bn — these are return-seeking allocators — and the funds that would have to change mandate are largely owned by states whose fiscal position depends on the returns. What would settle it: a single major fund adopting and holding a below-market-return civilizational mandate for a decade. None has.

Frontier “Only war-equivalent mobilisation works, and only under war-equivalent threat perception.” A minority position with the best historical support of any hypothesis here: the one unambiguous 40%-of-GDP mobilisation in the record had total war attached, an elevenfold expansion of the taxpayer base, a 94% top rate and a $185 billion retail bond drive. Against it are two direct counterexamples — China's sustained 41% gross capital formation in peacetime, and an interstate system delivered over thirty-five years without emergency framing. What would settle it: a peacetime democracy sustaining above 10% of GDP on a single programme for a decade. None has. Frontier The adjacent hypothesis is that China is the existence proof and the mechanism is state-directed credit rather than markets. Against that: the overseas arm of the same model has 80% of its portfolio in distressed countries and has shifted from infrastructure at 65% of commitments to 31% while rescue lending went from 13% to 58% — the mechanism has demonstrated a failure mode at the frontier of its reach. What would settle it is the return on the domestic build, which is not independently measurable.

Frontier “Green industrial policy is a durable new mobilisation vehicle.” This hypothesis has taken the sharpest single hit of any here within the last eighteen months. For: $2.2 trillion a year of clean energy investment. Against: $34.8bn of cancellations in one year against $12.3bn of announcements, and statutory repeal of essentially an entire consumer credit suite three years after enactment. What would settle it: whether the European and Chinese programmes survive their own political cycles. Speculative “Discounting is the wrong instrument entirely, and the profession is quietly abandoning it.” A minority but rising position whose capital-formation implication is radical: if civilization-scale projects are not to be appraised by discounting, there is no price at which such projects clear, and allocation must be by political decision rather than by return. Its institutional trace is real and is recorded in Long-Term Institutions rather than here.

Speculative “Intergenerational instruments can be designed.” Perpetual instruments with consumption-linked rather than rate-linked coupons; sovereign contingent claims indexed to the outcome the project is meant to produce; multi-generational trusts with constitutionally entrenched mandates. Held by a small design-oriented literature and some sovereign-debt reformers. The evidence base is essentially empty — GDP-linked warrants exist in restructurings and inflation-linked perpetuals have historical precedents, but nothing carrying a century-plus obligation has been tested. What would settle it: issuance, and then a shock. Frontier And the strongest deflationary hypothesis has to be stated honestly because it partly dissolves this slot: that there is no capital-formation problem at all, only a project-selection problem, in which case the same capital would deliver materially more under honest appraisal and no new financing instrument is required. That claim belongs to Megaproject Governance, which makes it and flags it as testable. If it is right, this brief's subject is a second-order problem.

Established The first bottleneck is duration arithmetic, and it is not negotiable. A bond's price sensitivity to rates rises with the time-weighting of its cash flows, so a low-coupon hundred-year instrument concentrates almost all of its value in a payment nobody alive will receive. The Austrian 2120 issue lost more than two-thirds of its capital value for issue-date buyers on a rate move that would have cost a ten-year holder a small fraction of that. Frontier The consequence is a buyer problem rather than an issuer problem. Any institution subject to a solvency ratio, a funding-level test or a daily mark cannot hold that volatility in size, so the natural buyer of the longest instrument is precisely the institution least able to own it. Issuance is easy; placement at scale is not.

Established The second is that the only naturally long liability is shortening. Global pension savings of $63.1 trillion sit against liabilities the IMF says are already longer than the assets backing them, and the pool is converting at roughly six percentage points a decade from defined benefit — which has an institutional obligation to match a decades-long liability — into defined contribution, which has a member, a retirement date and a daily-priced balance. Frontier Illiquid assets at 22% of EU defined-benefit portfolios show the appetite exists where the structure permits it, and the 2–27% spread across member states shows the structure is what varies.

