The world is ordering electricity faster than it can build the networks that carry it.

The direct answer is that this is a whole-electricity-system build-out. Industry, cooling, transport, heat and digital infrastructure can add demand faster than networks, equipment, permits, flexibility and cost rules can be delivered. Generation can also be proposed faster than it can be connected. For a customer choosing a site, backing a generation portfolio or planning a network, the decisive question is therefore not how many megawatts appear in an announcement. It is whether supply, the connection study, physical network work, equipment, commissioning and energisation form one dated and funded route to operation.S01S02

The scale change is real, but its wording matters. The IEA expects the world to add about 1,100 TWh of electricity demand in each year from 2026 to 2030, compared with roughly 700 TWh a year over the previous decade. That is the basis for saying fifty per cent more demand is added each year: it refers to the absolute TWh increment, not a fifty per cent acceleration in the percentage growth rate. The modelled annual rate is 3.6%, against 2.8% historically, and global consumption rises from an estimated 28,200 TWh in 2025 to 33,600 TWh in 2030. The distinction prevents a strong signal from becoming an exaggerated one.S01

The supply system is not moving on the same clock. The IEA's worldwide connection evidence places more than 2,500 GW of renewable-generation, storage and large-load projects in queues or stalled processes. That is a QUEUED / OFFICIAL ESTIMATE across different categories, jurisdictions and stages; it is not operating capacity and cannot be read as one additive supply pipeline. A grid project can typically take five to fifteen years, compared with one to three for a data centre and one to five for wind or solar. Those ranges are indicative, yet they expose the governing mismatch: the asset that carries the power can have the longest lead time in the chain.S02

Money provides another signal, but only after its basis is kept straight. The February planning comparison says annual grid investment must rise about half from roughly USD 400 billion, implying about USD 600 billion by 2030 in real 2024 dollars at market exchange rates. The May tracking series uses real 2025 dollars and estimates nearly USD 450 billion in 2025, followed by further growth in 2026. These are not values to splice into a synthetic gap. The later report also warns that equipment and input-price inflation accounts for part of the rise, so capital spending is neither delivered network volume nor supplier revenue.S02S04S05

The first practical consequence is a change in the unit of decision. An announced generator, a signed power contract, a queue position or an investment allowance can each be useful evidence, but none of them alone makes a site power-ready. The useful record joins the load's realistic ramp, the generation and flexibility available at the node, the studies and upgrades needed, the owner of each cost, the delivery dates for scarce equipment, and the tests required before full output. That record can be compared across very different electricity worlds even when their headline numbers cannot. It turns the supercycle from a slogan into a sequence that a buyer can challenge.S07S12S13

In the 2000s, the phrase “global electricity market” concealed two different motions. Many advanced economies were entering a long period of flat or weak demand, with mature networks managed around incremental replacement and reliability. Emerging industrial systems were adding customers, factories and urban load, while much of Sub-Saharan Africa still faced the more basic problem of first connection. Those starting points shaped planning institutions and investment habits. A network built for slow aggregate change can be reliable and still lack headroom at the precise place a large new load wants to arrive; a system adding access can need generation, transmission, distribution and affordable service at once.S11S18S22

The 2010s then changed the generation side faster than the carrying system. IRENA's series records worldwide renewable capacity rising from about 2,021 GW at the end of 2016 to about 5,149 GW at the end of 2025. For most countries those statistics represent capacity installed and connected, but they remain nameplate capacity rather than annual generation, firm deliverability or local network headroom. The build proved that modular generation could scale rapidly. It did not prove that every network, market rule, transformer fleet or planning process had been rebuilt around bidirectional flows, variable output and much larger connection volumes.S06

By the early 2020s, demand itself had begun to turn. US net energy for load grew by an average 1.7% a year in 2020–2025 after just 0.1% in 2005–2019. China continued to add demand on a much larger base, while India and Southeast Asia remained faster-growing systems. The causes also converged: industrial policy, electric vehicles, cooling, heat pumps and data centres all sought network capacity. This is why the present cycle is not simply another renewables wave. It places old and new generation, fast loads, distribution constraints, bulk transmission, flexibility and cost allocation inside the same delivery problem.S01S16S18

