Something unfamiliar is happening to the energy transition,as its hardest problem is no longer the technology it set out to build. In 2025 the world added 693 GW of renewable capacity, the largest addition ever recorded by IRENA, and cumulative solar passed 3 terawatts in mid-2026 on BloombergNEF's count. Yet the IEA now counts more than 2,500 GW of renewable, storage and large-load projects stalled in grid connection queues worldwide, which captures the imbalance in a single statistic: clean generation became affordable and fast while the system that must connect, move, balance and deliver it remained slow and expensive to expand. That collision between record generation, resurgent demand and a delivery network built for a slower century is now visible in prices, queues and politics on every continent, and it will do more than any technology cost curve to decide how fast electrification actually happens.
This Research investigates rather than assumes: that the transition is entering an infrastructure phase in which the speed of grid development, more than the cost of renewable generation, increasingly sets the pace of decarbonisation. We audited and re-verified a 90-claim research dossier, traced journalistic claims to primary sources, discarded figures that failed verification, and followed the causal chain from generation growth through connection queues, wires, machines, minerals, capital, permitting and politics to find which links actually bind.
What this research does not support is the easy conclusion that the world simply needs more transmission, because the constraint turns out to be different things in different places. Across the OECD the binding problem today is institutional, found in connection processes and permitting regimes that stretch a wire to between 8 and 13 years while a solar plant takes one to three. Just behind it sits a harder physical wall that the queue numbers obscure, a hollowed-out industrial base in which US transformer lead times average 128 weeks, Europe's three big cable makers hold order backlogs near €35 billion, and copper, trading near $14,500 a tonne in August 2026, comes from mines that take 18 years to develop. China faces a different limit again, one of absorption, where record grid spending still trails a generation boom that pushed solar curtailment to 9.2% in early 2026. And across most of Africa the constraint is capital itself, since the grid there is not congested but absent, with 560 million people unconnected and the continent drawing about 3% of global energy investment.
The sceptic's case gets a full chapter, because parts of it are right. Three quarters of queued US projects historically never got built, demand forecasts are demonstrably inflated, and flexibility tools that already exist could unlock capacity equal to more than half the global queue. Taken together, those corrections mean the bottleneck is smaller than the queue numbers suggest, yet also more durable than the reform headlines imply, because its slowest-moving components, the factories, the mines and the distribution networks, all run on decade-long clocks that no single policy cycle can compress.
For 20 years the energy transition had one master variable: the cost of clean generation. Every learning curve, every auction record, every gigawatt milestone tracked a single question, whether solar, wind and batteries could get cheap enough to displace fossil fuels on economics alone. By the middle of this decade the answer was so thorough that the industry stopped asking. Renewables took 85.7% of all capacity added worldwide in 2025 on IRENA's count, wind and solar met 99% of last year's demand growth on Ember's data, and renewables edged past coal as the world's largest electricity source. Generation, as a problem, is substantially solved.
What replaced it is harder to photograph. Electricity consumption reached 28,600 TWh in 2025 and the IEA expects growth to accelerate, 3.6% this year and 3.8% next, heading toward roughly 40% more demand by 2035 under its central scenarios. Behind that acceleration stands everything the transition was designed to produce: EVs and heat pumps landing on residential feeders, industry electrifying process heat, hydrogen electrolysers requesting firm connections, data centres arriving in 500 MW blocks that a distribution planner of 2015 would have mistaken for a typo. Every one of those loads, and every new generator built to serve them, must pass through the same physical funnel of connections, wires, substations and transformers.
That funnel is old and narrow. The world's grid runs to roughly 80 million km of lines, 93% of it distribution, and in advanced economies more than half was laid over 20 years ago, per the IEA's landmark grids study. The same study set out what national climate targets quietly assume: adding or refurbishing 80 million km by 2040, which amounts to rebuilding the entire existing global grid in 15 years. Against that requirement the money is flowing to the wrong side of the meter, roughly $1 trillion a year into power generation and about $400 billion into grids, a 40-cent ratio the IEA says must move toward parity with grid spending above $600 billion a year by 2030.
This is the shift this investigation examines: from generation economics to system economics. In the old model, value came from producing a megawatt-hour more cheaply than the incumbent. In the new one, value comes from producing it in a place, at a time, and through a connection where it can actually be delivered. The scarce asset is no longer the panel or the turbine but the network capacity behind it. A megawatt-hour of Scottish wind curtailed behind a constraint and re-bought as gas generation in the south of England is not cheap power; it is £2.7 billion a year of system cost, and rising.
None of this was unforeseeable, and grid planners have warned about it for a decade. What changed in the past 24 months is that the warning stopped being theoretical. It acquired prices, a tenfold rise in PJM capacity auctions. It acquired queues, 2,500 GW of them. It acquired casualties in offshore wind farms waiting on delayed onshore connections, and politics in a US administration cancelling transmission corridors while a Dutch minister explains to parliament why 14,000 companies cannot get a connection. The grid stopped being background infrastructure and became the story.
2,500 GW waits. The IEA's global count of projects stalled in connection queues equals roughly ten times the installed power capacity of the entire African continent, all of it parked in paperwork. The best-documented queue is American: Lawrence Berkeley National Laboratory's Queued Up census found 2,061 GW of generation and storage pending at end-2025, led by 773 GW of solar and 749 GW of storage, while Europe's queues held about 1,700 GW on European Commission figures and Britain's reached 722 GW, four times what its own 2030 plan requires, before regulators intervened. For the projects that survive the wait, the median time from US connection request to commercial operation now exceeds five years, which is longer than it takes to build the plant itself.
Read carefully, though, because the queue is two stories at once and the second undercuts the first. LBNL's cohort analysis is brutal: of everything that entered US queues between 2000 and 2020, just 13% reached commercial operation and 75% withdrew. Developers file speculative requests at multiple substations to discover their costs, then abandon the losers, so queue totals measure option-taking as much as intent. When Britain forced the question with its "first ready, first connected" reform, filtering 722 GW through readiness gates, just 283 GW of confirmed generation and storage connections emerged alongside 99 GW of demand connections. The queue was a fiction; the 283 GW is a plan. The US is running the same experiment at continental scale, and since FERC's Order 2023 reforms survived their court challenge in July 2026 the queue has shrunk for the first time in years. One technology bucked the trend. Queued gas capacity rose 86% in a single year, to 253 GW, as developers read the bottleneck and reached for the fuel that comes with its own delivery system.
Where projects do connect, the next filter is absorption, and the evidence that systems are saturating is now measurable in five different currencies. Britain curtailed 13% of its potential wind output in FY2024/25 and paid £1.7 billion in thermal constraint costs to manage the gaps. Germany held renewable curtailment to 3.5% but still spent €2.9 billion on congestion management in 2024. China, which builds grid faster than anyone, watched solar curtailment climb to 9.2% in January and February 2026, brushing the 10% ceiling Beijing itself set when it relaxed the old 5% rule. California curtailed 3.4 TWh in 2024, although batteries have begun pulling the rate down, and Australia's market curtailed roughly 8 TWh in 2025, up more than 60% in a year. The causes differ between hours, network limits in some and simple oversupply in others, and this investigation is careful not to blame every curtailed megawatt-hour on a missing wire. The direction, though, is the same everywhere: each new gigawatt added to a saturated system earns less and wastes more than the one before it.
Prices tell the same story from the other side. Negative wholesale prices, once a curiosity, reached about 6% of all hours in 2025 in France, Germany, the Netherlands and Spain on IEA data, upwards of 500 hours per market. A larger figure of 9,000-plus negative-price hours circulates for Europe as a whole, but it sums dozens of bidding zones and rests on a single trade report, so this investigation stays with the per-market numbers. Their financing consequence is already measurable: lenders to European merchant renewables have compressed loan tenors from 15-20 years to 5-10 and increasingly require co-located storage before term sheets move, per Renewable Energy World's reporting on the 2025-26 financing market. The market is repricing delivery risk in real time, even where regulators have not.
