Infrastructure Misallocation

The asset takes three years. The infrastructure it depends on takes ten.

Ontario Teachers’ Pension Plan has argued that the physical economy — generation, grids, storage, resilience — is where the next decade of infrastructure capital belongs. The thesis is right, and this firm shares it. This paper tests it against the live record to the end of October 2026, across Europe, North America, Australia and Southeast Asia, and reaches a sharper conclusion: the problem is no longer a power problem, and it is no longer a data-centre problem.

In eight sectors on four continents the same structure appears. Capital commits to the production asset on a one-to-five-year clock. The enabling layer that asset cannot operate without — transmission, firming, water supply, wastewater, ports, installation vessels — runs on a five-to-fifteen-year clock. The asset is financed, built and commissioned into an ecosystem that is not ready to receive it. That is not a delay. It is a misallocation, and it is priced as though it were a delay.

• The mismatch is structural, not cyclical. A semiconductor fab is built in about three years; the dedicated water scheme that supplies it takes seven to ten or more, and the transmission line to the same site takes about ten. No amount of capital compresses the second number, because the binding constraint is consent, land and public process, not money.
• Announcement is not capacity. Announced hydrogen pipelines worldwide exceed 40,000 km by 2035. Nine per cent of that length is operational or has committed investment. Germany’s hydrogen core network is 9,040 km on the plan and 537 km in the ground.
• The constraint has moved from power to water, ports and consent. Power queues are the known problem. Water is the one that stops a fab, a housing pipeline and a lithium basin alike; vessels and quay space are the ones that stop offshore wind; wastewater headroom is the one that stops tens of thousands of homes.
• The better risk-adjusted position is the bottleneck, not the asset. Shared power and water for industrial districts, grid reinforcement and firming, treatment and reuse, quay and vessel capacity, common-user minerals infrastructure. These are contracted, monopoly-adjacent, long-dated cash flows sitting upstream of assets that cannot run without them — and they are being funded last.

Capital runs on the short clock. Consent runs on the long one.

Every case in this paper reduces to the same arithmetic. The production asset is a private decision taken on a balance sheet: land, equipment, a construction contract, a commissioning date. It is quick because money makes it quick. The enabling layer is a public decision taken through a statutory process: route selection, easements, environmental assessment, objections, a regulated price determination, procurement, then construction. It is slow because process makes it slow, and capital does not accelerate process.

South Korea states the gap plainly. A semiconductor fab is built in about three years — TSMC completed its Kumamoto plant in twenty months. The dedicated pipeline and intake works that will supply the Yongin cluster are costed at seven years at best and more than ten at the longer end, with first water around 2031 and full volume in the mid-2030s. A transmission line to the same cluster takes about ten years. The fab is a corporate decision; the water and the line are not.

The consequence is not that the asset is never used. It is that the asset is used late, at lower utilisation, with contracted capacity it cannot draw and a cost base carried against revenue that has not arrived. On a project financing, that lands as a longer construction-to-cash period and a tighter cover ratio. On an equity case, it lands as a return earned over fewer years than the model assumed. Neither is what the committee approved.

The queue is the known constraint, and it is still growing.

The International Energy Agency puts more than 2,500 GW of renewable, large-load and storage projects in grid connection queues worldwide. That is the headline number, and it understates the problem, because a queue measures projects waiting rather than projects abandoned before they joined one.

On the generation side the symptom is curtailment. Across the European Union roughly 72 TWh of mainly renewable output was curtailed in 2024 on grid bottlenecks — about the annual electricity consumption of Austria — at a cost, including compensation, of some €8.9 billion. Germany alone spent €3.07 billion on congestion management in 2025 across 30.3 TWh of measures, of which 9.4 TWh was renewable output turned down. That is energy built, paid for, and thrown away because the wires to move it were not built alongside it.

On the demand side the symptom is the inverse: load that cannot connect. Utilities in the large North American data-centre hubs report reserved capacity far in excess of current draw, and “bring your own power” conditions and reserved-capacity charges are appearing in tariffs. Australia is the clearest published case. More than forty large data centres are planned across Sydney and Melbourne, carrying over A$100 billion of projected investment — and on Bloomberg’s count only 23 per cent of them have publicly confirmed electricity supply or a grid connection: seven of twenty-eight active projects in Sydney, three of fifteen in Melbourne. Transgrid states that Sydney lacks the transmission capacity to connect any of the 20 GW of inquiries it has received, with half the proposed demand sitting within twelve kilometres of a single substation and amounting, on its own, to the peak winter load of New South Wales. Nationally, more than 20 GW is announced or early-stage against roughly 1.5 GW operational, and AEMO reports that over a third of proposed grid-connected data centres were cancelled last year.

