The nuclear renaissance is real. The construction map is not evenly shared.

The global nuclear renaissance is being built, but not evenly. Policy support is broad; construction is concentrated; repeat commercial delivery is proven in only a small number of programmes. On 12 July 2026, IAEA PRIS listed 77 reactors and 80.720 GWe of net capacity under construction, of which China held 37 reactors and 39.952 GWe—48.05% of units and 49.49% of capacity. That makes China the fleet-scale case. Much Western re-entry remains spread across restart, review, licence, finance and a small set of builds, so the useful question is which project has reached which evidence state.S01S05S06S12S13S14S15S16S17S18

Those 77 reactors matter. They are physical projects that have crossed PRIS’s construction-start threshold, not a list of speeches or aspirations. But the total is not a forecast of 77 completions. PRIS begins under-construction status at the first major placement of concrete and ends it at first grid connection. It then calls a reactor operational from first grid, even when the period before declared commercial operation is still trial operation. A large live stock therefore proves that construction is under way. It says nothing by itself about when each unit will connect, when it will enter commercial service, how much older capacity will retire, or whether a programme can repeat the result.S01S02

The policy map is wider still. The IEA found support for expanding nuclear power in more than 40 countries, while the July 2026 PRIS construction table covered a much smaller set. National instruments also speak in unlike measures: the United States set a 400 GWe objective for 2050, the United Kingdom an ambition of up to 24 GWe by 2050, India a 100 GWe target for 2047, and Japan an outlook of approximately 20% nuclear generation in fiscal 2040. These are signals of direction, not quantities that can be added or translated into an implied reactor order. Every named project must still cross its own review, licensing, finance, construction and operating milestones.S01S06S26S27S28S29

This distinction changes the decision in front of a utility partner, supplier, lender or public authority. The decision-relevant unit is not a country headline or even a reactor name in a database. It is the latest achieved stage, the evidence that supports it and the next missing proof. The argument follows five steps: the live construction stock is real and concentrated; status is not a capacity forecast; serial delivery is the strategic advantage; fleet retention alters the net-capacity result; and the decisive proof in the 2030s will be successive units entering commercial operation. That sequence identifies which programme merits pursuit or deeper diligence now, and which still asks the reader to underwrite an announcement.S01S02S12S13S14S15S16S17S18S22

The present cycle is easier to understand against the first one. IAEA’s historical series records 305 construction starts totalling 264.128 GWe net during the 1970s. The annual high was 43 starts in 1976. First-grid connections peaked later, tying at 33 units in both 1984 and 1985; the 1985 cohort carried the higher capacity total, at 31.129 GWe net. That sequence does not prove one simple cause, but it shows the temporal shape of a build wave. Starts come first; grid connections arrive years later; commercial declarations can follow later again. Today’s policy and construction figures sit at different points on that same chain and should not be read as simultaneous output.S02S03

The earlier scale-up was not merely a set of reactor decisions. It depended on owners able to carry long projects, regulators able to license and oversee them, financing structures that could sustain construction, industrial supply chains that could make nuclear-grade systems, and a workforce able to repeat specialised work. The institutional point remains current. The IAEA’s Milestones Approach describes a nuclear programme as a system that also includes safety, security, safeguards, radioactive-waste management, legal and regulatory frameworks, human resources and decommissioning. Civil nuclear capacity is therefore not only an industrial-output question, and an increase in construction volume does not remove those continuing obligations.S03S30

After the first surge came a much quieter period. From 1988 through 2007, annual global construction starts never exceeded seven. Yet the interval still contained 73 starts and 61.505 GWe net. Pause is therefore too absolute. The more accurate description is a 20-year low-start interval: activity thinned sharply in aggregate while programmes in parts of Asia continued. That distinction matters because the capabilities inherited by countries were uneven. Some maintained active construction and regulatory pipelines; many advanced economies carried operating fleets forward while new-build experience became intermittent. The current re-entry challenge is partly the attempt to rebuild that delivery system while replacing or extending an ageing fleet.S03S04S06

The geography shifted further after 2010. The IEA’s 2025 review counted 52 reactor construction starts since 2017, of which 25 used Chinese designs and 23 Russian designs. Its later global review reported ten construction starts totalling 12.2 GW in 2025: nine in China and one in Russia. It also said that 94% of starts over the preceding decade used Chinese or Russian designs. These are measures of design origin and start location, not full attributions of ownership, finance or every supplier. Even with that boundary, they show why the contemporary cycle is different from a broad return to the old Western build map. The centre of active repetition has moved.S05S06