Strip away the queues and the politics and the bottleneck reduces to two clocks running at different speeds. A utility-scale solar plant takes about four years from conception to energisation on the Energy Transitions Commission's accounting, and with clean permitting can be done in one. A new transmission line takes 8 to 13 years on the IEA's reckoning of European and American experience, meaning a developer can build the same solar plant twice over, three times in some jurisdictions, in the period its evacuation line spends in development. Everything else in this investigation, the queues, the curtailment, the gas plants ordered as workarounds, follows from that arithmetic.
Why does a wire take a decade? Because it is less an engineering project than a serial negotiation. Route a line across 500 km and you negotiate with every landowner, council, environmental assessment and appeal along it, then with a regulator over who needs it and who pays, then with a sellers' market for cable and transformers, and only then with the terrain. Each stage is sequential, few can be parallelised, and any one can reset the clock. Australia has priced each failure mode with unusual candour: Project EnergyConnect, the South Australia interconnector, roughly doubled from A$2.28 billion to about A$4.1 billion and slipped years, with Transgrid now seeking A$1.14 billion of the overrun from consumers. VNI West nearly doubled to A$7.6 billion and moved its completion two years to 2030 amid sustained farmer protest that the state's own farming federation attributes to poor early engagement. HumeLink sits near A$4.9 billion. None of these are engineering surprises; they are the compounding cost of time.
America shows what happens when the negotiation simply fails. The US completed 888 miles of high-voltage transmission in 2024, against roughly 4,000 miles built in 2013 and the ~5,000 miles a year its own national planning study implies is needed, per Grid Strategies. Federal policy then turned actively hostile in 2025-26: the Department of Energy cancelled the remaining National Interest Electric Transmission Corridors in August 2026 and had earlier pulled a $4.9 billion loan guarantee from Grain Belt Express after a senator's campaign against it. Yet the deeper signal runs the other way. Invenergy responded by awarding $1.7 billion of Phase 1 construction contracts to Quanta and Kiewit anyway. SunZia, 550 miles of HVDC carrying 3,000 MW of New Mexico wind, reached commercial operation in June 2026 after 17 years of development, and Champlain Hudson energised the same month to carry up to a fifth of New York City's needs. The lines that survive the decade are transformative; the problem is how few enter the pipeline, and how long the decade is.
China is the control experiment. Where the state carries siting, cost allocation and procurement in one decision, the wire compresses: roughly 45 ultra-high-voltage lines spanning about 65,000 km with some 360 GW of transfer capability on Global Energy Monitor's compilation, with 15 more planned by 2030. GE Vernova's grid chief now describes 2 GW HVDC as the industry's emerging standard building block, and China deploys it as routine. The lesson is not that autocracy builds better; it is that the OECD's constraint is not concrete or conductor but the institutional metabolism that turns intent into rights-of-way.
Transmission gets the photographs, pylons against sunsets and subsea cables on drums. But 93% of the world's 80 million km of grid is distribution, the low-voltage capillaries running down every street, and that is where the transition physically lands. The heat pump, the EV charger, the rooftop array and the home battery all plug into feeders engineered decades ago for one-way flow and modest evening peaks. Distribution is also where the bottleneck is hardest to see: a saturated transmission line announces itself in congestion prices, while a saturated feeder announces itself only as a refusal letter.
The Netherlands received the refusal letters first. Roughly 14,000 companies were waiting for a new or heavier grid connection in late 2025, the climate minister told parliament, with network operators investing about €8 billion a year and still rationing capacity. Parts of the country face queues stretching into the 2030s, and housing developments wait alongside bakeries and factories. Nothing about Dutch demand is exotic; the Netherlands is simply the first dense, prosperous, rapidly electrifying economy to hit the ceiling of its distribution stock, which makes it a preview rather than an outlier.
Australia previews the opposite failure mode, not scarcity of capacity but abundance of generation in the wrong layer. At midday on 4 October 2025 the National Electricity Market's operational demand fell to a record-low 9,666 MW while rooftop solar met 61% of underlying demand from 15 GW of household panels. Feeders built to deliver power now spend their middays absorbing it, and the engineering questions around voltage management under reverse flow, hosting-capacity limits and who may export when have arrived years before the tariff structures and control systems designed to answer them.
Nigeria completes the triptych with distribution failure of a different order entirely. The constraint there is not hosting capacity but the commercial collapse of the distribution layer itself: sector-wide aggregate technical, commercial and collection losses hit 37.44% in Q1 2026 against a 16.92% regulatory target, costing ₦140.6 billion in a single quarter on NERC's figures, with more than 40% of registered customers unmetered. When the wires lose four naira in every ten to leakage and non-payment, no amount of upstream generation or transmission investment can make the system bankable. Chapter 12 returns to this in depth.
Which is harder, moving power between regions or orchestrating millions of small assets? The evidence suggests transmission is harder politically and distribution harder operationally, but the decisive asymmetry is simpler: transmission failure is priced and therefore managed, while distribution failure is silent until it is absolute. No market index tracks feeder headroom, and the Dutch waiting list was a surprise to almost everyone outside the network companies that maintained it.
Suppose every reform worked. Suppose queues were filtered to real projects, permits moved in two years, and cost allocation stopped being litigated. Would the wires then get built on schedule? This investigation's answer, tested against procurement data rather than policy documents, is no. Behind the demand for grid capacity sits a second, quieter constraint: the world's physical ability to manufacture it, a hypothesis the research began with and the evidence has now largely confirmed.
The distinction matters because the two constraints respond to different levers on different clocks. A queue rule can change in a regulatory cycle, and Britain proved it in under two years. A transformer factory takes roughly three years to build and qualify, a cable plant closer to five, and a copper mine, on S&P Global's data, an average of 17.9 years from discovery to production. If the binding constraint is institutional, the bottleneck is a policy failure and reform fixes it. If it is industrial, the bottleneck has a physical floor that no permitting bill can lower, and every national grid plan on earth is implicitly bidding against every other for the same machines. The evidence in the next chapter says the industrial floor is real, already binding at the margin, and geographically concentrated to a degree that would alarm any strategist who found it in semiconductors.
One number frames the stakes before the detail: between them, national plans imply rebuilding the world's entire grid inventory by 2040, while equipment order books at the three firms holding two thirds of the global HVDC cable market are already sold out to the end of the decade. The queue measures who wants capacity; the order book measures who will get it.
Start with the machine no substation works without. A large power transformer is a bespoke, 200-tonne object wound from grain-oriented electrical steel, filled with specialist components and tested for months, and almost nobody makes them anymore in the places that now need them most. US procurement lead times averaged 128 weeks in Q2 2025 on Wood Mackenzie's tracker, with generator step-up units near 144 weeks and the largest custom classes stretching to four years per the IEA, all of this against roughly one year before 2020. Prices have risen about 77% since 2019, a figure the IEA and Burns & McDonnell independently corroborate in the 75-80% range. The dependency is concentrated: the US imports 82% of its large power transformers per the Department of Energy's resilience report to Congress, its single domestic producer of transformer steel, Cleveland-Cliffs' Butler Works, covers a fraction of national demand, and Japan's Nippon Steel makes the only top grades.
Cables are tighter still, and the market structure tells you why. Prysmian's transmission order backlog stands near €17 billion, NKT's high-voltage backlog at €10.2 billion with another ~€2 billion converted in January, and Nexans' at €7.7 billion. That is roughly €35 billion of committed future work at the three European firms the IEA estimates hold about two thirds of the global HVDC cable market. The agency's transmission report puts AC cable procurement at two to three years, DC cable waiting lists beyond five, and European production slots effectively booked to 2030. A claim of 8-to-10-year HVDC delivery slots circulates in industry commentary, but it did not survive this investigation's verification; the IEA's booked-to-2030 finding is the strongest citable version, and it is quite strong enough. A developer signing today for a project needing DC cable is negotiating for the early 2030s.