Established The third is that the appraisal parameter has no market to read it off. There is no traded instrument at a two-century horizon, so the discount rate applied to a civilization-scale project is an assumption sourced from a professional distribution whose acceptable range implies present values a hundredfold apart. Frontier That is not a resolvable measurement problem — more data on hundred-year bonds would price a hundred-year bond, not a two-hundred-year benefit — and it means the go/no-go decision on this class of project is made by parameter choice before any engineering is examined.

Established The fourth is that the most-quoted requirement figure is a benchmark, not a physical need. The $15 trillion infrastructure gap is defined as the shortfall against matching best-performing peers, on a top-down econometric panel with two counterfactuals. That definition is almost never stated when the number is quoted. Frontier It matters because a benchmark gap is elastic to the choice of peer and a physical gap is not, and policy arguments that treat the two as the same thing are borrowing the authority of engineering for a statistical construction.

Frontier And the fifth is absorption. Nearly €270 billion of a €650 billion facility remained unspent against a hard deadline, with one member state declining €67 billion of its €83 billion loan allocation, attributed to administrative bottlenecks and limited implementation capacity. Capital that cannot be absorbed is not capital that has been mobilised, and the announced-to-disbursed ratio is a measurable quantity that almost nobody reports.

Established Nothing on this map produces a result this brief waits on, and that is the honest adjudication rather than a gap. The constraints here are an instrument that has not been issued, a mandate that has not been adopted and a statutory tenor that has not been written — things a treasury, a fund owner or a legislature could choose to supply. All three are recorded as typed requirements below, and none is a discovery.

Established Two seams run to briefs that already exist and are cross-referenced rather than restated. Megaproject Governance owns whether the money is spent well, and its bearing on this brief is the single sentence in section 1. Long-Term Institutions owns whether a long-horizon commitment binds — a statutory duty, an appraisal rate, a fund's fiscal rule — while this brief uses the same objects as a stock of capital and a pricing parameter. Frontier The distinction is not pedantic: a fund that is $15.2 trillion of investable assets and a fund that is a commitment device are the same institution described by two different questions, and merging them produces the common error of treating an asset total as evidence of a mandate.

Frontier What this brief waits on from research is narrow and mostly descriptive. A published series of announced-versus-disbursed ratios across large capital programmes, which nobody maintains; a duration profile of global institutional liabilities that separates contractual obligation from investment horizon; and a repeat of the expert discount-rate survey with the same instrument, so that the distribution can be tracked rather than cited once. Speculative None of those changes the tenor finding; each would let it be stated with a base rate instead of with cases.

Frontier The decisive experiment is an issuance, and it is available to any large sovereign that wants it. Issue an ultra-long instrument whose coupon is linked to consumption or output rather than to a nominal rate, and publish the order book. The tenor hypothesis predicts a placement failure at size among solvency-constrained institutions and take-up only from unconstrained holders; the design literature predicts the opposite. Either way the result is data, and the Austrian 2120 issue is the control.

Speculative Second: one major sovereign fund adopts a published below-market-return mandate for a defined share of assets and holds it for a decade. That is the only test that separates “permanent horizon” from “total-return allocator with a long name”. The measurable outputs are ticket size, geographic distribution and realised return against the fund's own benchmark. None of the roughly 800 state-owned investors has run it.

Established Third, and the cheapest: publish the announced-to-disbursed ratio for every large capital programme, prospectively. The European facility's 58% at a fixed deadline, the 3:1 cancellation-to-announcement ratio in one country's clean-energy tally, and the shift of an overseas lending programme from 65% infrastructure to 58% rescue lending are three data points that happen to have been reported. A base rate would convert a set of anecdotes into the most useful number in this subject, and every input is already in public accounts.

Frontier Fourth: track the statutory tenor of capital programmes against the physical tenor of the assets they finance. The 2022-to-2025 repeal cycle is one observation; the record contains many more, and the comparison is mechanical — enactment date, first repeal or curtailment date, and the construction period of the financed asset class. Speculative If the median policy tenor is shorter than the median asset tenor across a large sample, the finding in section 2 stops being a case study and becomes a design constraint that any programme has to be built against.