Networks entered that convergence with inherited constraints. The European Commission says the region has about eleven million kilometres of electricity networks and estimates that roughly forty per cent of distribution assets are more than forty years old. China is investing in grid and distribution service alongside its exceptional generation build. India's public plan separates additions already commissioned from those under construction, under bidding and still planned, and it records the right-of-way, forest-clearance, land and contractual problems that affected earlier delivery. These are different systems, but the common lesson is that a generation total says little about the condition or readiness of the path between plant and load.S08S10S17

The 2026–2030 period is therefore a build test, not an endpoint. Forecast demand and announced capital create pressure; commissioning records reveal whether the system answered it. In the 2030s, the confirming evidence will be less glamorous and more useful: transmission and distribution assets entering service, connection studies becoming constructed works, equipment arriving without repeated slippage, large loads ramping as forecast, renewable output moving with less avoidable curtailment, reliability holding and access becoming affordable service. If those outcomes arrive, the supercycle will have rebuilt a system. If they do not, it will have produced a large inventory of plans, orders and queue positions.S09S10S12S25

The new demand is plural. Industry adds large process and manufacturing loads, sometimes in clusters created by trade and industrial policy. Cooling grows with income, appliance ownership and hotter conditions. Transport adds vehicles, depots and charging whose peaks depend on behaviour and tariffs. Electrified heat changes seasonal and local distribution demand. Digital infrastructure adds concentrated facilities with high power density and unusually short commercial schedules. The categories overlap in national statistics, so the honest picture is a causal map rather than a neat stack whose pieces can all be added without adjustment.S01S16S20S21

The IEA attributes forty-nine per cent of additional global electricity demand through 2030 to buildings, with cooling and heat pumps among the important drivers inside that subtotal. Transport contributes more than one-tenth of the increase. Those shares describe additions, not total demand, and the building components must not be counted again outside their parent. Industry and digital infrastructure complete the picture without a fabricated residual. For system planning, the engineering profile matters as much as the annual energy: coincidence, ramp speed, voltage, power quality, location and willingness to accept curtailment determine which part of the network must change.S01

AI belongs inside that map, not above it. All data centres used an estimated 415 TWh in 2024, about 1.5% of global electricity. The IEA base case reaches roughly 945 TWh in 2030, still below three per cent of world consumption. Data centres account for less than one-tenth of global electricity-demand growth in the base case, but around half of the increase in the United States through 2030. That is the defining contrast: modest in the world total, potentially dominant in a particular country, utility territory or substation catchment. The category also includes non-AI computing, even though accelerated servers drive much of the projected growth.S01S03

The range around digital demand is wide enough to change local build decisions. In the IEA's conditional 2035 cases, data-centre electricity use is about 700 TWh under Headwinds, about 970 TWh with high efficiency and more than 1,700 TWh in Lift-Off. These are not assigned probabilities. They show how adoption, server efficiency, utilisation and location can alter the load. A prudent network or site plan should therefore test the low and high cases, require evidence for the ramp, and avoid financing an irreversible solution against one unexamined forecast. Flexibility, staged connection and credible cancellation rights can be more valuable than false precision.S03S19

Once several fast loads arrive together, the constraint moves through the physical chain. A bulk system may have enough annual generation but insufficient transfer capacity. A transmission node may have headroom while the distribution feeder, fault level or transformer does not. Equipment can be specified before permits or cost ownership are settled, then sit outside the critical path while a study is redone. Non-firm service can bring a project forward, but the curtailment right changes its operating risk and economics. The build-out is therefore an exercise in coordination across assets and rules, not a contest to announce the largest generation or load number.S02S07S13

Three electricity worlds make the comparison legible. Mature systems are seeing demand return to networks built during a slower era. Fast-growth industrial systems must add enormous volumes of demand, generation and networks together. Access-constrained systems still need first connection, dependable service and affordability, often with limited fiscal space and a high cost of capital. This is not a ranking. Each world is tested on the same six questions: what drives demand, what generation is being added, what network work is required, what flexibility exists, whether reliability or access improves, and who ultimately pays.S01S08S11S22