The grid is answering, but on its own clock. AEMO’s National Electricity Market connection pipeline reached 80.9 GW in the September 2026 quarter, up 43 per cent on the year and, for the first time, larger than the 77 GW of capacity actually in service. Batteries are 55 per cent of it. The qualification matters more than the headline: 88 per cent of that capacity sits in the earliest application and proponent-implementation stages. A pipeline is not a connection, and a record pipeline against a lagging reinforcement programme is the mismatch restated, not resolved.

Water is the constraint that stops a fab, a suburb and a lithium basin alike.

Power queues are now priced. Water is not, and it is the harder constraint, because a grid can be reinforced in a corridor already in public hands while a catchment cannot be enlarged at all.

Semiconductors

TSMC’s Arizona complex carries a total planned investment of about US$265 billion across twelve facilities. The first fab draws roughly 4.75 million US gallons of water a day — about 18,000 cubic metres, or the supply of around 15,000 Phoenix households. Sixty-five per cent of intake is currently reclaimed against a target of ninety per cent or better, and the industrial water reclamation plant that delivers it is not fully operational until 2028.

The basin is going the other way. Under the post-2026 Colorado River operating rules — Record of Decision issued 21 August 2026, not a consultation — the Lower Basin takes a 1.25 million acre-foot reduction in each of 2027 and 2028, of which Arizona carries 760,000 acre-feet a year. On groundwater the position is unresolved rather than tightening: the Arizona Department of Water Resources’ unmet-demand rule, which had constrained new assured-supply determinations in the Phoenix Active Management Area, was struck down by the Maricopa County Superior Court in April 2026 on procedural grounds and is subject to appeal. The 100-year assured water supply requirement remains. An investor should read that as uncertainty in both directions, not as relief.

This is not an Arizona peculiarity. On SEMI’s analysis, more than 40 per cent of semiconductor facilities announced since 2021 sit in watersheds projected to face high or extremely high water stress between 2030 and 2040. The industry is siting its next decade of capacity in the catchments least able to supply it.

Housing

The same constraint, one sector across, stops houses. In the United Kingdom, wastewater and sewage headroom is holding up large numbers of consented units, with Northern Ireland the most acute case; one study puts the cost to the British economy of undelivered housing on water grounds at £25 billion over five years. In Ireland, Uisce Éireann’s capacity in the key growth areas supports well below the Government’s housing targets. In Canada, a material share of housing-enabling infrastructure is in poor condition and local capacity is being rationed. In Western Sydney, industrial and residential pipelines have contracted on road, water and wastewater access.

Where national water frameworks exclude or poorly integrate industrial and data-centre demand — which is common — the two queues compete for the same headroom without anyone having allocated between them. That is a policy gap, and it is also a pricing opportunity for whoever builds the treatment and reuse capacity that resolves it.

Announced is not built, and in hydrogen the gap is now measurable.

Hydrogen is the cleanest demonstration of the thesis, because both numbers are published: what was announced, and what exists.

Worldwide, announced hydrogen pipeline projects — new build and repurposed gas line together — exceed 40,000 km by 2035. Nine per cent of that length is operational or carries a committed investment. Germany, the most advanced programme in Europe, has an approved core network of 9,040 km; 537 km of it is in operation, with a further 900 km expected by the end of 2027, taking the network to 16 per cent of its approved length eight years after approval.

The demand side has now moved to meet the supply side, downward. ENTSOG’s draft Hydrogen Infrastructure Gaps Identification Report for TYNDP 2026, issued for consultation on 9 October 2026, cuts the European demand outlook for 2030 to around 335 TWh a year and for 2040 to 1,296 TWh a year.

Even on the reduced case the infrastructure does not reach the demand. ENTSOG finds that 12 per cent of European hydrogen demand in 2030 goes unsatisfied for want of transmission and storage; that curtailment occurs in thirteen of twenty-seven member states; and that nine countries face curtailment running from 70 per cent to full disruption. South-Eastern Europe and the isolated systems are structurally the most exposed. Paid capacity reservations exist but are thin, and electrolysis capacity remains far below the targets set against it.

Australia shows the same conversion failure from the other side. The country holds roughly 15 per cent of announced global electrolyser capacity, yet announced projects fell from 90 to 76 during 2025 and those operating or under construction from 30 to 28. BP cancelled the 105 MW H2Kwinana project in June 2026 after its Hydrogen Headstart application was unsuccessful; CQ-H2 at Gladstone, Whyalla, Trafigura’s Port Pirie electrolyser and the Hunter Valley ammonia project have all been abandoned, terminated or stalled. The common feature is not engineering. It is that the offtake, the firm power and the transmission were never contracted on the same clock as the plant.