Speculative Fifth, and the one nobody will run: a matched comparison of appraisal outcomes under discounting against appraisal under published undiscounted time profiles. The decision rule differs by a factor of a hundred at two centuries, so the two methods will select different portfolios from the same candidate set. Running both on the same pipeline and recording where they diverge would price the parameter choice in projects rather than in percentages, which is the form in which a decision-maker can actually see it.

Established The financing mechanics of the historical cases are more instructive than their headline magnitudes, because the mechanics are what does not transfer. The wartime case ran on three simultaneous instruments: a tax base that went from 4 million to 43 million filers, a top marginal rate of 94%, and a retail bond drive of more than $185 billion from individuals plus $24 billion from commercial institutions, against a total war cost of $304 billion split roughly $136.8bn taxes and $167.2bn borrowing. A peacetime programme has access to none of the three at those magnitudes.

Established The Interstate case is the transferable one and its engineering is a cost-share formula. A 90% federal / 10% state split, rising to 95% in public-land states, sustained across thirty-five years and seven administrations, delivered 46,876 miles for a final estimate of $128.9 billion against a 1956 vote taken on $27 billion. Frontier What made it durable was not the magnitude but the funding mechanism's independence from annual appropriation, and that is a design feature a modern programme can copy without copying the emergency framing.

Frontier The Marshall design is a structural adjustment programme, and reading it as a financing template inverts it. The capital was small — under 3% a year of recipients' national income, with the investment channel contributing about 2% of national product by 1951 — and the reference assessment attributes the effect to the policy environment it bought rather than to the money. A programme copying the cheque without the conditionality is copying the part its own historians say did not do the work.

Speculative The instrument that would actually close the gap does not exist, and its specification is short. It must be large — 1% of global GDP is about $1.1 trillion a year; tolerant of a payback beyond any holder's horizon; insulated from a policy reversal inside the construction period; and priced at a rate the profession cannot agree on within two orders of magnitude. What exists instead is $15.2 trillion of sovereign fund assets on total-return mandates, $68.3 trillion of pension assets whose liabilities already exceed their asset duration and which are converting to defined contribution, a century bond that has lost two-thirds of its value, and a legislative record showing a twenty-year capital programme repealed in three. Speculative The gap is not money. It is a claim on a future society, held by a present institution, that neither can revoke. Sovereign debt is revocable by default or inflation; equity carries no horizon obligation; a statutory duty is repealable; and a fund's mandate is amendable by its owner. This is the most speculative paragraph on this page and is flagged accordingly: it is a framing claim, not a measured result.

Speculative Three candidate designs have been proposed and none has been tested at scale. Perpetual instruments with consumption-linked rather than rate-linked coupons, which would remove the duration problem by removing the fixed nominal claim. Sovereign contingent claims indexed to the outcome the project is meant to produce, which align the holder's return with the programme's purpose. And multi-generational trusts with constitutionally entrenched mandates, which attempt to make the obligation survive the issuer. Handwave Each has an internal logic and none has an evidence base; the nearest precedents — GDP-linked warrants in sovereign restructurings, historical inflation-linked perpetuals — are orders of magnitude short in both size and tenor.

Within this map: Megaproject Governance, which owns delivery and whose bearing here is one sentence; Long-Term Institutions, which owns whether any of these commitments bind; Future Capital Markets, which owns the instruments and intermediaries up to the point where tenor runs out; Future Taxation Models, which owns the revenue side of any programme financed by tax rather than by borrowing; Civilizational Planning, which owns what such a programme would be for; and Economic Resilience, where the cost of buying insurance against shocks is priced.

Outside it: public finance and sovereign debt management, which supply the duration arithmetic; economic history, which supplies the mobilisation record; the intergenerational-ethics literature, which supplies the discounting dispute; development finance, which supplies the overseas lending record; and pension actuarial practice, which is where the liability-duration measurement actually lives.

Established Almost every number on this page comes from a party with an interest, and the direction of that interest is load-bearing. Running against its own interest and therefore weighted up: a clean-energy advocacy organisation reporting that clean energy contracted three dollars for every one announced. Running with its interest and therefore cited for figures rather than framing: a space advocacy organisation on Apollo's cost, the delivering agency on the Interstate's, a commercial data platform on sovereign fund totals, an investment-consultancy affiliate on pension assets, a trade credit insurer on European absorption, and an organisation that advocates lower taxes on the statutory repeal dates.

