Energy Futures: Backgrounder

Future of Energy: Research Foundation

Cheap electrons reshape the energy future to 2046

October 2026

Executive summary

Plan for an electricity-centred energy system in which cheap solar, wind and batteries supply most new energy, while grids, finance and politics, not generation cost, set the pace. In 2025, renewables overtook coal as the world's largest source of electricity (33.8% vs 33.0%). Solar met about 75% of demand growth, and fossil generation fell slightly (−0.2%) outside a recession for the first time (Ember) [High]. Fossil fuels still supply about 86% of primary energy, every fuel hit a record, and energy CO2 rose 0.4–1.1% depending on the dataset (EI via Carbon Brief; Global Carbon Budget) [High]. The system is changing quickly at the margin and slowly in total. Both facts are true at once, and any plan for 2026–2046 has to hold both.

Five findings anchor the planning:

• The official forecasters have been too pessimistic about modular technology. The IEA has underestimated solar in every World Energy Outlook since 2002 (pv magazine USA). Experience-curve methods, back-tested on more than 50 technologies, have predicted better (Way et al.) [High]. The same record shows big site-built projects (Western nuclear, CCS, hydrogen, fusion) arriving later and costing more than forecast.

• AI is a regional shock, not the main global driver. Data centres used about 415 TWh in 2024 (about 1.5% of world electricity). The IEA's central case reaches about 1,200 TWh by 2035, still under 10% of global demand growth, although data centres drove about half of US growth in 2025 (IEA) [Medium].

• Grids are the binding constraint. More than 2,500 GW of projects sit in connection queues worldwide. Grid spending (about USD 400 bn/yr) is growing at less than half the pace of spending on new generation (IEA Electricity 2026) [High].

• The Global South's leapfrog is under way, but the poorest are being left behind. Pakistan, Africa and India are adopting solar because it is cheapest. However, Africa receives about 2% of clean-energy investment while holding about 20% of the world's population, and it pays 2–3 times higher capital costs (IEA WEI 2025) [High].

• 1.5°C overshoot is now the base case. Human-induced warming crosses 1.5°C around 2030 (Forster et al.) [High]. Damages and tipping risks therefore rise through the planning window whatever is done. Fossil air pollution, meanwhile, kills millions of people every year now.

The system that delivers the broadest benefit has four parts:

• electrified end uses;

• solar-led generation firmed by batteries, hydro, geothermal and, where institutions support it, nuclear;

• heavy and early grid investment, including flexible demand from data centres and EVs;

• cheap capital routed to emerging economies.

Planning anchor20352046ConfidenceGlobal electricity demand vs 2025+29% (BNEF) to +40% (IEA)about +70% (interpolated from 2050 outlooks)MediumRenewables' share of generation>50% (IEA STEPS)about 60–65% (interpolated)MediumOil demandPlateau about 2029–2030 at about 102–105 mb/d; OPEC disputes thisFlat-to-declining; 2050 range 75–124 mb/d across outlooksMedium, contestedData-centre electricityabout 1,200 TWh (range 700–1,700)1,500–5,000 TWh (derived extrapolation)Medium / LowWarmingHuman-induced 1.5°C crossed about 2030Current trajectories point to 2.2–2.9°C by 2100High / Medium

Confidence tags used throughout:

• High means several independent primary sources agree.

• Medium means one primary source, or sources that differ by definition.

• Low means a secondary source, a single research thread, or the researchers' own extrapolation.

The 2026 baseline

The world is electrifying fast at the margin while total fossil use is still at a record. That is a plateau, not yet a decline.

Energy supply. Total energy supply passed 600 EJ in 2025 (+1.7% per the Energy Institute; +1.3% per the IEA) (EI; IEA GER 2026). Fossil fuels supplied 86.2% of primary energy, a record low share but a record absolute volume (DieselNet/EI). DNV describes the same system as roughly 80/20 fossil to non-fossil (DNV). The gap is an accounting convention for non-combustion sources, so cross-source comparisons need care [Medium].

Electricity. Demand reached 31,779 TWh (+2.8%). Low-carbon generation grew by 887 TWh, more than the 849 TWh rise in demand (Ember) [High]. Electricity is still only about 21–22% of final energy (IEA WEO 2025).

Emissions. Sources disagree on 2025:

SourceMeasure2025 changeLevelIEAEnergy-related CO2+0.4%(not stated)Global Carbon BudgetFossil CO2+1.1%38.1 GtEnergy InstituteEnergy-sector CO2+1.1%35.8 Gt

The differences come from different boundaries [High that it is a plateau; Medium on direction]. The geography of the increase has inverted. China's power-sector CO2 has been falling since early 2024, and China and India cut coal power together for the first time since 1973 (Carbon Brief). Meanwhile US emissions rose 3.2% as US coal power rose 13% (EI). The remaining 1.5°C budget is about 170 Gt CO2, roughly four years of current emissions (GCB).