Offshore wind is short of vessels and quay, and neither can be ordered in a year.

Offshore wind has moved past the stage where the turbine is the hard part. The hard parts are a vessel that can lift it and a berth that can stage it, and both are capital goods with multi-year lead times held by a small number of yards.

Equinor’s Empire Wind is the instructive case. The 810 MW project is about 80 per cent complete — all fifty-four foundations, the offshore substation and the cabling are in — yet first power has slipped to the middle of 2027. The cause is not the turbines. The installation vessel is foreign-built and so cannot carry components between United States ports; the project therefore depends on American-built tugs and barges shuttling out from Brooklyn, and those tugs and barges are themselves still being built. A Jones Act constraint on a hull became a delay on a gigawatt-scale generation asset.

The fleet behind it is thin. Outside mainland China there were 38 wind turbine installation vessels operating at the end of 2025 with five on order, and 51 heavy-lift vessels with seven on order — orderbooks of about 13 and 14 per cent of fleet against a segment described as structurally undersupplied. Marketed utilisation for turbine transport and installation fell to 64 per cent at the end of 2025 from 70 per cent a year earlier, and is forecast back to roughly 79 per cent in 2026. The nuance worth holding is that the newbuilds that have been ordered are largely being ordered at Chinese yards by non-Chinese owners. The shortage is of building capacity, not of willingness to order.

Quay space is the same problem fixed to the ground. The Port of Rotterdam has taken a 45-hectare site at the Maasvlakte to the market for a dedicated offshore wind terminal with 835 metres of heavy-duty deep-sea quay able to take jack-ups, targeted to open in mid-2029. The project is still at request-for-information stage — no final investment decision, no operator award — precisely because suitable space in a working port is scarce and has to be assembled, not bought.

Meanwhile the money has turned. Global offshore wind investment fell 72 per cent year on year in the first half of 2026, to roughly US$12 billion, while European awards continued. The capacity target and the capital to install it are now moving in opposite directions.

The factory slips two years; the charger slips a quarter at a time.

Volkswagen’s PowerCo cell plant at St. Thomas, Ontario — C$7 billion, about €4.8 billion, and 90 GWh at full build — has pushed production start from 2027 to 2029. PowerCo attributes the two-year slip to market demand, technology and group strategy rather than to the grid, and that is fair; but a plant of this size is also a connection, a substation and a transformer order, and those lead times do not shorten while the decision is reconsidered. Several Canadian and European battery investments have slipped on comparable timelines.

Charging shows the distribution-network version of the same constraint. The United States passed 80,000 public DC fast-charging ports in the third quarter of 2026 — 80,964 across 14,614 stations — but the rate of addition is falling: 4,106 ports in the quarter, 12 per cent fewer than a year earlier, across 20 per cent fewer new stations. The federal programme behind a slice of that build has been cut rather than redirected: roughly US$500 million of NEVI funding was rescinded in February 2026, a further US$879 million is proposed for transfer and US$4.2 billion of unobligated funds is proposed for cancellation. In Europe, AFIR is driving corridor coverage and power density, but the binding constraint is low-voltage reinforcement and the concentrated depot loads that heavy vehicles impose — again a network question, not a charger question.

Battery storage itself is the exception that proves the thesis. Deployment is rising strongly, because storage is the one asset whose economics improve as the mismatch worsens: it is paid to absorb the curtailment and firm the connection. That is the shape of the opportunity in this paper, visible in the one place where the market has already found it.

In minerals, the enabling layer has finally been named out loud.

This firm is for the metals — copper, gold, silver, lithium, the critical minerals and the rare earths — without qualification. Nothing the modern economy intends to build, the energy transition and the data centre included, is built without them, and the capital case for them is stronger today than at any point in this cycle. Which is exactly why the infrastructure point matters here more than anywhere else.

BloombergNEF puts the copper investment requirement at about US$332 billion in new mines, against a deficit beginning in 2028 and reaching 20 million tonnes a year by 2050 — some 44 per cent of projected demand. That is not a geology problem. The orebodies are known. What is missing at remote sites is power, water, roads, rail and port access, and those are the items no single project can justify alone and no single project should be asked to.