Frontier The institutional requirement that would matter most is a statutory tenor that matches an asset tenor. A credit or subsidy whose repeal requires more than an ordinary majority, or whose withdrawal triggers compensation to sunk investors, would convert a policy commitment into something a lender can price. No large jurisdiction has done this for a capital programme, and the 2022-to-2025 cycle is what its absence costs. Speculative Whether such an entrenchment is desirable is a genuine question — a legislature that cannot repeal a bad subsidy is also a legislature that cannot repeal a bad subsidy — and this brief does not resolve it.

Established The second is a mandate, and the institutions that would have to write it are identifiable. Roughly 800 state-owned investors hold $60 trillion, of which $15.2 trillion sits in sovereign wealth funds and 81% of that in fifteen of them. A civilizational mandate is therefore not a diffuse political achievement; it is fifteen decisions. Frontier That the concentration makes it tractable and that none of the fifteen has made it are both facts, and the second is better evidence than the first is an argument.

Frontier The third is absorption capacity, which is an administrative rather than a financial constraint. A facility 58% disbursed against a fixed deadline, with a member state declining four-fifths of its loan allocation, is a system whose limiting factor is the ability to write and supervise contracts. Capital-formation policy is usually argued as if the binding constraint were the raising and never the spending rate, and the one recent case with published figures says otherwise.

Established The distributional facts here are sharper than the intergenerational ones, and they are measured. Africa receives 2% of global clean energy investment with 20% of world population. The largest cross-border capital programme in the record has 80% of its portfolio supporting countries in financial distress, having shifted from infrastructure at 65% of commitments to rescue lending at 58%. Whatever civilization-scale investment has meant so far, it has not meant capital flowing to where the marginal person is.

Frontier The intergenerational question is decided by a parameter, and the parameter is an ethical position expressed as arithmetic. A modal pure time preference of zero among surveyed experts is a statement that future people count equally; a rate calibrated to observed capital returns is a statement that they count as the market counts them. Between the two lies a factor of a hundred in the present value of a two-century benefit, and that factor decides whether a project is built. Handwave Presenting either position as the technically correct one, rather than as a normative choice with a technical implementation, is where most of the dishonesty in this subject lives.

Established And the wartime precedent carries an ethical warning that the mobilisation enthusiasts rarely quote. The financing that achieved 40% of GDP required an elevenfold expansion of the taxpayer base, a 94% top marginal rate and a retail bond drive conducted under total mobilisation. Those are the instruments of a society that has suspended ordinary politics. Speculative An argument that a peacetime programme should reach wartime magnitudes is, if taken at its own arithmetic, an argument for wartime instruments, and the people making it seldom say so.

Frontier The last ethical fact is a lock-in one and it runs the other way. An instrument designed so that a future society cannot revoke it is an instrument designed to bind people who cannot consent. The tenor problem and the legitimacy problem are the same problem seen from opposite ends: everything that would make civilization-scale capital durable also makes it undemocratic, and no design in the literature has escaped that trade-off.

Established The terminal position is that societies mobilised capital at civilizational scale repeatedly, and none of the flagship cases worked the way the framing assumes. The Marshall Plan, the canonical proof, delivered under 3% of recipients' national income a year and contributed about 2% of national product through the investment channel by 1951, with its own historians concluding it worked by changing the policy environment. Apollo was 0.7% of federal spending on average and is now 0.36%. The Interstate took thirty-five years. And the most recent attempt reversed inside three, with $34.8 billion of cancellations against $12.3 billion of announcements in a single year, reported by an organisation founded to advocate for the sector.

Established So the framing survives in a changed form. Societies can mobilise, and then stop. The specific and tractable version of the finding is that the tenor of the political commitment is systematically shorter than the tenor of the asset — which is a design statement rather than a lament about will, and it points at instruments rather than at exhortation.

Frontier The long-run consequence is that a whole class of project is unbuildable for reasons that have nothing to do with engineering or with money. A programme with a two-century payback cannot be priced, because no market prices that horizon; cannot be held, because no institution's constraints permit the duration; and cannot be relied upon, because the statute funding it is repealable by ordinary majority. Each of those is separately fixable and none has been fixed.