The most recent complete annual IAEA series records nine construction starts and six first-grid connections in 2024. The IEA then records ten starts in 2025. These adjacent figures are evidence of continuing activity, but they remain modest beside the 1970s peak and should not be merged into a forecast. They also capture different milestones: a start is first major concrete, while a grid connection marks the beginning of the PRIS operational state and may precede commercial operation. The current cycle becomes a renaissance in the stronger sense only if the start cohort moves through trial into commercial service and is followed by later units, rather than accumulating as a long-lived construction stock.S02S03S05

The first change is a stricter evidence ladder. Policy or target establishes direction. A named proposal establishes that an official body or authorised owner has identified a project. Regulator-accepted review establishes a live application, not a licence. A licence establishes a defined permission, often with conditions and regulatory control points. Finance or final investment decision shows capital commitment, but it does not substitute for construction authority. First major concrete establishes the PRIS construction start. First grid establishes connection and begins the PRIS operational state. Trial operation sits between first grid and declared commercial operation. A later unit reaching commercial operation begins to demonstrate repeat delivery. The order can overlap in practice, but no stage is evidence of the next.S02S14S15S16

China’s own records illustrate why these distinctions are substantive rather than semantic. At the end of May 2026, the national regulator reported 37 construction-licensed units and another 11 approved units awaiting construction. A paired licence for Xuwei Units 1 and 2 released the first-concrete control point for Unit 1 while retaining it for Unit 2. PRIS recorded Unit 1 entering construction four days later; Unit 2 had no construction-start row at the 12 July cut-off. The absence of that row is not evidence of cancellation. It shows that approval, a paired construction licence, release of a unit-specific regulatory control point and first concrete are different achieved states even inside a serial-build system.S01S09S10S11

The second change is concentration at both project and design level. China accounts for roughly half of the live PRIS construction stock by unit count and net capacity. The IEA’s recent-start record is also dominated by Chinese and Russian designs. This does not mean every project has the same delivery condition, nor that design origin identifies the full supply chain. It does mean that broad interest elsewhere has not yet produced an equally broad construction map. When a buyer asks where repeatable near-term demand may exist, the live evidence points first to programmes already licensing, pouring first concrete and advancing multiple units, not simply to jurisdictions with large 2050 ambitions.S01S05S06

The third change is the strategic meaning of repetition. A transparent sample of 12 completed units across five site programmes—Fuqing, Fangchenggang, Taishan, Sanmen and Barakah—contains seven same-site successors. Only two of those successors reached commercial operation faster than the selected first unit at their site. Construction-to-commercial durations in the sample range from 68.8 months for Fuqing 5 to 113.1 months for Sanmen 1. The sample is too small and varied to estimate a universal learning rate, and it contains no cost evidence. It shows something more useful: repetition creates programme continuity and reusable institutions, but the label repeat unit does not guarantee monotonically shorter elapsed time.S03S09S22

The United States and Canada show three distinct modes. Vogtle Unit 4 is a recent large-reactor completion, evidence that the United States can finish a modern new-build unit; it is not yet proof of a repeated multi-site programme. Palisades remained a restart project in licensing and environmental review at the cut-off. The rescission of earlier shutdown certifications did not itself authorize restart or establish output. At Darlington, the Canadian regulator had issued a construction licence for one BWRX-300 and the project had entered licensed SMR construction, while a separate future operating-licence process remained ahead. Completion, restart licensing and SMR construction are all forms of re-entry, but they retire different uncertainties.S12S13S14S31

The United Kingdom and France add finance and operating-status distinctions. Hinkley Point C remained under construction under regulatory oversight. Sizewell C had a nuclear site licence and a final investment decision, but the site licence did not permit nuclear-related construction; finance and construction therefore remained separate milestones. Flamanville 3 had connected to the grid and remained in start-up or testing without a public commercial-operation date at the cut-off. PRIS would call that unit operational from first grid, but the comparison must call it connected or in trial until an authorised commercial declaration appears. These cases are not weak because their labels are precise. The precision tells a buyer which work is real and what proof is still missing.S02S15S16S17S18S32S33

Japan and Korea demonstrate a different split between fleet restoration and new-build continuation. Onagawa Unit 2 completed a regulated restart into commercial operation. That is meaningful retained capacity, but it is not a new-build completion. Saeul Unit 3 had an operating licence and public start-up generation evidence while still lacking an achieved commercial declaration at the cut-off. An operating licence authorizes a stage; observed generation shows connection or trial; neither should be promoted into the owner’s later commercial milestone. For a market participant, opportunities around restart, commissioning and new construction may overlap, but the delivery evidence and risk still differ.S19S20S34S35

India separates an achieved cohort from a long-range target. Kaiga Units 5 and 6 reached first concrete as the first project in a sanctioned fleet-mode cohort. That is materially stronger evidence than the national objective of 100 GWe by 2047, but it does not convert the whole objective into financed or licensed units. The useful diligence question is what follows Kaiga: whether subsequent units in the sanctioned cohort pass their own permissions, funding and construction-start milestones. Treating the target as a build count would hide the difference between a policy direction and the actual throughput of India’s delivery system.S21S28