Money. Global energy investment was USD 3.3 tn in 2025: about USD 2.2 tn clean and about USD 1.1 tn fossil. Solar (USD 450 bn) was the single largest item (IEA WEI 2025). The total rose to a record USD 3.4 tn in 2026 (IEA WEI 2026) [High]. Data-centre investment (about USD 580 bn) now exceeds global oil supply investment (USD 540 bn) (IEA WEO 2025).

China dominates the build-out. In 2025 China accounted for:

• 378 GW-DC of solar (58% of the world total) and 119 GW of wind (72%);

• nearly 75% of EV production;

• more than 80% of battery cells (Ember; IEA GEVO 2026).

Clean energy was 11.4% of China's GDP and about 37% of its GDP growth in 2025 (Carbon Brief/CREA) [Medium].

2025 electricity generation by source (Ember):

SourceTWhShareChangeCoal10,47633.0%−0.6%Gas6,91921.8%+0.5%Hydro4,43714.0%flatNuclear2,8128.9%+1.3%Solar2,7788.7%+30%Wind2,7158.5%+8.2%

All renewables together supplied 10,730 TWh (33.8%). The Energy Institute's equivalent split is 33.4% renewables vs 32.6% coal, a small definitional difference (Connaissance des Énergies/EI).

source,generation_2025_TWh,share_2025_pct,change_2025_pct
coal,10476,33.0,-0.6
gas,6919,21.8,0.5
hydro,4437,14.0,
nuclear,2812,8.9,1.3
solar,2778,8.7,30
wind,2715,8.5,8.2
renewables_total,10730,33.8,

What the forecasters project, and their track record

The major outlooks agree on direction but split sharply on oil and gas. Those split by policy assumptions far more than by technology. All of them have electricity demand growing 2.5–4 times faster than total energy, solar as the largest growth source, coal in decline, and "current trajectory" warming of 2.2–2.9°C.

Outlook (edition)ApproachElectricity demandOilWarming by 2100IEA CPS (WEO 2025)Policies frozen; solar additions held at about 540 GW/yr+~40% by 2035113 mb/d in 2050~2.9°CIEA STEPS (WEO 2025)Stated policies+~40% by 2035Plateau about 102 mb/d around 2030, then slow decline2.5°CIEA NZE (WEO 2025)Normative net zero>+50% by 2035Steep declinePeak about 1.65°C around 2050BNEF ETS (NEO 2026)Economics only, no new policy+29% by 2035, +69% by 2050Peak about 2029; early-2000s levels by 20502.4°CBNEF NZS (NEO 2026)Net zeroDominant carrier by 2037—1.73°C (peak 1.81°C in 2049)DNV ETO 2025Single best estimate+120% by 2060Fossil demand falls before 20302.2°CExxonMobil (2026)Company view+~65% by 2050>100 mb/d through 2050CO2 −20% by 2050OPEC WOO 2026Producer view—124.1 mb/d in 2050—

Sources: IEA STEPS; IEA CPS; IEA overview; BNEF NEO 2026; DNV; ExxonMobil; OPEC via Energy News Beat. The OPEC figures come from a secondary summary.

Policy-tracking outlooks. These measure what current and pledged policies would deliver, rather than modelling the energy system:

TrackerCurrent policiesPledgesUNEP Emissions Gap 2025up to 2.8°C2.3–2.5°C if NDCs are fully implementedClimate Action Tracker2.6°C2.2°C with net-zero pledges included

Sources: UNEP; CAT.

The US government outlook has gone dark. The EIA cancelled its 2025 International Energy Outlook after staff losses; the latest edition is IEO2023 (EIA).

Track record. Official agencies have under-forecast exponential technologies and over-forecast "big-build" ones [High]:

ForecastProjectionOutturnIEA, early 2016~50 GW/yr of solar, levelling off76 GW in 2016 (ScienceAlert); ~600–650 GW-DC in 2025IEA, 2000 vintageEU renewable power share of 5.2% (reference) / 10% (ambitious)34% in 2019 (BU ECI)EIA, 2000US wind in 2020About 4× the high case and 16× the reference case (BU ECI)IEAGlobal solar capacity +6%/yr over 2020–2050+35%/yr achieved in 2010–2020 (BU ECI)DNVHydrogen at 4.8% of 2050 energyCut to 3.5% over three editions (DNV)RMI, 2023Fossil power plateaus in the late 2020sFossil generation −0.2% in 2025 (RMI)

Way, Ives, Mealy and Farmer (Joule, 2022) found that most energy-economy models have historically overestimated clean-tech costs and underestimated deployment. In their analysis a fast transition saves "at least USD 12 trillion" against continued fossil use (University of Oxford) [Medium; full paper not re-read].