Beneath the machines sits the metal. Copper broke $12,000 a tonne for the first time on 23 December 2025 and traded near $14,500 in mid-August 2026 with exchange inventories falling for 42 consecutive sessions. The IEA's Global Critical Minerals Outlook 2026 projects a 25% copper supply gap by 2035 in its base case, and Wood Mackenzie sees demand rising from 34.5 to 42.7 million tonnes a year over the same period, requiring more than 8 Mt of new annual mine capacity against that 17.9-year development clock. Grids are the demand driver hiding in plain sight: under IEA scenarios, grid build alone consumes 9 to 12 Mt of copper and 21 to 27 Mt of aluminium annually by the 2040s. The final input is people. Goldman Sachs Research tallies 750,000 new power-sector workers needed across the US and EU by 2030, including 207,000 in US transmission and interconnection, while Australia's official skills agency wants 32,000 more electricians by decade's end.
Can the world manufacture grid infrastructure fast enough to meet its own plans? On current evidence, not on the plans' timelines, but the response has begun and its dates can be read off press releases. Hitachi Energy is spending $9 billion globally, including a ~$457 million Virginia large-transformer plant operational around 2028. Siemens Energy's $421 million Charlotte factory produces from 2027. GE Vernova committed roughly $600 million to US facilities, Cleveland-Cliffs is adding a $150 million transformer plant in West Virginia, and the White House invoked the Defense Production Act for grid equipment in April 2026. Every one of those dates lands after 2027, which defines the shape of the decade: a supply crunch through roughly 2028, easing thereafter if and only if the announced capacity is built and the order books keep justifying it. Meanwhile the marginal transformer increasingly ships from China, whose power-equipment exports rose 26% in 2025 to $37.4 billion, with transformers up 45%.
If new wires take a decade and new factories land after 2027, the decisive question for the next five years is how much capacity can be conjured from infrastructure that already exists. The answer, on the IEA's arithmetic, is larger than most policymakers appreciate: grid-enhancing technologies could unlock 450 to 700 GW of queued capacity and flexible, non-firm connection arrangements another 750 to 900 GW. Together that is 1,200 to 1,600 GW, more than half the global queue, available without a single new corridor. Each tool deserves scrutiny rather than enthusiasm, so what follows takes each in turn, its size, speed, cost and limits.
Batteries. The fastest-scaling grid asset in history, with 108 GW added worldwide in 2025 on IEA figures (BNEF counts 112 GW and 307 GWh; both are stated because their methods differ), and deployable in one to two years. The relief is measurable where density is high: California's 15.7 GW of storage cut the solar curtailment rate from 13% to 11.5% in early 2025 even as solar output grew, and Australia's Waratah Super Battery, 850 MW and 1,680 MWh, operates as a system-protection scheme that raises transfer limits into Sydney, effectively doing a transmission line's job under contract. The limit is temporal and spatial: batteries shift energy across hours, not seasons, and they relieve congestion only where they are sited.
Demand flexibility. The cheapest tool on the list and the least deployed. Duke University's Nicholas Institute finding reframes the entire load-growth panic: 76 GW of new US load could connect if curtailable just 0.25% of hours a year, and 126 GW at 1%, because the grid is sized for peaks that occur a few dozen hours annually. US virtual power plants reached about 33 GW against a DOE target of 80-160 GW by 2030. Google has contracted roughly 1 GW of data centre demand response (company-reported), and Sunrun's August 2026 agreement to back Voltus's hyperscaler capacity deals from residential batteries in PJM and MISO extends the model, though its size is undisclosed. The limit is contractual rather than technical: almost no market yet pays flexibility what a deferred wire is worth.
Grid-enhancing technologies. Reconductoring with advanced conductors raises line capacity 50 to 110% at roughly half the cost of new build, and the Berkeley/GridLab study finds it could quadruple US interzonal capacity growth to 2035 while saving $85 billion. Dynamic line rating adds 10-40% headroom on existing conductors, per DOE. Deployment remains thin for a reason this investigation regards as the single most fixable distortion in the sector: utilities earn regulated returns on capital deployed, so a cheap fix that avoids a large asset is, under most remuneration rules, a worse business than the large asset. The WATT Coalition's filings to DOE document the result in detail.
Stability engineering. The April 2025 Iberian blackout settled a debate that had run for years on assertion alone. ENTSO-E's 49-member expert panel found a voltage-control failure in which conventional plants delivered less than 75% of required reactive power at the critical moments, over 2.5 GW of generation disconnected within about 50 seconds, and Iberia lost synchronism with continental Europe. Of its 22 recommendations, none is "build less renewables"; they concern voltage ranges, ride-through and control speed. The tools exist and are scaling. Grid-forming inverters feature in 74% of Australia's 33 GW battery pipeline, China mandates grid-forming capability in tenders, and the world's largest grid-forming battery, 1 GW and 4 GWh, connected in Inner Mongolia in August 2026.
None of this substitutes for wires. Flexibility arbitrages time while transmission arbitrages geography, and a system needs both. What flexibility buys is the scarcest commodity in this story: years. Deployed hard, it can hold systems together through the equipment crunch of 2026-28 while the factories and corridors catch up.
Be precise about causality, because the loudest storyline of 2026 gets it backwards. AI did not create the grid bottleneck; underinvestment, queue design and a hollowed industrial base did that over 15 quiet years. What data centres did was arrive suddenly, in concentrated gigawatt blocks, at the exact moment the system had no slack left, and make the bottleneck impossible to ignore. They are the flash that revealed the photograph.
The demand numbers deserve their error bars. Data centres consumed about 415 TWh globally in 2024, some 1.5% of electricity, and the IEA's base case takes that to roughly 945 TWh by 2030, with a 2035 scenario span of 700 to 1,700 TWh. EPRI's high case has US data centres at 9.1% of national generation by 2030, Goldman Sachs projects 165% growth, and an IEA-linked review of the forecasting field documents divergence wide enough to swallow whole national grids. Every figure in this paragraph is a scenario, and this investigation treats each accordingly. Concentration, by contrast, is observed fact. Data centres already take 22% of Ireland's metered electricity, Virginia's Dominion zone winter peak is up 45% on 2019-20, and ERCOT holds connection requests summing to 367,790 MW by 2032 against an all-time actual peak of 85,508 MW, a number ERCOT's own chief executive discounts and one analyst calls impossible. Hyperscalers shop the same gigawatt to multiple utilities, and every duplicate inflates every forecast it touches.
Where the load is real, the price response has been violent. PJM's capacity auction cleared at $28.92 per MW-day for 2024/25 delivery, rose to $269.92 the next year, climbed through $329.17 and $333.44, then hit its negotiated $325 cap in July 2026 for the third consecutive cap-bound auction, with capacity costs up roughly $12.5 billion in two years and PJM attributing nearly all incremental peak growth to data centres. Utilities are answering in the oldest currency they know: TVA's board approved a plan on 20 August 2026 targeting 7 to 26 GW of new gas by 2040, and Duke Energy's Carolinas filing three days earlier seeks 14 GW of gas over 15 years. The politics arrived on schedule. A majority of North Carolina voters now oppose data centres in their communities, over 20 of the state's local governments have moratoriums, and a non-binding White House "ratepayer protection" pledge collected 23 governors and 187 companies in July.
Yet the same load carries the seed of the opposite outcome. A data centre is the most controllable gigawatt-scale load ever connected, and its work can, in principle, wait minutes or move continents. Google has proved the principle at 1 GW scale, EPRI's DCFlex demonstrations are testing it with utilities on three continents, and the Duke University arithmetic shows that accepting curtailment for one quarter of one percent of hours would let roughly 76 GW connect without new peak capacity. Nuclear PPAs, some 13 GW of announced hyperscaler commitments led by Microsoft's 835 MW Crane restart, are real but land years out on Carnegie's sceptical accounting. The fork is stark and contractual. Hyperscalers that accept curtailability become the grid's largest flexibility resource this decade; those that insist on firm 24/7 connections will keep commissioning gas plants, and the emissions cost of AI will be set not by chips but by interconnection clauses.