Speculative The most interesting scenario is the deflationary one. If the world already invests $3.3 trillion a year in energy and about 3% of GDP in infrastructure, and if the honest appraisal of what that capital delivers is materially worse than what is promised, then the binding constraint on civilizational outcomes is selection rather than formation — and the same money, honestly appraised, would build more. That claim belongs to another brief on this map and this one records it rather than arguing it. Handwave Which of the two constraints binds harder is not something either brief can currently measure, and the fashionable answer changes with the decade.

These horizons track issuance windows, statutory deadlines and fund mandates rather than technology:

• 10 yr: Established The dated items are hard. The European facility's end-2026 deadline arrives with roughly €270 billion at stake, and the disbursed share at that date is the cleanest available reading of absorption capacity. Frontier The 2027 construction-start and in-service cut-offs in the repealed credit suite become outturns rather than forecasts, and the cancellation-to-announcement ratio for that programme becomes a completed series. Frontier Expect at least one further ultra-long sovereign issue and expect it to be small; the placement size, not the coupon, is the datum to watch.
• 25 yr: Speculative On current growth rates defined contribution passes three-quarters of the largest pension markets' assets, which would make the tenor finding structural rather than directional. Speculative Either a major sovereign fund adopts a published sub-market civilizational mandate, in which case there is finally something to evaluate, or none does and the hypothesis stays untested for a second generation. Frontier The infrastructure gap as currently defined closes or does not on a benchmark that will itself have moved, which is a reason to watch spending as a share of GDP rather than the gap figure.
• 50 yr: Speculative If nothing changes in instrument design, the class of projects with payback beyond fifty years continues to be funded only where a state chooses to fund it directly out of current revenue — which is the arrangement that exists now and is what the record describes. Speculative The alternative is an entrenched instrument, and the only known routes to entrenchment are constitutional or supranational. Handwave Both are political outcomes and any confident statement about which arrives is assertion.
• 100 / 250+ yr: Handwave Beyond useful forecasting, and the honest content at this horizon is a single observation: the longest-dated instrument anyone has actually issued matures in 2120 and has already cost its first buyers two-thirds of their capital. Handwave One issue, six years old, is not a base rate for a century.

• Depends on Nothing on this map. This brief waits on no result another brief produces: what it lacks is an instrument, a mandate and a statutory tenor, none of which is a discovery. No typed depends-on edge is claimed. Two seams run to briefs that exist and are cross-referenced rather than restated — Megaproject Governance on whether the money is well spent, and Long-Term Institutions on whether the commitment binds.
• Requires (not on this map) The central constraint here is duration-shaped, and the token says so: a claim whose tenor outlives the institution that issued it. Nothing of the kind has been built. Sovereign debt is revocable by default or inflation, equity carries no horizon obligation, a statutory duty is repealable by ordinary majority, and a fund's mandate is amendable by its owner — so a programme with a two-century payback has no counterparty that can promise anything for two centuries. The second token is the buyer side of the same fact: the flagship hundred-year sovereign bond, Austria's 2120 issue, has cost issue-date buyers more than two-thirds of their capital value, because a low-coupon century instrument has a duration so extreme that a two-point rate move destroys most of the principal — which means every institution with a solvency ratio, a funding-level test or a daily mark is disqualified from holding it in size, and the natural buyer of the longest instrument is the one least able to own it. The instrument exists; the holder does not. The third is fiscal machinery rather than market structure: a statutory tenor that matches the asset tenor it finances. A capital programme whose returns accrue over twenty to thirty years was legislated in 2022 and substantially repealed in 2025, with a ten-year scored revenue effect of $484.49 billion reversed inside a single investment cycle and shorter than the construction period of the assets it funded; the measured consequence in the following year was $34.8 billion of cancellations against $12.3 billion of announcements. None of the three is a research result. Each is a choice a treasury, a fund owner or a legislature could make, and none has been made.
• Enables Every programme on this map whose payback is measured in generations inherits this constraint: planetary-scale energy systems, multi-century waste isolation, terraforming, and any climate intervention whose benefits accrue after its sponsors are dead. No typed enabling edge is claimed, because what would enable them is an instrument that has never been issued rather than a result this brief could produce.
• Adjacent Public finance and sovereign debt management; economic history, which supplies the mobilisation record; intergenerational ethics, which supplies the discounting dispute; development finance; pension actuarial practice; and within this map Future Capital Markets, Future Taxation Models and Civilizational Planning.