The market said so plainly in Sydney this month. At IMARC, from 27 to 29 October, mining financiers argued that governments should stop selecting individual winners and fund shared infrastructure instead. Traxys’ chief executive put the state’s role as water and power infrastructure, roads and transport, together with long-term debt; Argonaut’s executive chairman called multi-user infrastructure the fairest form of support and observed that picking winners “is not a skill which is held in government”; Appian noted the obvious arithmetic that an electoral cycle is a fraction of the fifteen years it takes to develop a mine.

The Australian Government’s updated Critical Minerals Strategy, released on 27 October 2026 in the same week, adopts the language: coordinated infrastructure development, enabling infrastructure in roads, rail, ports, energy and water, and a dedicated commitment on common-user processing facilities. It should be read for what it is. The update carries no new money — it restates the existing A$17.5 billion production tax incentive and the A$5 billion critical minerals facility — and it reframes the policy from climate toward national security and economic resilience. The diagnosis has been accepted. The balance sheet behind it has not yet moved, and that gap is where private capital is invited in.

In the region, the constraint is arriving before the capital does.

Johor is the clearest Southeast Asian case. It has become a major data-centre hub on the strength of land, proximity to Singapore and cheap power, and power is now the binding constraint: large reserved capacity is only partially drawn, and approvals are tightening around demonstrated access to power and water rather than around land and capital. Batam faces the parallel problem in a smaller system, where data-centre interest tied to Singapore meets electricity and water reliability that were not sized for it. Elsewhere in the region, charging networks are lagging vehicle growth for the same distribution-network reasons they are in Europe.

This firm works from the region, and the pattern on the ground is more pointed than the published data suggests. The investment proposition is presented as land, tax and proximity. The question that decides the project is firm power and reliable water over a twenty-year horizon, and it is usually the last question asked. Where a host system is small, the data centre is not a marginal load on the grid — it is a material share of it, and the political economy of that only becomes visible after the commitment is made.

Eight sectors, four continents, one structure.

Set the cases side by side and the common element is not the technology, the geography or the policy regime. It is that in every one of them the asset was financed on a short clock and the thing it depends on was not.

Buy the bottleneck, not the asset that is stuck behind it.

Announced capacity is a forecast. Connected capacity is an asset.

The investment consequence of a persistent timing mismatch is not that the production asset is worthless. It is that its return is earned later and over fewer years than the model assumed, while the asset that resolves the mismatch earns a contracted return from the day it is commissioned and has no competition, because nobody else built one.

For special-situations, infrastructure and physical-economy mandates, four positions follow.

1. Underwrite the ecosystem before the asset. Secured firm power, secured water over the asset’s life, secured berth or vessel slots, secured offtake. Treat an unsecured enabling input the way a lender treats an unsecured title: not as a discount, as a condition precedent.
2. Price announced capacity as a forecast. Nine per cent of announced hydrogen pipeline is built or funded; 23 per cent of Australian data-centre projects have confirmed power; 88 per cent of the record Australian connection pipeline is at the earliest stages. A queue position is not a connection and a memorandum is not a contract.
3. Go up the chain to the bottleneck. Shared power and water for industrial districts; grid reinforcement and firming; treatment and reuse; quay, yard and vessel capacity; common-user processing for minerals. These are long-dated, contracted, often regulated cash flows sitting upstream of assets that cannot operate without them. They are funded last and they should be funded first.
4. Read the public balance sheet honestly. Governments have adopted the enabling-infrastructure language — Australia’s critical minerals update is the clearest example — without attaching new money to it. That is not a reason to wait for the state. It is the reason the opportunity is available.

Where this firm stands.

The physical economy thesis is right and this firm is long it. We are for the metals, for industrial and commercial property, for the data centres and for the generation that has to be built to run them. Nothing in this paper is an argument against any of that.

The argument is about sequence. The last three years have funded the visible end of each value chain on the assumption that the invisible end would arrive on time, and across eight sectors and four continents it has not. The enabling layer is slower for structural reasons that capital cannot fix, and precisely because it is slow and unglamorous it is under-owned. That is not a defect in the opportunity. That is the opportunity.

The question this firm puts to a board or an investment committee before it commits to a production asset is short, and it is not about the asset. It is: show me the grid connection, show me the water licence, show me the berth. If any of the three cannot be shown, the better trade is usually to own the thing that is missing.

Where every claim in this paper comes from.