Governments have started treating grid equipment the way they treated semiconductors five years ago, and the paper trail is specific. In April 2026 the White House invoked the Defense Production Act for transformers, high-voltage components and power electronics; the Department of Energy's resilience report had already put the vulnerability in one figure, that 82% of US large power transformers are imported, with a destroyed unit having no off-the-shelf replacement. The EU's December 2025 Grids Package designates eight fast-tracked "energy highways" and recycles 25% of TSO congestion revenues into priority projects. India's ~$110 billion transmission plan is explicitly a state programme. China pairs a CNY 4 trillion five-year grid plan with an export machine whose power-equipment shipments grew 26% in 2025 to $37.4 billion, transformers up 45%. The pattern is legible: grid capacity is being reclassified from utility procurement to industrial policy.
Concentration is what makes it strategic rather than merely commercial. China refines about half the world's copper. Japan and Korea dominate the transformer-steel grades the US cannot make, with Nippon Steel alone producing the highest specification. Three European firms hold two thirds of global HVDC cable capacity, a chokepoint Europe rarely counts as bargaining power but could. Every one of these concentrations is a single point of failure in someone else's transition plan, and unlike oil none has a spot market. You cannot buy a transformer at Rotterdam; you wait 128 weeks, or you have a relationship, or you have a factory.
Electrification is also import substitution, a fact the 2026 Hormuz crisis demonstrated in wholesale prices. When roughly a fifth of global LNG supply briefly went offline, EU and Japanese power prices jumped more than 30% year on year while US prices barely moved, and the IEA calculates clean energy investment avoided $260 billion in fossil import costs in 2025. A grid-dependent electricity system converts fuel-security exposure into equipment-security exposure. That trade is favourable because equipment fails less often than straits close, but it is still a trade, and the countries treating transformer plants and cable capacity as strategic industries have noticed before the ones still treating them as line items.
A line energised in 2030 will still be carrying power in the 2080s, in a climate none of its design codes anticipated. The summer of 2026 offered previews on three continents, cutting in both directions. Europe's heatwave pushed cooling demand to records while stressing thermal plants, and Ember's analysis credits record solar output with keeping the grid stable through peak heat, a resilience finding as significant as any failure. In the US, Lake Powell and Lake Mead fell to record lows in August, threatening the hydropower that firms the Southwest, while parts of Indiana went more than a week without power after summer storms. A developing super El Niño is expected to reshape weather patterns well into 2027.
The engineering consequences are unglamorous and specific. Conductors sag and derate in heat precisely when air-conditioning demand peaks. Wildfire risk forces pre-emptive de-energisation and is redirecting utility capital from expansion to hardening across California and Australia. Drought curtails hydro and the cooling water that thermal and nuclear plants require, while floods take substations that sit, for historical reasons, disproportionately on low ground. This investigation looked for a credible global figure for grid climate-adaptation costs and reports plainly that none exists; the gap is flagged in the register rather than papered over with an estimate. What can be said with evidence is structural: resilience spending draws on the same crews, transformers and regulatory headroom as expansion, which means adaptation and transition are one budget competing with itself, and the competition tightens with every hot year.
Run the same eight questions across six systems and the global picture stops being a blur of national stories. What drives demand, what is being built, where it binds, how long resolution takes, who pays, what policy is doing, what technology is helping, and what happens if nothing changes. The answers reveal one structural problem wearing six faces.
America is running both directions at once. The reform machinery held, with FERC's Order 2023 interconnection rules surviving the DC Circuit in July 2026 and the queue shrinking for the first time in years. The federal build machinery reversed: corridors cancelled, the Grain Belt loan pulled, $7.6 billion of clean energy grants cancelled on what a DOE court filing admitted were political grounds. What is actually getting built now runs sideways through merchant developers, states and RTO portfolios, with SunZia and Champlain Hudson both energised in June 2026, while data centre load runs the capacity market at its price cap.
Europe has counted its gap with precision: €584 billion of grid investment needed this decade, 1,700 GW queued, 72 TWh curtailed in 2024 at a cost of €8.9 billion. It has legislated at its own pace, a Grids Package in December 2025 and an Electrification Action Plan in July 2026 targeting 46% of final energy by 2040. Germany's SuedLink arrives end-2028 versus an original 2022 date, the six-year gap bridged by €2-3 billion a year of redispatch. The Iberian blackout report has become the continent's grid-code moment, and negative prices its market-design one.
China falsifies the comfortable theory that money alone dissolves the bottleneck. State Grid spent a record CNY 650 billion in 2025 and will average roughly CNY 800 billion a year to 2030, with the UHV network spanning about 65,000 km. Deployment is faster still. The new Five-Year Plan targets more than 2,800 GW of wind and solar by 2030, Jiangsu province alone crossed 100 GW of solar in July, and curtailment climbed toward the state's own 10% ceiling anyway. Beijing's answer is to engineer absorption directly by mandating grid-forming storage in tenders and connecting the world's largest grid-forming battery in August.
India wrote the sequencing lesson into statute with a ₹9.15 lakh crore (~$110 billion) National Electricity Plan for transmission to 2032 covering more than 191,000 circuit-km of lines, inter-regional transfer capacity rising from 119 to 168 GW, all sized for a 500 GW non-fossil target that reached 300.5 GW in July 2026. The plan's soft underbelly is the layer below: distribution companies still lose 16.12% of power to aggregate technical and commercial losses, and their finances remain the chronic risk between build-out and bankability.
Australia is running both endgames simultaneously and publishing the invoices. The transmission-led path doubled in price project by project: EnergyConnect from A$2.28 billion to about A$4.1 billion, VNI West to A$7.6 billion and two years late against farmer opposition. The distributed path set records of its own, with rooftop solar meeting 61% of demand at the October 2025 minimum and grid-forming inverters featuring in three quarters of a 33 GW battery pipeline. The 2026 system plan trimmed remaining transmission needs to ~5,800 km, partly because batteries keep absorbing the jobs wires were once planned for.
Africa does not belong in this list as a sixth variation on congestion, which is why it gets the next chapter whole. One casebox here marks the contrast the rest of this chapter exists to sharpen.
560 million people. That is sub-Saharan Africa's unconnected population in 2024, within 655 million worldwide, per the multi-agency Tracking SDG7 report published in June 2026, which also finds the pace of electrification must triple to reach universal access by 2030. Set that against everything preceding this chapter and the inversion comes into focus. The OECD's grid problem is a system too full; Africa's is a system that barely exists. Sub-Saharan consumption outside South Africa averages around 180 kWh per person per year, against 6,500 in Europe, and low-income countries hold 33.6 watts of renewable capacity per person while high-income countries hold 1,224. These are not two points on one spectrum but different problem classes, and treating Africa as a smaller Europe has misallocated a decade of policy attention.
The finance numbers explain the gap better than any engineering account. Africa attracts about $110 billion of energy investment in 2026, some 3.3% of the global total for a fifth of the world's population, and under 2% of global clean energy spending on the IEA's earlier count. Capital costs run two to three times advanced-economy levels for utility-scale renewables, up to four times for access projects. The IEA's access arithmetic is stark: universal access by 2035 needs about $15 billion a year, split roughly $7 billion for grid extension, $5 billion for mini-grids and $3 billion for solar home systems, while actual tracked commitments run below $2.5 billion with private capital contributing $640 million. Mission 300, the World Bank and AfDB drive that has connected more than 50 million people since 2024 on ~$15 billion of commitments, is the scale of response required, and it remains a fraction of the need.
What grid exists, leaks. Only four sub-Saharan utilities keep losses at or below the World Bank's 10% benchmark, and eight countries exceed 30%. The regional pools that should pool scarcity instead demonstrate it: in one month the Southern African Power Pool recorded 6.1 GWh of unserved trade, matched transactions blocked by transmission limits, while the 2,000 MW Ethiopia-Kenya HVDC line runs at 200 to 400 MW of utilisation for want of downstream wires and demand contracts. West Africa's 15-country grid achieved full synchronisation with Nigeria for the first time only in November 2025. South Africa, the continent's one industrialised grid, proves the classification by exception: it has a true OECD-style congestion problem, with procurement rounds returning zero wind because the Cape grids are full, answered by a curtailment framework and Eskom's R390 billion, 14,000 km transmission plan.