Established “A Marshall Plan for X.” The most common rhetorical move in this subject and it invokes the opposite of what the reference study found. Marshall aid was under 3% a year of recipients' national income and its investment channel contributed about 2% of national product by 1951; its own historians conclude it worked as a structural adjustment programme that changed the policy environment. Citing it for the proposition that a large cheque produces growth cites the part its assessors specifically said did not do the work.

Established “An Apollo programme for X.” Apollo cost $25.8 billion nominal, roughly $309 billion in 2025 dollars, against a space agency that has averaged 0.71% of federal spending since the 1970s and is at 0.36% today. Frontier The analogy is usually deployed to signal enormous national effort and it denotes something under one per cent of a budget — which is, as it happens, the actual scale of the infrastructure gap, so the analogy is accidentally apt and rhetorically backwards.

Established “Just issue century bonds.” The flagship century issue has cost buyers at issuance more than two-thirds of their capital value. Frontier The failure is structural rather than bad luck: duration at that tenor is so extreme that no solvency-constrained institution can hold the instrument in size, which is why the market for it is thin at any coupon. Issuing is not the hard part.

Frontier “Sovereign wealth funds are civilization-scale investment vehicles.” Their deployment pattern says otherwise: an average ticket of $0.49 billion and 47% of 2025 capital into the United States. Established $15.2 trillion is a stock of capital, not a mandate, and the frequent slide from the first to the second is the single most common error in this field. Whether such a mandate would bind if adopted is a different question and belongs to Long-Term Institutions.

Frontier “There is a $15 trillion infrastructure gap.” There is a $15 trillion shortfall against matching best-performing peers, computed on a top-down econometric panel by a commercial forecaster under a G20 initiative. Established That definition is almost never stated when the figure is quoted, and it is not a physical requirement. The related and better-behaved statement is that current infrastructure spending is about 3% of global GDP against 3.7% on the same method — a gap under a percentage point.

Handwave “A currency issuer faces no financing constraint, only a real-resource constraint, so the funding question is vacuous.” This circulates widely and it is worth covering because its premise is genuinely underappreciated: the wartime record is partly consistent with it, since the United States borrowed $167 billion and monetised a share of it. Frontier The real-resource point is defensible. What is not defensible is the extension — that inflation and exchange-rate constraints can be managed away — and nothing would settle the dispute cleanly, because proponents attribute every disputed case to policy error. The strong version does its work by assertion at exactly that step.

Handwave “The capital exists; what is missing is political will.” The first half is true and measured: $15.2 trillion in sovereign funds, $68.3 trillion in pensions, $3.3 trillion a year in energy investment, and a sub-1%-of-GDP gap. Established The second half is not an explanation, because “will” has no measurement and no falsification condition — which is why the claim survives every disconfirming episode. The specific version that is testable is that policy tenor is shorter than asset tenor, and that one has a case series behind it.

Established “The problem is that nobody has ever mobilised at this scale.” Societies mobilised at four times this scale within living memory, and one country is doing something comparable now at 41.13% gross capital formation. Frontier The record refutes the scarcity framing and replaces it with a harder one: the mobilisations that worked either had a war attached, or ran for thirty-five years on a funding mechanism insulated from annual appropriation, and no modern proposal has either.

Speculative And the claim this brief itself makes should carry the same scepticism it applies to others. That the missing thing is “a claim on a future society, held by a present institution, that neither can revoke” is a framing, not a finding. It organises the evidence well and it has not been tested, because the instrument it describes has never been issued. Handwave If a consumption-linked perpetual or an entrenched multi-generational trust were placed at scale and failed for some other reason entirely, this brief's central claim would be the thing that had been refuted.