The structural assumption of flat additions persists in the IEA's policy-baseline scenario. The IEA's CPS freezes solar additions at about 540 GW/yr, already below 2025's actual installations. Treat it as a floor, not a forecast.

source,scenario,metric,warming_C_low,warming_C_central,warming_C_high
DNV ETO 2025,best estimate,2100,,2.2,
BNEF NEO 2026,Economic Transition,2100 (67th pct),,2.4,
BNEF NEO 2026,No transition counterfactual,2100,3.6,,
BNEF NEO 2026,Net Zero,2100,,1.73,
BNEF NEO 2026,Net Zero,peak 2049,,1.81,
IEA WEO 2025,STEPS,2100,,2.5,
IEA WEO 2025,CPS,2100,,2.9,
IEA WEO 2025,NZE,peak ~2050,,1.65,
UNEP EGR 2025,current policies,2100,,2.8,
UNEP EGR 2025,full NDCs,2100,2.3,,2.5
Climate Action Tracker,current policies,2100,,2.6,
Climate Action Tracker,pledges + net-zero targets,2100,1.8,2.2,2.8
NGFS Phase V,Current Policies,2100,2.75,,3.0

Solar, wind, batteries and EVs

Solar and batteries keep falling down their learning curves. The live risks are integration, trade walls and policy, not cost [High].

Solar costs. Module prices bottomed at about USD 0.09/W at end-2025 and rebounded to about USD 0.115/W in January 2026 (ITRPV via PV Tech; pv magazine). The rebound came from Chinese "anti-involution" consolidation and the removal of export rebates. For comparison, modules cost USD 106/W in 1976 (Our World in Data).

• Learning rate: 26% cost fall per doubling of cumulative output over 1976–2025 (ITRPV), against about 20% in the academic literature summarised by Our World in Data.

• Generation cost: utility solar averaged USD 0.043/kWh in 2024, 41% below the cheapest fossil option (IRENA).

• Deployment: a record 664 GW-DC in 2025, but SolarPower Europe expects a first contraction in more than 20 years in 2026 (−8% to 612 GW). The cause is China's shift from fixed tariffs to auctions (SolarPower Europe).

Wind is cheapest onshore (USD 0.034/kWh, IRENA). Offshore is the laggard: 9.3 GW was added in 2025, and UK offshore costs are 69% higher than five years ago (GWEC; BNEF) [High].

Batteries.

• Packs: the average lithium-ion pack cost USD 108/kWh in 2025. Stationary-storage packs fell 45% to USD 70/kWh, and LFP packs were USD 81/kWh (BNEF).

• Complete 4-hour storage systems: USD 73/kWh in China, against USD 219/kWh in the US (Energy-Storage.news). Planners should carry two cost curves: a "global/China" curve and a "tariff-walled" curve.

• 24-hour solar: Ember costs round-the-clock solar plus batteries (97% of hours) in Las Vegas at USD 104/MWh. That is below new coal (USD 118/MWh) and new nuclear (USD 182/MWh) (Ember) [Medium].

• Sodium-ion: CATL planned 2026 mass production of sodium-ion cells at about USD 70/kWh, against about USD 40/kWh for LFP cells. Delivery is not verified (Battery-Tech/Reuters) [Low].

EVs.

• Sales: about 20.7–21 million in 2025, roughly 25% of new car sales (IEA GEVO 2026; BNEF EVO 2026).

• Price: 70% of battery-electric cars sold in China were cheaper than the average combustion car.

• Forecasts: BNEF has EVs at 38% of sales in 2030, 52% in 2035 and 66% in 2040, though it has downgraded its outlook two years running on US and Chinese policy. The IEA has about 50% by 2035 [Medium].

Source conflicts to keep visible:

ItemFiguresWhy they differ2025 solar additions664 GW-DC (SolarPower Europe); 647 GW-DC (Ember); 655 GW (BNEF); 511 GW-AC (IRENA); ~600 GW (IEA)DC module rating vs AC grid rating (ratio ~1.3), and scope2025 wind additions165 GW grid-connected (GWEC); 178 GW delivered (GWEC supply-side); 170 GW (Wood Mackenzie); 159 GW (IRENA)Grid connection vs delivery2025 grid storage112 GW / 307 GWh (BNEF); 108 GW (IEA, secondary); ~250 GWh (Ember)Scope and estimation method2025 EV sales20.7 m (Benchmark); >20 m (IEA); 21 m (BNEF). China's domestic EV share: 55% (IEA) vs 64% (BNEF)Vehicle definitionsOil displaced by EVs, 20251.7 mb/d (IEA) vs 3.2 mb/d (BNEF, includes efficiency gains)Scope

Extrapolations (researcher's own, not a source forecast) [Low]. If historical learning rates hold:

• modules reach about USD 0.05–0.07/W by 2035 and USD 0.03–0.05/W by 2046;

• battery packs fall to about USD 70/kWh by 2035 and USD 45–55/kWh by 2046.