Nigeria compresses every layer of this investigation into a single system, which is why it anchors this chapter. The capacity ladder tells the story: roughly 13.6 GW installed, 4,458 MW available in Q1 2026 on NERC's count, a transmission network whose operator claims 8.5 GW of wheeling capability (a figure TCN defends and critics dispute), a historic peak delivery of about 5.8 GW, and average delivery of 4,113 MWh per hour in Q1 2026. The grid collapsed 12 times in 2024 and twice more around the turn of 2026, more than 60% of manufacturers have left it entirely, and the standard if dated estimate puts self-generation capacity at roughly eight times grid peak. The sector needs about $12 billion a year and attracts about $1 billion, per ESI Africa's August 2026 analysis, whose title asked the question this publication exists to answer: Nigeria is attracting capital, but is the power following?
August 2026 supplied evidence in both directions inside two weeks. On the forward side, TCN commissioned 672 MW of new transmission capacity across four Lagos substations on 19 August, with the World Bank funding two of them and Japan's JICA committing ₦20 billion to the Apapa substation. On the backward side, NERC dissolved the board of Kaduna Electricity Distribution Company on 10 August over ₦456.5 billion in cumulative market debt, a 41.93% remittance rate and ATC&C losses of 71.88%, installing an interim board of special directors chaired by Abdullahi Garba with Abubakar Hashidu as administrator while Afreximbank runs a 12-month search for a new core investor. The pattern is the investigation's thesis in miniature: capital reaches the layers that can be structured, generation and now transmission via DFI money, and dies at the layer that cannot, distribution, where losses, debt and unmetered customers make revenue unbankable. Meanwhile the subsidy bill hit ₦1.98 trillion in the 12 months to September 2025, 15 states now regulate their own electricity markets under the 2023 Act, and NERC's 2026 net billing rules give commercial and industrial solar exporters billing credits, a quiet acknowledgment that the distributed layer is where Nigerian electricity actually works.
Can Africa build a different architecture rather than reproducing the OECD grid? The evidence supports a portfolio answer, not a slogan. On the distributed path, mini-grid costs fell from $0.55 to $0.38/kWh between 2018 and 2022 with a $0.20 target by 2030; the World Bank reckons 217,000 mini-grids could serve 490 million people as the least-cost option; Nigeria's $750 million DARES programme is the largest public mini-grid push ever mounted; Zambia scaled solar from 88 to 841 MW in five years; and Namibia just commissioned Africa's first integrated hydrogen hub as a fully off-grid industrial facility. Against the pure leapfrog, the IEA's own least-cost modelling still assigns grid extension the largest share of access investment at 47%, because cities, industry and anything electro-intensive need wires, and mini-grids do not power smelters. The convergent future is mini-grids built now where the grid is distant, designed to interconnect when it arrives, with the central spine reserved for the loads that justify it. Africa's advantage is precisely that it is not locked into the architecture the OECD is now finding so expensive to modify.
Follow any thread in this investigation far enough and it ends at the same question. Grids are paid for through a short menu of network charges on consumer bills, government spending, developer connection charges, targeted large-load tariffs, green bonds and infrastructure funds, and concessional finance where none of the above works. What is being contested on four continents is not the menu but the allocation, and the fights of 2026 are worth reading closely because they preview the politics of the entire grid decade.
Britain shows the cost of delay made legible. Balancing the system already adds about £3 a month to a typical domestic bill, and NESO's projection has constraint costs peaking near £8 billion in 2030 without faster network build, roughly triple today's level, while accelerated delivery could save up to half of that. The honest framing, which NESO itself now uses, is that consumers are already paying for the missing wires; they are just paying for their absence rather than their construction.
America shows the allocation fight at full volume. PJM's capacity costs rose roughly $12.5 billion in two years, ICF projects bills in data-centre-heavy regions rising 15-40% by 2030, and the political response so far is a voluntary pledge from 23 governors and 187 companies that is explicitly non-binding. TVA moved first on the substantive question by introducing a Capacity Commitment Charge that raises data centre bills about 10% over three years, the first serious attempt to make the load that causes the cost carry it. California's decade-long version of the same fight, over rooftop solar, now spans a four-fold disagreement: an $8.5 billion annual cost-shift on the Public Advocates Office's count, roughly $4 billion on UC Berkeley's, and a claimed $1.5 billion net saving on the solar industry's. This investigation presents the range because the range is the finding. Nobody agrees who owes whom, and the methodology chosen determines the answer.
Nigeria states the emerging-market version without euphemism: tariffs below cost for most customers produced a ₦1.98 trillion subsidy in 12 months, paid through the fisc rather than the bill, while the DisCos' unbanked losses ensure that neither consumers nor taxpayers get a working grid for the money. Australia's Transgrid, seeking A$1.14 billion of cost overrun from consumers, completes the tour. In every system, the gap between grid cost and grid charge eventually finds a payer, and the payer is usually the party with the least say in the planning.
Underneath the fights sits the structural asymmetry that makes grid finance hard everywhere: costs land now and locally while benefits accrue later and diffusely, and no regulator has found a popular bridge across that gap. Institutional capital is not the missing piece. Pensions and infrastructure funds hold exactly the duration grids need, and Australia's competitively tendered REZ concession shows them taking it when the structure exists. What is scarce is the structure itself: allowed returns that track equipment inflation, anticipatory investment rules, and in the South, guarantees that absorb the currency and offtaker risk private capital will not. Who captures the value if the bottleneck breaks? In order: consumers who stop paying congestion premiums, equipment makers with sold-out order books, flexibility aggregators paid to substitute for wires, and the landowners and jurisdictions that host what gets built. Who pays if it does not? The same consumers, invisibly, indefinitely.
A publication that only stress-tests other people's optimism is not doing its job, so what follows takes the strongest version of the case against this investigation's thesis, argued from the same evidence base.
First, the queues are paper. Seventy-five percent of two decades of US queue entries withdrew, Britain's 722 GW melted to 283 GW under a readiness filter, and the US queue is now shrinking. The 2,500 GW headline measures cost-discovery behaviour, not suppressed projects. Second, the demand forecasts are inflated exactly when inflation is most profitable. ERCOT's 368 GW is a sum of duplicate requests its own chief executive discounts, the forecasting literature diverges by a factor of two, and every gigawatt a hyperscaler shops to five utilities appears in five plans; some of the panic is book-keeping. Third, the cheap fixes are visibly working. Batteries cut California's curtailment rate while volumes grew, Waratah substitutes for a Sydney transmission upgrade, and the IEA's own 1,200-1,600 GW of flexibility headroom equals more than half the queue. Fourth, some waste is optimal. An electricity system with zero curtailment is overbuilt in wires, Germany's €435 million curtailment compensation is a rounding error in a €300 billion-plus system, and negative prices partly signal abundant clean generation, which was the point. Fifth, distributed energy shrinks the problem from below. Australian rooftops meeting 61% of demand at the midday minimum is transmission demand that vanished.
Weigh it fairly and the sceptic wins real ground without taking the field. Queue inflation is real, but the projects that did complete still waited more than five years, and shrinking queues have not shortened waits. Forecast inflation is real, but the strongest evidence in this investigation is not a forecast; it is cleared auction prices at caps, curtailment rates at records, refusal letters in the Netherlands, and order books sold out to 2030, all of which measure the present. Flexibility headroom is real and this investigation counts it as the decade's best lever, but it is unlocked by exactly the regulators whose incentive structures currently point the other way. Optimal curtailment exists, but 13% of British wind and 9% of Chinese solar sit far above any optimum in the literature, and distributed defection relieves feeders while creating coordination and payer-base problems that no distribution operator has yet priced.
Where does that leave the thesis? Qualified, and stronger for it. The evidence does not support the claim that the grid has stopped the transition, because renewables met effectively all demand growth in 2025. It does support the narrower claim this investigation set out to test: the pace-setting constraint has moved from generation cost to system delivery. The sceptic's corrections shrink the bottleneck's measured size by perhaps a third, and nothing in the sceptic's case touches the two slowest components, the industrial base and the distribution layer, at all.