In the United States, national demand growth has returned and data-centre concentration makes local capacity, generation adequacy and cost shifting visible. FERC's 2026 show-cause action covers six organised markets but explicitly recognises regional differences in definitions, studies, operating requirements, upgrades and cost allocation. In Europe, electrification meets ageing distribution networks and cross-border planning. Great Britain's former queue exceeded 700 GW and included waits of up to ten years, but that old queue was not demand or a needed build programme. Reform reordered projects; only realised connection times will show whether the clock changed.S07S08S09S18S19

China and India reveal a different scale. China is expected to account for nearly half of the global increase in electricity demand to 2030 while continuing very large renewable and network additions. High national build and utilisation do not eliminate provincial or node-level constraints. India has the fastest demand growth among the large regions in the IEA comparison, and its transmission plan provides unusually useful stage evidence: by March 2024, some inter-regional additions were commissioned, others were under construction or bidding, and another tranche had not yet entered those stages. That separation is more decision-useful than one combined plan total.S01S06S10S17

Southeast Asia combines industry, rising cooling demand and digital infrastructure with national-grid expansion and cross-border ambition. Its modelled average demand growth of 5.3% in 2026–2030 is close to India's fast-growth world, but its delivery institutions span several countries. The Middle East adds cooling, desalination and industrial projects, yet Gulf interconnection, import-constrained systems and conflict-affected markets should not be compressed into one capacity story. In both regions, a headline generation plan can be real while finance, interconnection, fuel, storage or network synchronisation still owns the in-service date.S01S20S21

Africa keeps the universal thesis honest. In 2024, an estimated 655 million people worldwide still lacked electricity access, more than 560 million of them in Sub-Saharan Africa. A connection can exist and still be unaffordable or unreliable, so the outcome cannot be measured only in installed generation. Grid extension, mini-grids and stand-alone systems each have a role, with concessional finance and utility viability affecting delivery. The supercycle is incomplete if it serves new industrial and digital loads while leaving first connection, distribution quality and basic affordability outside the frame.S11S22

Latin America adds hydro dependence, transmission distance and climate exposure. Brazil's system institutions forecast average load growth of 3.4% in 2025–2029, while the PDE 2035 reference case estimates roughly BRL 120 billion of transmission investment through 2035. The first is a forecast and the second a planning need, not an award or construction result. Curtailment can arise from transmission congestion, operational conditions, contract rules and changing hydrology. Comparing the region well therefore means following load, lines, storage, dispatch and drought together rather than assigning every lost unit to one missing asset.S23S24

A credible power-ready date begins with a lifecycle. “Planned” means an intention; “contracted” means commercial obligations exist; “financed” means capital has been committed on stated terms; “under construction” means physical delivery has begun; “connected” means the network interface has been energised under an agreement; and “operating” means the asset or load is actually producing or consuming. These states are not interchangeable. The Australian Energy Market Operator makes the distinction concrete by tracking projects from application through registration and commissioning to full output, with modelling and performance standards as explicit parts of the process.S12S13

AEMO's March 2026 scorecard recorded 67.3 GW of generation and storage progressing through the National Electricity Market connection process, compared with about 73 GW of existing capacity excluding consumer energy resources. The pipeline had grown by a third in a year. None of that makes the full pipeline operating capacity. The value of the scorecard is its movement between states: applications can be approved, projects can register, commissioning can advance, and only then can full output be recorded. India offers a comparable discipline from a government plan, separating commissioned transmission from construction, bidding and later perspective stages.S10S12

The worldwide queue estimate needs the same discipline. Its more-than-2,500-GW headline is a QUEUED / OFFICIAL ESTIMATE that mixes renewable generation, storage and large loads at different stages; it is not operating capacity. The IEA's high-level work suggests that reform, grid-enhancing technology and non-firm arrangements could unlock substantial advanced-stage capacity, but the ranges are non-additive and omit some local technical constraints. Non-firm service can accelerate access while permitting curtailment. The honest funnel therefore ends with “connected outcome not yet observed.” No global conversion rate should be invented from a changing collection of queue definitions, duplicate projects, cancellations and reassessments.S02