On the published side, BNEF projects solar costs −30% and storage costs −25% by 2035 (BNEF).

series,year,value,unit,source
pv_module_price,1976,106.09,USD/W,Our World in Data
pv_module_price,2019,0.377,USD/W,Our World in Data
pv_module_price,2025,0.09,USD/W (end-year),ITRPV 2026
pv_module_price,2026,0.115,USD/W (Jan FOB China),OPIS via pv magazine
utility_pv_installed_cost,2010,5283,USD/kW,IRENA
utility_pv_installed_cost,2024,691,USD/kW,IRENA
utility_storage_installed_cost,2010,2571,USD/kWh,IRENA
utility_storage_installed_cost,2024,192,USD/kWh,IRENA
li_ion_pack_avg,2025,108,USD/kWh,BNEF
li_ion_pack_stationary,2025,70,USD/kWh,BNEF
li_ion_pack_LFP,2025,81,USD/kWh,BNEF
bess_turnkey_global,2024,165,USD/kWh,BNEF survey 2024
bess_turnkey_global,2025,117,USD/kWh,BNEF survey 2025
bess_turnkey_4h_china,2025,73,USD/kWh,BNEF survey 2025
bess_turnkey_4h_china,2035,41,USD/kWh (forecast),BNEF

Firm clean power: nuclear, fusion, geothermal and long-duration storage

Firm clean supply grows through 2035 mainly from Chinese nuclear and pumped hydro. Enhanced geothermal is the fastest-moving Western story, and fusion is a post-2046 capacity option. "Firm" means power available on demand, regardless of weather.

Nuclear.

• Output: a record 2,702 TWh in 2025 per the World Nuclear Association. Ember reports 2,812 TWh; the two differ in scope (WNA).

• Pipeline: 80 reactors (87 GWe) under construction, 38 of them in China and none in the US (WNA) [High].

• Cost gap: Western new-build costs USD 12,300–17,570/kW, or USD 175–255/MWh (Lazard 2026). China builds at about USD 2,500/kW (WNA). Recent Western projects show the problem:

  • Vogtle: about USD 36 bn and 15 years for two units (POWER).

  • Flamanville 3: €23.7 bn (ANS).

  • Hinkley Point C: £35 bn in 2015 prices (Industrial Info).

• Projections: the IAEA's 2050 range is 561–992 GWe, from 377 GWe in 2024 (ANS).

• Small modular reactors: the only Western SMR under construction is OPG's Darlington BWRX-300. The first unit costs C$7.7 bn and targets grid connection at end-2030 (ANS).

• US policy versus delivery: the first US pilot reactor went critical in August 2026, after the executive-order deadline of July 2026 (Oklo).

Fusion shows real science but no electricity. The National Ignition Facility reached a target gain of 4.13 (8.6 MJ) in April 2025 (LLNL). ITER's deuterium-tritium operation has slipped to 2039 (Physics World). Commonwealth Fusion Systems targets first plasma in 2027 and has raised USD 4 bn (CFS). Helion no longer frames its goal as "net electricity" (Helion).

Geothermal. Fervo's Cape Station (Utah) began commercial operation in October 2026 with 33 MW, targets about 500 MW by 2028, and aims for USD 3,000/kW (Energy Global; Fervo Q1 2026). The IEA sees up to 800 GW worldwide by 2050 if costs fall 80% to about USD 50/MWh by 2035 (IEA) [Medium].

Long-duration storage. Pumped hydro passed 200 GW worldwide in 2025, with 243 GW under construction, mostly in China (IHA). Form Energy's iron-air factory makes about 2 GWh/yr, and its first 150 MWh system is due in 2027 (MIT Technology Review).

Hydrogen and CCS lag forecasts badly. Low-emissions hydrogen output is only about 1 Mt/yr, and just 9% of the 2030 project pipeline has reached a final investment decision (IEA). CCS operating capture capacity is 64 Mtpa according to the Global CCS Institute (GCCSI), but the Statistical Review reports 258.9 Mt captured in 2025 (EI summary). That conflict is unresolved and likely definitional.

Why firm power still matters. Even in a solar-led system, the modelling literature finds firm low-carbon resources cut fully decarbonised system costs by 10–62% (Sepulveda et al.). Storage lasting 10 hours or more cuts 100%-renewable system costs by about 32% (Dowling et al.).

Scenario ranges (researcher synthesis from the notes, not a published forecast) [Low–Medium]. Each cell shows slow / central / fast.

Technology20352046Cost signalNuclear (GWe in operation)440–470 / 500–540 / 560–600520–600 / 700–800 / 950–1,100Fleet-built $40–70/MWh; Western first-of-a-kind $175–255/MWhWestern SMRs1–3 GW operatingdepends on unit-to-unit learningDarlington ~C$20,000–25,000/kW (derived)Next-gen geothermal5–10 / 20–40 / 60–100 GW30–60 / 100–200 / 300–500 GWTarget $45–50/MWh by 2035Pumped hydro350–450 GW500–650 GW—Multi-day storage5–20 GW50–200 GW$1,900–2,500/kW todayFusion~0 (pilots <0.5 GW)0 / <2 / 5–20 GWnot demonstrated

AI and data-centre demand

AI will reshape where, how fast and with what fuel firm capacity is built, especially in the US. It is unlikely to dominate global electricity growth.