What follows are scenarios, not predictions, and no probabilities are assigned because no credible model supports them. Elements of all three are already observable in different places, which is the point of drawing them.
Investment closes on the IEA's $600 billion-plus path. Readiness-gated queues generalise, GETs earn a remuneration route, the 2027-28 factory vintage relieves equipment, and flexible connection becomes the default offer. Watch for the WEI grid line rising toward $500 billion, US transmission completions recovering toward 2,000 miles a year, NESO's constraint costs peaking below projection, and transformer lead times falling back under 80 weeks. Consequence: electrification accelerates and the queue converts at rising rates. Britain's reform and China's spending are this scenario running early.
Demand and generation keep outrunning wires. Constraint costs compound toward NESO's £8 billion marker, capacity auctions stay capped or crisis-priced, connection dates slip past 2030, and the queue's fastest-growing technology remains gas. Watch for rising redispatch records, large loads defecting to on-site generation, and gas turbines ordered for "reliability" that run for decades. Consequence: the transition slows without any single visible failure, and the 2030s inherit both the emissions and the bill. PJM and the Netherlands are this scenario running early.
The grid stops being the default. Data centres build microgrids, industry self-supplies, households defect at the margin, and the network becomes a backup service with a shrinking payer base carrying fixed costs. Watch for behind-the-meter capacity growth outpacing grid connections and utility death-spiral economics in high-tariff jurisdictions. Consequence: faster decarbonisation for the connected few, stranded costs for everyone else, and the equity problem compounds. Lagos, where the productive economy already runs on roughly 40 GW of self-generation, and Perth's disconnection debate are this scenario running early.
What we know. The transition's constraint has moved. Generation is abundant, with 693 GW added in 2025, solar past three terawatts, and renewables now the largest source of electricity on Ember's accounting. Delivery is not. There are 2,500 GW in queues, a 40-cent grid dollar against the generation dollar, wires that take a decade against plants that take three years, and waste in curtailment, congestion and balancing that now runs to $12.3 billion a year in the US, £2.7 billion in Britain and €8.9 billion in the EU. These are observed prices and audited reports, not projections. We also know the constraint is not one thing: it is institutional in the OECD, absorptive in China, financial in Africa, and industrial everywhere underneath.
What we think. Ranked by the evidence, the biggest constraint this decade is institutional throughput: the queue rules, permitting clocks and remuneration models that decide how fast everything else is allowed to move. The fastest solution is flexibility, meaning batteries, curtailable load and grid-enhancing technologies, deployable in one to two years and able on the IEA's numbers to unlock capacity equal to more than half the global queue. The longest-term constraint is the industrial base, because transformer plants, cable factories, copper mines and a 750,000-worker gap all move on clocks no legislature can compress. The most underestimated risk is distribution, the 93% of the grid with no price signal, where saturation arrives as a refusal letter and the Netherlands has already received 14,000 of them. The most promising opportunity is the collision of AI money with grid scarcity, because hyperscaler capital is the first force in decades rich enough and impatient enough to pay for grid modernisation, if regulators price flexibility before utilities finish pouring gas-plant concrete. And the biggest uncertainty is demand itself, with a 700 to 1,700 TWh data centre range wide enough to swallow every plan written this year.
What we cannot yet prove. Whether the bottleneck slows aggregate decarbonisation rather than redistributing it, since renewables still met effectively all demand growth in 2025, and the sceptic's corrections in Chapter 14 stand. Whether the 2027-28 factory vintage arrives on time and sufficient. Whether reformed queues shorten waits for real projects or merely thin the paperwork around unchanged build rates. And whether Africa's financing gap closes; Mission 300's early pace is evidence it can, not that it will.
What happens next. Five dials through 2027 will tell the story: the IEA's grid investment line against $500 billion, US transmission miles against 2,000, transformer lead times against 80 weeks, PJM's first uncapped auction price, and whether Nigeria's new Lagos transmission capacity converts into delivered megawatts or strands against the DisCo wall. Those five will say which scenario the world is actually funding long before any ministry announces it.
The transition spent 20 years making clean electrons cheap. Its next decade will be decided by something less photogenic: whether societies can permit, manufacture, finance and staff the machinery that moves electrons, at the speed they have already committed to generating them.
This Deep Dive was produced by auditing, re-verifying and reorganising a 90-claim TCL research dossier compiled on 21 August 2026 from six parallel evidence sweeps (global baseline, queues and curtailment, supply chain, data centres and flexibility, country case studies, Africa and Nigeria). Every load-bearing figure was checked for source, publication date, underlying data period, geographic scope, units, capacity-versus-generation basis, and scenario-versus-forecast status. Journalistic claims were traced to primary sources wherever they exist; where a primary document could not be reached, the register below says so.
Kept (strong evidence, strong value): the IEA investment and queue series; LBNL queue and completion data; NESO, Grid Strategies and BNetzA cost series; ENTSO-E's Iberian blackout findings; the transformer, cable and copper supply-chain evidence; PJM auction results; IEA data centre scenarios; NERC Nigeria quarterlies; the country case studies; the sceptic's case and scenarios. Kept with qualification: China's UHV totals (derived sums, labelled approximate); Australia's 2025 curtailment (single-source aggregation); Google's 1 GW demand response and all factory-expansion figures (company-reported); TCN's 8.5 GW wheeling claim (disputed, both sides carried); the 2019 Nigeria genset estimate (dated, flagged); the 3 TW solar milestone (single analyst house); ERCOT's 2032 figure (requests-based, discounted in text). Verified this pass: the Kaduna DisCo order against NERC's primary text, which names a chair and administrator but no board headcount; the previously reported "5-member" figure was removed. Replaced in the research stage: the stale 2.6 TW US queue, the $20bn+ US congestion figure (2022 vintage), the 2024-vintage 10,000 km Australian ISP figure, and the superseded "600 million without power" (now 560 million in sub-Saharan Africa, 655 million globally). Excluded as unverifiable: "9,000 negative-price hours" as a single-market claim (it is a cross-zone sum); "8-10 year HVDC delivery slots"; Nigeria's "1.56 million meters in 2026"; the preliminary ~$550bn WEI 2026 grid figure (secondary source only; this publication retains the verified 2025 basis); and any global figure for grid climate-adaptation costs, for which no credible source was found.
Fifteen charts appear, each with source, period, scope, units, method and caveat stated beneath it. Three candidate charts were declined because their datasets are not comparable: a global connection-queue league table (queue definitions differ irreconcilably), a cross-regional congestion-cost comparison (US locational congestion, German redispatch and GB constraints measure different things), and an Africa access-versus-investment time series (no continuous comparable series exists). A proposed TCL Grid Bottleneck Index was assessed and deferred: of eleven candidate indicators, only three exist with comparable definitions across more than five major markets. A six-market pilot scorecard (US, GB, Germany, China, India, Nigeria) on five quarterly-sourceable indicators is recommended as a future TCL franchise instead. No figure in this publication was invented, interpolated or estimated; where an estimate appears, it is labelled as whose.
Five analytical claims in this investigation are The Climate Ledger's own synthesis rather than any source's finding, stated here for accountability: (1) the bottleneck is a sequencing problem more than an investment problem, the plant clock against the wire clock, and capital volume cannot fix a clock mismatch; (2) queue reform without industrial expansion moves the bottleneck to the factory door rather than removing it; (3) the binding constraint migrates downward into distribution as electrification spreads, and distribution's lack of a price signal makes that migration invisible until refusal; (4) the honest metric of the grid crunch is shifting from queue volume to the earnings spread between delivered and undeliverable power; and (5) AI capital is the first force rich and impatient enough to finance grid modernisation, and interconnection contract design, not chip efficiency, will decide whether it does.