Newcomer programmes show how far apart nominally similar national stories can be. The United Arab Emirates completed commercial operation of the four-unit Barakah programme. Turkiye remained in construction and commissioning at Akkuyu; fuel delivery or a target for first electricity did not establish grid connection. Egypt had all four El Dabaa units under construction without achieved grid dates. The Philippines remained further upstream: its government had adopted a nuclear-energy roadmap and was developing national infrastructure, but it had no reactor in the July PRIS construction set. These are programme completion, active construction or commissioning, multi-unit construction without connection, and policy or institutional preparation. They belong on one global map only if the map keeps their stages visible. The IAEA newcomer framework also makes clear that construction is not the whole national undertaking: regulatory independence, safety, security, safeguards, waste, human resources and decommissioning remain enduring responsibilities.S01S22S23S24S30S36

Latin America exposes a second distortion: equal unit counts can conceal profoundly unequal scale and condition. PRIS assigns one under-construction unit to Argentina’s CAREM25 prototype at 25 MWe net and one to Brazil’s Angra 3 at 1,340 MWe net, a 53.6-to-one capacity ratio. Neither had a recorded first-grid date. The administrative label is useful because it preserves a common lifecycle definition, but it is insufficient as a delivery judgement. A credible comparison must pair unit count with capacity, construction age, achieved milestones and current authoritative project evidence. One reactor is not one standard quantum of programme progress.S01S25

New-build volume is only one side of the capacity equation. PRIS listed 417 operational reactors with 379.700 GWe net on 12 July 2026. Of that fleet, 283 reactors and 256.719 GWe were at least 30 years from first grid; 207 reactors and 182.478 GWe were at least 40 years; and 50 reactors and 38.550 GWe were at least 50 years. These are nested cohorts, not quantities to add. Age is not a retirement forecast, licence expiry, generation measure or safety judgement. It is evidence that a large share of the fleet will require continuing owner, regulator and investment decisions while new projects advance.S04

Life extension can therefore affect the capacity outcome as much as construction. The IEA reported more than 60 reactor life-extension decisions in the five years preceding its 2025 report, covering nearly 15% of the global fleet. Separately, the OECD NEA Low Scenario expects more than 50 GWe to retire by 2040, while its Current Trends Scenario expects 10 GWe as long-term operation and investment retain more capacity. Those are conditional model outputs, not predictions of individual reactor closure. A life-extension decision is not a guarantee of extra operating years either: inspections, refurbishment, licensing and owner economics still intervene. The contrast shows why an outlook cannot subtract assumed retirements mechanically or treat all older reactors as fixed future capacity. The net path is the result of commercial additions, permanent shutdowns, restarts and authorised continued operation, each on its own evidence trail.S04S06S07

The same bottleneck applies to future scenarios. The OECD NEA’s 2050 cases reach 347 GWe in Low, 619 GWe in Current Trends, 883 GWe in Ambitious and 1,324 GWe in Transformative. Only the Transformative case exceeds the tripling objective. These are conditional modelled outcomes, not forecasts. The higher cases depend on linked changes in project execution, industrial capability, workforce, supply chains, investment frameworks and finance, together with long-term operation and new build. The capacity endpoints cannot be detached from the delivery systems assumed to produce them.S07S08

That is the practical bottleneck: conversion. More policy support can widen the top of the funnel, but it cannot by itself supply qualified people, regulator capacity, repeat orders, finance structures or commercial completions. An announcement-heavy pipeline can coexist with a concentrated build map and an ageing operating base. The programmes most likely to matter to a buyer are those that can show both sides of the equation: a credible route through the next project milestone and a realistic account of fleet retention or replacement. The gap between 619 GWe and 1,324 GWe is not a difference in optimism. It represents a much more demanding execution system.S04S06S07S08S13

The next 60 months should be read as a sequence of achieved proofs, not as a race between target totals. Four signals would show whether construction is broadening beyond its present concentration and whether re-entry is becoming a repeatable delivery system. Each signal has a public evidence owner and a condition that would materially change the answer.S01S02S24S25

Taken together, the signals create a usable scorecard. First concrete tests whether policy and finance are converting. Commercial operation tests whether construction is completing. A repeat unit tests whether delivery is becoming institutional rather than exceptional. Net capacity tests whether new build is outrunning loss from the existing fleet. The buyer should ask for the next dated record in that chain, identify who must issue it and decide what evidence would stop the commitment. Evidence-state discipline is more demanding than counting announcements, but it is also more actionable: it shows where a project can still fail to prove its case.S01S02S04S07S08S25