Size today. Data centres used about 415 TWh in 2024 (about 1.5% of world electricity): 45% in the US, 25% in China and 15% in Europe (IEA Energy and AI). Use grew 17–20% in 2025 (IEA June 2026 update). Estimates for 2025 conflict by definition: the Energy Institute reports 788 TWh, while BNEF reports 500 TWh (84 GW) (EI; BNEF) [Medium].

Projections.

• IEA Base case: about 945 TWh in 2030 and about 1,200 TWh in 2035, within a 2035 range of 700 TWh (Headwinds) to more than 1,700 TWh (Lift-Off). That is under 10% of global demand growth, but nearly half of US growth (IEA).

• BNEF: data centres at 5.4% of global demand by 2035 (BNEF NEO 2026).

• DNV: AI at 3% of electricity by 2040 (DNV).

• The full spread: more than 50 studies reviewed by DNV span 210–7,900 TWh for 2030, the widest of any uncertainty in the system (Pielke).

Hype versus reality.

• The four largest hyperscalers' capex roughly doubled, from about USD 370 bn (2025) to about USD 705 bn guided for 2026 (Om Malik) [Medium; secondary].

• The Stargate Abilene expansion was cancelled in March 2026 over financing, power delays and revised demand forecasts (TechRadar).

• US utilities' data-centre forecasts appear about 40% overstated when checked against chip shipments (Grid Strategies).

• Texas's ERCOT large-load queue went from 63 GW to more than 410 GW in about 15 months (Latitude Media; RTO Insider).

• Queues are statements of intent, not forecasts.

Supply is going mostly to gas in the near term.

• GE Vernova's gas-turbine backlog plus reservations reached 100 GW in Q1 2026 (SEC filing).

• Meta contracted 7.5 GW of new gas plants in Louisiana (Fortune).

• The IEA still expects renewables to meet nearly half of data-centre demand growth to 2030 (IEA).

Local stress is real.

• PJM, the largest US grid operator, saw its capacity auctions hit the price cap ($329–333/MW-day). Its market monitor attributes about USD 6.3 bn of about USD 16.4 bn in 2028/29 capacity costs to data centres (Industrial Info).

• Data centres use 23% of Ireland's electricity (Irish Times).

• 7 in 10 Americans would oppose a nearby AI data centre (Stateline).

The flexibility opportunity. If data centres accept 0.25–1% curtailment, about 76–126 GW of new load fits on existing US grids (Duke via Utility Dive) [Medium].

Two signals that keep AI in proportion.

• Efficiency: leading AI chips improve about 40%/yr in energy efficiency, while frontier compute grows 4–5×/yr (Epoch AI).

• Other demand: IEA projects EV electricity above 1,500 TWh by 2035, larger than data centres' ~1,200 TWh (IEA GEVO 2026). Air conditioning adds about 500 GW of peak demand by 2035 (IEA WEO 2025).

2046 (derived extrapolations, not any source's forecast) [Low]. Data-centre use plausibly reaches 1,500 TWh (saturation), 2,500–3,000 TWh (central) or about 5,000 TWh (high). For AI to drive most global growth, it would need about 7,000–9,000 TWh.

year,case,basis,TWh,source
2024,actual,consumption,415,IEA Energy and AI
2024,actual,generation incl. losses,460,IEA Energy and AI
2025,estimate,facility incl. broader scope,788,Energy Institute 2026
2025,estimate,IT-load basis,500,BNEF NEO 2026
2025,derived (415 x 1.17-1.20),consumption,485-500,notes (derived)
2030,central (IEA Base),consumption,945,IEA Energy and AI
2030,central (IEA Base),generation incl. losses,1000,IEA Energy and AI
2030,low (researcher range),consumption,700,notes (derived)
2030,high (researcher range; McKinsey/RAND GW converted),consumption,2300,notes (derived)
2035,low (IEA Headwinds),consumption,700,IEA Energy and AI
2035,High Efficiency,consumption,970,IEA Energy and AI
2035,central (IEA Base),consumption,1200,IEA Energy and AI
2035,high (IEA Lift-Off),consumption,1700,IEA Energy and AI
2035,low (IEA Headwinds),generation incl. losses,790,IEA Energy and AI
2035,central (IEA Base),generation incl. losses,1300,IEA Energy and AI
2035,high (IEA Lift-Off),generation incl. losses,2000,IEA Energy and AI
2046,low (derived),consumption,1500,notes (derived)
2046,central (derived),consumption,2500-3000,notes (derived)
2046,high (derived),consumption,5000,notes (derived)

Grids, storage, flexibility and electrification

Grids are the slowest-moving part of the system and the most likely binding constraint over 2026–2035. The fastest fixes avoid building new lines [High].

Investment. Grid spending was about USD 400 bn in 2025. It must rise about 50% by 2030 (IEA Electricity 2026), and above USD 600 bn/yr on net-zero paths (IEA). A secondary summary of IEA WEI 2026 reports about USD 550 bn in 2026 (Mercom) [Low].