Type: P = primary (agency, regulator, operator, court, company filing, official data); S = specialist research; J = journalism. Verification: V = verified against the stated source; V1 = verified but single-source; CR = company-reported; 2nd = reached via secondary source, primary pending. Retrieval date for all rows: 21 August 2026.
| Claim / figure | Source | Date | Data period | Type | Ver. | Caveat |
|---|---|---|---|---|---|---|
| Global electricity consumption 28,600 TWh; +3.0% (2025), +3.6% (2026f) | IEA Electricity Mid-Year Update 2026 | 23 Jul 2026 | 2025-27 | P | V | Consumption basis; 2026-27 are forecasts |
| Renewables 9,836 TWh in 2024 (+9.8%); 5.2 TW installed; 693 GW added 2025 (85.7% of additions) | IRENA Renewable Energy Statistics 2026 | 14 Jul 2026 | 2024-25 | P | V | Generation data lag capacity by a year |
| Renewables 33.8% of generation vs coal 33.0%; wind+solar met 99% of 2025 demand growth | Ember Global Electricity Review 2026 | 21 Apr 2026 | CY2025 | P | V | Generation basis, differs from IEA consumption basis |
| Solar past 3 TW cumulative, mid-2026 | BloombergNEF via Bloomberg | 3 Aug 2026 | mid-2026 | S | V1 | Single analyst house; crossing date debated |
| Grids ~$400bn vs generation ~$1,000bn a year | IEA World Energy Investment 2025 | Jun 2025 | 2025e | P | V | 2026 preliminary (~$550bn) excluded, 2nd only |
| >$600bn/yr grid investment needed by 2030; >80m km built/refurbished by 2040; >50% of advanced-economy grid over 20 yrs old | IEA Electricity Grids & Secure Energy Transitions | Oct 2023 | to 2040 | P | V | Requirement under announced policies, not forecast |
| >2,500 GW stalled in queues; GETs unlock 450-700 GW; non-firm connections 750-900 GW | IEA Electricity 2026, grids chapter | Feb 2026 | current | P | V | The One Number |
| US queue 2,061 GW end-2025 (solar 773, storage 749, gas 253 +86%, wind 220); median wait >5 yrs; 13% built / 75% withdrawn (2000-2020 cohorts) | LBNL Queued Up 2026 | May 2026 | end-2025 | P | V | Requested capacity, not expected build |
| FERC Order 2023 upheld | DC Circuit opinion 23-1282 | Jul 2026 | n/a | P | V | Court opinion is primary |
| GB queue 722 GW pre-reform; Gate 2 confirmed 283 GW + 99 GW demand | Ofgem TMO4+ / NESO via Knight Frank | Apr-Dec 2025 | 2025 | P/S | 2nd | Gate 2 results reached via secondary; NESO primary exists |
| GB balancing £2.7bn FY24/25 (£2.5bn prior); wind curtailment 13%; ~£8bn 2030 projection; ~£3/month on bills | NESO Balancing Costs Report + explainer | Jun 2025 / Apr 2026 | FY24/25; 2030 | P | V | £8bn contingent on build pace; savings up to £4bn |
| US congestion $12.3bn (2024), peak $20.8bn (2022); 888 miles of 345 kV+ built 2024 vs ~4,000 (2013), ~5,000 needed | Grid Strategies / Fewer New Miles | Nov 2025 / Jul 2025 | 2024 | S | V | Methodology stated in reports |
| Negative prices ~6% of hours FR/DE/NL/ES 2025; country counts 513-593 h Jan-Oct | IEA Electricity 2026 / pv magazine | Feb 2026 / Nov 2025 | 2025 | P/J | V | "9,000+ hours" is a cross-zone sum; excluded as headline |
| Germany congestion management €2.9bn (2024); curtailment compensation €435m (2025); 3.5% curtailed (2024) | BMWK/BNetzA via Clean Energy Wire | 24 Mar 2026 | 2024-25 | P/J | V | Ministry answer to Bundestag underlies |
| China curtailment: solar 9.2%, wind 8.5% (Jan-Feb 2026), ceiling relaxed 5%→10% | National Monitoring Center via Bloomberg | 8 Apr 2026 | Jan-Feb 2026 | P/J | V | Early-year reading is seasonally inflated |
| CAISO 3.4 TWh curtailed 2024; rate 13%→11.5% Jan-May 2025; 15.7 GW storage | US EIA / pv magazine USA | May-Jul 2025 | 2024-25 | P/J | V | Absolute curtailment still rose 4.1% |
| Australia NEM ~8 TWh curtailed 2025 (+62%) | RenewEconomy (NEM data) | 27 Nov 2025 | 2025 YTD | J | V1 | Single-source aggregation; AEMO corroborates direction |
| Iberian blackout: <75% reactive power delivered; >2.5 GW lost in ~50 s; 22 recommendations | ENTSO-E expert panel final report | 20 Mar 2026 | 28 Apr 2025 | P | V | Spanish government report differs in emphasis; both noted |
| Transmission lead times 8-13 yrs; AC cable 2-3 yrs; DC cable >5 yrs, slots booked to 2030; largest transformers ~4 yrs | IEA Building the Future Transmission Grid | Feb 2025 | 2024 | P | V | "8-10 yr HVDC slots" claim failed verification; excluded |
| Solar ~4 yrs (reducible to ~1), onshore wind ~10 (to 4.5), offshore ~12 (to 5.5), incl. permitting | Energy Transitions Commission | Jan 2023 | n/a | S | V | Includes full pre-development |
| US transformer lead time 128 wks avg, GSU ~144; prices +77% since 2019; 82% of LPTs imported; one US GOES producer | Wood Mackenzie via Power Magazine / DOE LPT Resilience Report | 2025 / Jul 2024 | Q2 2025 / 2019 | S/P | V | Price range corroborated by IEA and Burns & McDonnell |
| Cable backlogs: Prysmian ~€17bn; NKT €10.2bn; Nexans €7.7bn; big three ~2/3 of HVDC market | Prysmian FY25 / NKT FY25 / Nexans FY25 / IEA | Feb 2026 | end-2025 | P | CR | Segment definitions differ between firms |
| Copper >$12,000/t on 23 Dec 2025; ~$14,500/t Aug 2026; 25% supply gap 2035; demand 34.5→42.7 Mtpa; 17.9-yr mine lead time | Mining.com / IEA GCMO 2026 / Wood Mackenzie / S&P Global | Dec 2025-Aug 2026 | 2025-2035 | J/P/S | V | IEA and Wood Mackenzie boundaries differ; not combined |
| Grids need 9-12 Mt Cu + 21-27 Mt Al per year by 2041-50 (scenarios) | IEA grids report | Oct 2023 | 2041-50 | P | V | APS/NZE scenario ranges |
| 750,000 new power workers US+EU by 2030 (US 510k incl. 207k transmission); Australia 32,000 electricians | Goldman Sachs Research / Jobs and Skills Australia via AEC | Jul 2025 / Jul 2024 | to 2030 | S | V | Draws on CEWD/DOL surveys |
| Data centres 415 TWh (2024) → ~945 TWh (2030 base); 700-1,700 TWh 2035 range | IEA Energy and AI | Apr 2025 | 2024-2035 | P | V | Scenarios, not forecasts; rival estimates differ materially |
| Ireland: data centres 22% of metered electricity (2024) | CSO Ireland | 10 Jun 2025 | 2024 | P | V | 2025 figure (~23%) not independently fetched |
| PJM: $28.92 → $269.92 → $329.17 → $333.44 → ~$325 cap; costs +~$12.5bn in two years | PJM releases + Jul 2026 auction | 2024-26 | DY 24/25-28/29 | P/J | V | Latest two auctions cap-bound; scarcity price suppressed |
| ERCOT requests-based 367,790 MW by 2032 vs 85,508 MW actual peak | ERCOT release | 15 Apr 2026 | to 2032 | P | V | Requests-based ceiling; discounted by ERCOT's own CEO |
| TVA IRP 7-26 GW gas + data centre capacity charge (approved 20 Aug 2026); Duke Carolinas 14 GW gas filing (17 Aug 2026) | TVA via Utility Dive/PRN / Utility Dive | Aug 2026 | to 2040-41 | P/J | 2nd | Gas-plant $ figures vary $3-3.5bn across outlets |