One disclosed US case shows how the gates separate. A Talen filing described power-sale options associated with the Susquehanna data-campus transaction in increments up to a 960 MW contractual ceiling. PJM reported no material transmission impact only up to 480 MW, including 300 MW previously authorised; transfer above the study boundary produced deliverability violations that required upgrades. FERC rejected the amended interconnection agreement without prejudice in November 2024. The cited records do not establish completed network works, commissioning or operating load at the ceiling. The right conclusion is not that the project failed or succeeded, but that the public evidence stops before power-ready operation.S14S15

Cost ownership can hold the same critical path. A network upgrade may benefit the connecting project, later users and system reliability in different proportions. The connecting party may accept a bespoke contribution, self-build work, staged service or curtailment; the network company may recover allowed investment from tariffs; regulators may need to decide how costs move between existing and new customers. FERC's large-load action, the European Commission's grid programme, Ofgem's transmission allowance and India's plan all show public institutions working on this boundary. None of those policy or allowance records, by itself, is completed construction.S07S08S10S25

The practical diligence test is short enough to use. First, state the load and ramp rather than only the ultimate ceiling. Second, identify available supply and its physical node, not just a contractual clean-power claim. Third, name the study, voltage, upgrades and equipment. Fourth, record permits, land, modelling and performance standards. Fifth, show who pays and what can be curtailed. Sixth, separate ordered, financed, construction, connected and operating evidence. Finally, ask which dependency has the least schedule evidence and the weakest owner. That dependency, not the largest megawatt on the page, owns the date.S02S07S13

The next five years should be watched as a conversion problem. Demand forecasts need to become measured load without wholesale cancellation. Queue reform needs to produce completed studies, constructed upgrades and shorter realised connection times. Capital spending needs to translate into lines, substations, transformers and digital control rather than mainly higher unit prices. Flexible service needs to reveal its curtailment frequency and commercial value. In access-constrained systems, connection counts need to become reliable and affordable service. Each signal has an owner, a time window and a condition that would weaken the argument.S01S02S05S11

Demand is the first falsification test. The IEA outlook is a central forecast, and data-centre scenarios span a wide range. If industrial projects, electric-vehicle uptake or data-centre campuses are cancelled or ramp much more slowly, networks may gain time and some local scarcity could ease. That would not erase ageing assets or access needs, but it would reduce the urgency of capacity built solely for one high case. Conversely, measured load growth across several end uses would strengthen the whole-system interpretation because no single driver could explain the outcome.S01S03S18S23

Delivery is the second test. Great Britain's reordered queue, AEMO's state-by-state scorecard and India's staged transmission plan provide templates for evidence that can move. The relevant changes are not a smaller headline alone, but more projects clearing studies, reaching construction, commissioning and full output on disclosed dates. FERC's action will matter when tariff changes are final, studies become more predictable and cost allocation ceases to reopen the schedule. If reform compresses realised connection time without the expected rise in capital, the present thesis weakens in a useful way: institutions and flexibility will have substituted for part of the build.S07S09S10S12

Physical productivity is the third test. The World Energy Investment series warns that spending can rise because transformers, cables, metals, labour and project complexity cost more. A stronger result would pair money with completed circuit length, transformation capacity, connection throughput and reliability. Ofgem's initial electricity-transmission allowance, European investment needs and Brazil's reference scenario are inputs to that test, not answers. When public reporting distinguishes initial allowances, contingent funding, awards, construction and commissioned assets, customers can see whether the capital cycle is expanding the system or compensating for inflation and delay.S04S05S08S24S25

The final test is distributional. A system can connect a prestigious large load while shifting excessive costs to existing customers, or add generation while access and service quality remain poor. Cost-allocation decisions, curtailment terms, reliability outcomes and affordability therefore belong beside megawatts and dollars. The confirmation signal is a portfolio of commissioned networks, connected loads, maintained reliability and wider access, with costs assigned through rules that remain workable as the system grows. That is a more demanding measure than investment, and the only one that shows whether the supercycle delivered usable electricity.S07S11S17S22