Bottlenecks.

• Equipment: large transformers take up to 4 years and DC cables more than 5 years to procure. Lead times have almost doubled since 2021 (IEA).

• US build rate: the US built 536 miles/yr of 345 kV+ lines in 2020–24, against about 5,000 miles/yr needed (Grid Strategies).

• US queues: more than 2,060 GW of projects are waiting to connect. Only about 13% of historical requests reached operation, and median wait times now exceed five years (LBNL).

• China's counter-example: more than 62,000 km of ultra-high-voltage lines built (Global Transmission), and a CNY 4 tn (about USD 570 bn) grid plan for 2026–2030 (Enerdata).

BottleneckBypass leverSize of leverNew lines take 5–15 yearsReconductoring existing lines with advanced conductors>80% of US interzonal need to 2035, at less than half the cost (GridLab/Berkeley)Connection queuesNon-firm connections plus grid-enhancing technologies1,200–1,600 GW of stalled projects globally (IEA)Peak capacityFlexible large loads76–126 GW in the US (Duke)Evening ramp after sunsetCo-located batteries (~275 days to build)California batteries met 36% of evening demand (July 2026); South Australia 40% (Nov 2025)Household distributed resourcesVirtual power plants (VPPs)37.5 GW in North America vs a DOE target of 80–160 GW by 2030

Sources for the last two rows: WINS Solutions; RenewEconomy; Wood Mackenzie.

Reliability is an engineering and governance problem, not a physics limit. The 2025 Iberian blackout was traced to voltage-control failures, not to renewables as such. ENTSO-E's 22 recommendations, all deployable now, focus on reactive power and inverter "ride-through" settings (pv magazine).

Oversupply is the growing operational issue. EU negative-price hours doubled in one year, from 593 (Q1 2025) to 1,223 (Q1 2026) (pv magazine).

Electrification and the efficiency dividend. Only 37% of primary energy becomes useful energy. EVs and heat pumps are 2–4× more efficient than the combustion equipment they replace (RMI). This is why electrified pathways need less total energy.

Electrification levels. Electricity's share of final energy is 21% globally, forecast to reach 24% by 2030 (IEA). For China, the IEA reports above 27% (2025) while Ember reports 32% (2023), a definitional conflict (Ember). US and EU shares have stagnated since 2010. Heat pumps now outsell gas boilers in Germany and gas furnaces in the US (IEA).

Who pays for the grid. In Pakistan, cheap rooftop solar cut grid sales 11% in three years while fixed capacity payments of about Rs 2 trillion/yr stayed. That pushed tariffs up and left poorer households paying more: solar adoption is 88% among high-spending households and 38% among low-spending ones (The News; Payne Institute) [Medium]. Tariff design, not technology, decides whether this becomes a utility "death spiral".

The cost of delay. The IEA's Grid Delay Case adds about 60 Gt of cumulative CO2 over 2030–2050 (IEA).

Energy access and the Global South

The Global South is already leapfrogging to solar on household economics. The binding constraints are finance, utility solvency and reaching the poorest.

How many people lack access.

SourceWithout electricityWithout clean cookingUN SDG7 tracking (2023 data)666 million (~85% in Sub-Saharan Africa)2.1 billionIEA WEO 2025~730 millionnearly 2 billion

The counts differ by definition and year (SEforALL; IEA) [High on scale; Medium on exact count].

Progress is far too slow. Electricity access improves by about 19 million people a year net, against about 136 million a year needed for universal access by 2030. In Sub-Saharan Africa, net progress was only about 5 million in 2023, because population growth absorbed most new connections (Ecofin/IEA).

The cost is small. Universal access costs about USD 23 bn/yr for electricity (to 2035) plus USD 4 bn/yr for clean cooking (to 2040). That is under 1% of global energy investment (IEA) [Medium].

Mission 300 shows why counting matters. The World Bank and African Development Bank's programme claims 50 million connections (Ecofin). An ODI audit found much of that figure is legacy pipeline counted regardless of approval date (ODI). Treat the claim as an upper bound [Low].

The distributed solar boom.

• Pakistan: about 50 GW of panels imported by August 2025, at about USD 0.08/W (Payne Institute).

• Africa: imported 15 GW of panels in the 12 months to June 2025 (+60%). Ember expects a record 17 GW installed in 2026, about 75% of it distributed (pv magazine; SolarQuarter/Ember) [Medium; secondary].

• Off-grid kits: a record 10.2 million sold in 2025, half of them pay-as-you-go. Sales of productive-use appliances such as pumps and fridges fell (GOGLA).

India added about 38 GW of solar in 2025, and coal generation fell about 3% (Ember; CREA). It also set the most transferable price point: solar with 4-hour storage at INR 2.86/kWh (about 3.3 US¢) (pv magazine) [High].