• Ontario Teachers’ Pension Plan, The Physical Economy: Investing Through a Changing Energy Landscape
• International Energy Agency, Electricity 2026, Grids chapter — over 2,500 GW stalled in grid queues worldwide
• International Energy Agency, Global Hydrogen Review 2026, Trade and infrastructure — announced pipelines exceed 40,000 km by 2035, 9 per cent operational or committed
• Aurora Energy Research, via Clean Energy Wire, EU grid package — 72 TWh curtailed and €8.9bn of curtailment cost, 2024
• Bundesnetzagentur / SMARD, German congestion management, full year 2025 — €3.071bn, 30.3 TWh of measures, 9.4 TWh of renewable curtailment
• ENTSOG, TYNDP 2026 Hydrogen Infrastructure Gaps Identification Report, draft for stakeholder consultation, 9 October 2026
• S&P Global, ENTSOG cuts the 2040 European hydrogen demand outlook by a third
• FNB Gas, Two Years of the Hydrogen Core Network, 22 October 2026 — 9,040 km approved, 537 km in operation, 900 km more by end-2027; Bundesnetzagentur core network approval, 22 October 2024
• Geoscience Australia, Australia’s Energy Commodity Resources 2026 — hydrogen project counts, data to December 2025
• Argus Media, BP drops the 105 MW H2Kwinana project after an unsuccessful Hydrogen Headstart application, 24 June 2026; CSIRO HyResource, H2Kwinana
• Arizona Commerce Authority and City of Phoenix, TSMC’s additional US$100bn, taking total Arizona investment to US$265bn, 16 July 2026
• AZFamily, TSMC Arizona water demand — 4.75 million gallons a day, 65 per cent reclaimed against a 90 per cent target, reclamation plant fully operational in 2028
• US Bureau of Reclamation, Colorado River post-2026 operations, Record of Decision 21 August 2026; Arizona’s 760,000 acre-foot annual reduction
• KJZZ, Maricopa County Superior Court strikes down the ADWR unmet-demand rule, April 2026; KTAR, ADWR to appeal
• Korea JoongAng Daily, fabs in about three years against transmission lines in ten, 12 August 2026; Tom’s Hardware, Yongin water supply schedule
• SEMI, Ripple Effects: Water Risk and Resilience Across the Semiconductor Value Chain — over 40 per cent of facilities announced since 2021 in watersheds projected for high or extremely high water stress by 2030–2040
• The Maritime Executive, Equinor on Empire Wind — about 80 per cent complete, first power mid-2027, delayed on United States-built tugs and barges, 28 October 2026
• Westwood Global Energy Group, offshore vessel fleets — WTIV and heavy-lift fleet and orderbook excluding mainland China, utilisation to end-2025
• Port of Rotterdam Authority, offshore wind terminal — 45 hectares, 835 metres of quay; request for information and the mid-2029 target, November 2025
• BloombergNEF, renewable energy investment, first half 2026 — offshore wind investment down 72 per cent year on year
• Electric Autonomy Canada, PowerCo delays St. Thomas to 2029, 24 September 2026; Volkswagen Group, up to €4.8bn / C$7bn by 2030
• Paren, State of the EV Charging Industry, Q3 2026 — 80,964 fast-charging ports, additions down 12 per cent year on year; NEVI programme status, rescission and proposed transfers
• Home Builders Federation, A Drain on the Nation — about 30,000 homes blocked in England on wastewater capacity
• Wastewater Infrastructure Group, Wastewater Constraints and the Cost of Delay in Northern Ireland, July 2026 — 55,000 homes impeded, construction halted in 23 towns and villages
• The Irish Times, Uisce Éireann capacity against the 300,000-home target, 10 August 2026
• Policy Connect, Bricks and Water, 14 July 2026, restating Public First for Enabling Water Smart Communities, January 2025 — £25bn over five years in undelivered housing
• Financial Post, Canadian water and wastewater constraints on housing — over 60,000 homes delayed in Toronto; over 11 per cent of assets in poor or very poor condition
• Bloomberg, Australian data-centre projects stalling on grid strain and community opposition, 22 October 2026
• AEMO, NEM Connection Scorecard, September 2026 quarter — 80.9 GW pipeline, up 43 per cent, 88 per cent at the earliest stages
• BloombergNEF, Transition Metals Outlook 2026, 20 October 2026 — US$332bn copper investment gap
• Reuters via Mining.com, financiers at IMARC Sydney on shared infrastructure, 29 October 2026; IMARC 2026, 27–29 October, ICC Sydney
• Australian Government, Critical Minerals Strategy 2023–2030 update, 27 October 2026, and Focus Area 5, coordinated infrastructure and common-user facilities
• Brook, Don’t Tar All Natural Resources With the Same Brush, PanEuro, October 2026; Brook, Australia Has Repriced Its Largest Asset Class, PanEuro, September 2026