| 76-126 GW of new US load integrable at 0.25-1.0% annual curtailment | Duke Univ. Nicholas Institute (Norris et al.) | 19 Feb 2025 | US system | P | V | Working paper, widely cited |
| Reconductoring +50-110% capacity at ~half new-build cost; ~4x interzonal growth to 2035; $85bn savings; DLR 10-40% (DOE) | UC Berkeley / GridLab / DOE | Apr 2024 | to 2035 | P | V | Modelled US findings |
| Batteries: 108 GW added 2025 (IEA); 112 GW / 307 GWh (BNEF); Google ~1 GW data centre DR; Sunrun-Voltus deal (MW undisclosed) | IEA GER 2026 / BNEF / Google / PV Tech | 2026 | 2025-26 | P/S | V/CR | Storage methodologies differ; DR figures company-reported |
| Waratah 850 MW/1,680 MWh as SIPS; grid-forming in 74% of 33.2 GW NEM pipeline; NEM min demand 9,666 MW, rooftop 61% | RenewEconomy / AEMO via ESN / AEMO via PV Tech | 2025-26 | 2025 | P/J | V | AEMO data underlies |
| Netherlands ~14,000 companies waiting; ~€8bn/yr grid investment | Minister Hermans / Netbeheer NL via DutchNews | 6 Oct 2025 | 2025 | P/J | V | Parliamentary statement underlies |
| EU €584bn to 2030; 1,700 GW queued; 72 TWh / €8.9bn curtailed 2024; Grids Package (10 Dec 2025); Electrification Plan 46% by 2040 (17 Jul 2026); ~80% to miss 15% interconnection target | EC / CLEW / EC Jul 2026 | 2023-26 | 2024-2040 | P | V | Electrification plan date corrected (Jul 2026) |
| SuedLink €10bn, completion end-2028 vs original 2022 | energynews.pro | 22 Sep 2025 | 2025 | J | V1 | TenneT/TransnetBW primary exists |
| China: CNY 650bn grid spend 2025; CNY 4tn 2026-30 (+40%); 45 UHV lines ~65,000 km ~360 GW; 2,800+ GW wind+solar target; Jiangsu 100 GW | Xinhua via Asia Financial / GEM / Carbon Brief | Jan-Aug 2026 | 2025-30 | P/S | V | UHV totals are derived sums, labelled approximate |
| India ₹9.15 lakh crore (~$110bn) transmission plan to 2032; 191,000 ckm; 119→168 GW inter-regional; AT&C 16.12%; 300.5 GW non-fossil (Jul 2026) | Ministry of Power (PIB) / PFC via Power Line / MNRE | Oct 2024-Aug 2026 | to 2032 | P | V | PFC rupee aggregates flagged inconsistent; % used |
| Australia: 2026 ISP ~5,800 km; EnergyConnect A$2.28→~4.1bn (A$1.14bn recovery sought); VNI West A$3.9→7.6bn, 2030; HumeLink ~A$4.9bn; CWO REZ A$5.5bn concession | AEMO ISP 2026 / ABC / RenewEconomy / IPA | 2025-26 | to 2050 | P/J | V | 10,000 km figure is 2024 ISP vintage; superseded |
| US federal reversals: NIETC corridors cancelled; Grain Belt loan pulled, $1.7bn Phase 1 proceeding; $7.6bn grants cancelled on politics (court filing); SunZia + CHPE energised Jun 2026 | DOE / Utility Dive / AP / NYSERDA | 2025-26 | 2025-26 | P/J | V | Grant admission anchored in court filing |
| Factory response: Hitachi $9bn programme incl. ~$457m Virginia (op. ~2028); Siemens $421m Charlotte (2027); GE Vernova ~$600m; Cliffs $150m Weirton; US DPA determination (Apr 2026); China equipment exports $37.4bn (+26%), transformers +45% | Hitachi / Siemens / Caixin / Utility Dive | 2024-26 | 2024-28 | P/J | CR | Expansion figures are company-reported |
| 655m without electricity globally; 560m in SSA; pace must triple; 33.6 vs 1,224 W/person renewables | Tracking SDG7 2026 | 24 Jun 2026 | 2024 | P | V | Supersedes "600 million" |
| Africa $110bn energy investment 2026 (3.3% of global); <2% of clean energy spend; capital costs 2-3x (access up to 4x); access needs ~$15bn/yr vs <$2.5bn committed ($640m private) | IEA Financing Clean Energy in Africa / IEA access finance / WEI 2026 via ThisDay | 2023-26 | 2023-26 | P | 2nd | $110bn via secondary; confirm in WEI 2026 Africa chapter |
| Mission 300: 50m+ connected, ~$15bn AfDB/WB committed | AfDB | 16 Jun 2026 | 2024-26 | P | V | n/a |
| SAPP 6.1 GWh unserved trade (one month); Ethiopia-Kenya 2 GW line at 200-400 MW; WAPP full sync Nov 2025; 4 SSA utilities ≤10% losses, 8 countries >30% | SAPP via ESI Africa / KETRACO via AEP / World Bank research via USITC | 2024-25 | 2015-25 | P/J | V1 | Import volumes single-source; loss data 2015 vintage |
| Nigeria: available 4,458 MW; avg 4,113 MWh/h; ATC&C 37.44% vs 16.92% target; ₦140.6bn quarterly loss; >40% unmetered | NERC Q1 2026 (via Arise) | 7 Jul 2026 | Q1 2026 | P | V | Installed capacity varies 12.0-13.6 GW by source |
| Kaduna DisCo: board dissolved 10 Aug 2026; ₦456.5bn debt; remittance 41.93%; ATC&C 71.88%; interim board chaired by A. Garba, administrator A. Hashidu | NERC interim order | 10-11 Aug 2026 | 2025-26 | P | V | NERC page states no board headcount; earlier "5-member" reports dropped |
| TCN +672 MW Lagos (19 Aug 2026); JICA ₦20bn Apapa; TCN 8.5 GW wheeling claim; sector needs ~$12bn/yr vs ~$1bn; >60% manufacturers off-grid; 12 grid collapses 2024; subsidy ₦1.98trn; 15 states self-regulating; gensets ~8x grid peak (2019) | Vanguard / NAN / ESI Africa / NERC / Dalberg/A2EI | 2019-2026 | as stated | P/J | V | TCN disputes "stranded" framing; genset figure dated, flagged |
| Mini-grids $0.38/kWh (2022), $0.20 target 2030; 217,000 could serve 490m; DARES $750m; Zambia solar 88→841 MW; Namibia off-grid hydrogen hub | World Bank/ESMAP / REA / CleanTechnica / ESI Africa | 2022-26 | 2018-26 | P/J | V | Zambia figures pending ZESCO/ERB primary check |
| South Africa: R390bn / 14,000 km TDP; BW6-7 zero wind on Cape grid constraints; R3.58-4.2trn to 2050 | Eskom / Pinsent Masons / ESI Africa | 2024-26 | to 2050 | P/S/J | V | n/a |
| California cost-shift range: $8.5bn (PAO) / ~$4bn (Berkeley) / -$1.5bn (M.Cubed); ICF: bills +15-40% by 2030; Hormuz: EU/Japan power prices +>30% y/y Q2 2026; $260bn avoided fossil imports (2025) | Canary Media/CPUC / Fortune / IEA | 2025-26 | 2024-26 | P/J | V | Cost-shift presented as a contested range by design |
| Climate previews: solar credited with European grid stability in heatwave (Ember); Lake Powell/Mead record lows; Indiana week-long outages; super El Niño developing | Ember via REW / REW / Carbon Brief | Aug 2026 | 2026 | P/J | 2nd | Ember report primary; fetched via trade press. No credible global adaptation-cost figure exists; none used |
The Climate Ledger tracks climate, capital and the industries shaping a net-zero future, with a founding focus on where clean energy money flows across Africa, where it stalls, and what it means for the hundreds of millions still without power. Policy, markets, deployment gaps. Monthly. This Deep Dive is Volume 02 in the series, following Who Owns the Batteries?. TCL analyses capital flows and does not provide investment advice. Corrections and primary documents are welcome at blogpost@theclimateledger.org.