The cost ladder has inverted. Solar now sits at the cheap end almost everywhere:

OptionCost (US¢/kWh)SourceNigeria diesel generator~22 (derived)Arise/EmberPakistan grid, high-use customers10–18PayneSoutheast Asia average electricity cost~12IEA SEANigeria panel only, no battery~11 (derived)Arise/EmberPakistan rooftop solar~5PayneBrazil average solar (2024)4.3TaiyangNewsIndia solar + 4-hour storage3.3pv magazineSaudi utility solar bid1.29MEED

Finance is the obstacle, not technology.

• Cost of capital: utility solar in Kenya, Senegal and South Africa faces 8.5–11%, against 4.7–6.4% in North America and Europe. Contracts are in hard currency, and only two African countries are investment-grade (IEA).

• Southeast Asia gets 2% of global clean spending while holding 9% of the world's population. Its coal plants average under 15 years old (IEA SEA).

• Clean cooking is the hardest gap. Africa's deficit is still growing, and the IEA's roadmap needs a 7-fold acceleration (IEA).

The costs of climate change and pollution

Pollution costs are large, immediate and well measured. Climate damages are large, rising and contested in size. A scenario should carry both, with ranges.

Warming so far.

• 2024 was 1.55°C (WMO) to 1.60°C (Copernicus) above pre-industrial, and 2025 was 1.43–1.47°C (WMO; Copernicus).

• Human-induced warming was 1.37°C in 2025 and is rising at a record 0.27°C per decade. It will reach 1.5°C around 2030 (Forster et al.) [High].

• The two published 1.5°C carbon budgets (130 Gt from Forster et al.; 170 Gt from the Global Carbon Budget) both run out before 2030.

Damage estimates are contested.

• The IPCC position: limiting warming to 2°C costs "a few percent" of 2050 GDP, and its benefits exceed costs in most of the literature (IPCC) [Medium, as assessed by the IPCC].

• A retracted headline study: the widely cited Kotz et al. (2024, Nature) "19% income loss by 2049" paper was retracted in December 2025 over data errors. A revised version, not yet peer-reviewed, widens the range to 6–31% (Nature retraction). Do not treat any figure from it as settled [Low].

• Central-bank scenarios depend on it: the NGFS Phase V damage function relied on Kotz and is being revised in Phase VI (NGFS).

• The social cost of carbon: a working paper puts it at USD 1,065/t (Bilal & Känzig), against the EPA's USD 190/t. US agencies were barred from using the EPA figure in May 2025 (E&E News) [Low–Medium].

Cost categoryMagnitudeYearConfidenceAir-pollution deaths, all sources7.9 million (State of Global Air)2023HighFossil-fuel air-pollution deaths2.52 million from outdoor fine particles (Lancet Countdown) vs 5.13 million incl. ozone (Lelieveld, BMJ); methods differ2022 / 2019Medium (conflict)Fossil subsidiesUSD 725 bn explicit + USD 6.7 tn implicit (unpriced pollution and climate) (IMF)2024MediumHeat546,000 deaths/yr; USD 1.09 tn labour lost (WHO/Lancet)2012–21 / 2024MediumInsured natural-catastrophe lossesUSD 107–129 bn (2025); USD 137–147 bn (2024) (Artemis)2024–25HighAdaptation finance gapneeds USD 310–365 bn/yr by 2035 vs USD 26 bn delivered (UNEP)2023High

Methane is the cheapest near-term lever. Fossil methane exceeds 120 Mt/yr. About 70% is avoidable with existing technology, about 35 Mt of it at no net cost, which would avoid about 0.1°C of warming by 2050 (IEA).

Tipping points break the trend lines. Warm-water coral reefs have passed their tipping point (central estimate 1.2°C). Ice sheets and permafrost cluster around 1.5°C, and the Atlantic overturning circulation (AMOC) and Amazon around 2°C (Global Tipping Points). AMOC timing is contested: statistical methods give 2037–2109 (Ditlevsen), while 34 climate models show weakening but no collapse this century (Baker et al.). Each century above 1.5°C adds about 40 cm to the sea-level commitment for 2300 (Schleussner).

Clean tech has costs too, but of a different kind. Its footprint is mainly supply-chain concentration (72% average share for the top mineral refiner) and future waste (IEA). That is not the millions of deaths a year attributable to fossil combustion.

The economic prize, and who wins and loses

On the sourced evidence, a fast electrified transition is the cheaper path, with large health and import-bill gains. The gains are unevenly shared, and the losers can be identified and compensated [Medium].

The net-cost question is mostly about accounting baselines. Oxford's experience-curve study finds a fast transition saves at least USD 12 tn against staying on fossil fuels (University of Oxford). McKinsey's USD 9.2 tn/yr is gross spending; its own business-as-usual comparison puts the increment nearer USD 1 tn/yr (Inside Climate News). BNEF counts USD 2.3 tn of transition investment in 2025, against about USD 2.9 tn/yr needed in its base case (BNEF).

Security gains are already measurable.

• Clean energy avoided about USD 260 bn of fossil imports in 2025, about USD 110 bn of it in China (GreentechLead/IEA) [Medium; secondary].

• Europe's 2021–23 gas crisis cost governments €651 bn in consumer support (Bruegel).

• About 80% of humanity lives in fossil-importing countries (Ember).

Jobs.

• Energy employs about 76 million people, and electricity is now the largest energy employer (IEA).

• Renewables employ 16.6 million, 44% of them in China (IRENA/ILO).

• The IEA's net-zero path adds about 15 million energy jobs by 2035, but cuts coal-supply jobs by more than 60%.

• Renewables wages grew 0.8% in 2025, against 3.7% in oil and gas.

Real system costs are part of the picture.

• Germany puts its grid-balancing and related system costs at about €36 bn/yr (Wikipedia/Energiewende) [Low; tertiary source].

• Gas-price spikes still raised wholesale power prices in 21 EU countries in 2025 (Ember).

GroupGains or lossesEvidenceChinaLargest winner: one-third of clean investment, 44% of renewable jobs, USD 110 bn of imports avoidedIEA, IRENALarge fossil importers (EU, India, Japan, Pakistan)Lower import bills, fewer price shocksIEA, EmberPeople in polluted cities (China, India)Health gains; fossil air pollution kills 2.5–5.1 million a year globallyLancet, BMJFossil workers (18.4 million in supply)Coal jobs −60% and oil and gas −40% by 2035 under net zero; under 15% of coal workers covered by transition policiesIEAAsset ownersUSD 1.4 tn of stranded oil and gas value, two-thirds held via funds and pensionsSemieniuk et al.PetrostatesPlateauing demand; the Middle East holds 20% of upstream investmentIEAEmerging economies outside ChinaUnder 30% of energy investment for two-thirds of humanity; about 2× borrowing costs350.org/IEAUS householdsEnergy bills up USD 78–415/yr by 2035 after the 2025 tax-credit repeal (OBBBA)Rhodium; Kleinman

Fossil subsidies are regressive. The poorest 20% receive about 8 cents of each explicit subsidy dollar. Removing subsidies and recycling the revenue can therefore be pro-poor (IMF).

Policy, geopolitics and critical minerals

Policy is fragmenting into a fast, cheap China-linked bloc and a slower, costlier Western "trusted supply" bloc. Minerals risk is about concentration and weaponisation, not geological scarcity.

United States.

• The 2025 tax law (OBBBA) set a cliff for wind and solar credits: projects must start construction by 5 July 2026 or enter service by 31 Dec 2027. Storage, geothermal and nuclear keep credits into the 2030s (Latham & Watkins).

• US utility solar is now forecast roughly flat at about 43–44 GW/yr through 2031 (pv magazine USA; Solar Power World).

• The administration has paid about USD 3.9 bn to buy back offshore wind leases (OPB/AP) [High].

China.

• Emissions were −0.3% in 2025 and −1% in Q2 2026, with falling oil use now driving the decline (WINSS/CREA).

• China still added 30 GW of coal power in H1 2026.

• Its 2035 pledge (7–10% below peak) is judged consistent with about 3°C by CREA (Carbon Brief).

• Its 15th Five-Year Plan targets 30% wind and solar generation and 35% electricity share of final energy by 2030 (TaiyangNews).

European Union. The EU adopted a −90% target for 2040, delayed its second emissions-trading scheme (ETS2) to 2028, and proposed softening the main ETS for 2031–40 (OneStopESG; ASUENE). The carbon border tax (CBAM) went live on 1 Jan 2026, with carbon prices around €75–82/t.

Critical minerals.

• The top refiner's average share was 72% in 2025, and China leads refining for 19 of 20 strategic minerals (IEA GCMO 2026; IEA GCMO 2025).

• Mineral investment fell 9% in 2025, and copper faces a 25–30% supply deficit by 2035.

• Minerals are about 3% of an EV's price, so availability matters far more than price.

• Full Chinese export controls would put about USD 6.5 tn/yr of downstream production at risk (IEA) [High].

Two 2026 discontinuities come from a single research thread with partly secondary sourcing (to be verified against the IEA Oil Market Report and Reuters):

• US–Iran conflict and Strait of Hormuz disruption from about late February 2026. Reported details:

  • over 11 mb/d of Gulf crude and over 80 Mtpa of LNG curtailed (Wood Mackenzie);

  • Brent above USD 126/bbl at peak (IDN Financials/Reuters poll);

  • renewed disruption in October 2026, with about 4 transits/day against a baseline of 85 (Straits.live).

  The IEA's own EV outlook independently refers to a 2026 "largest oil supply shock in history" (IEA).

• 10 November 2026 US–China expiry cliff. China's suspension of its October 2025 rare-earth controls expires together with the US reciprocal-tariff suspensions (White House; Mining Technology).

Hard dateEvent5 Jul 2026 (passed)US wind and solar construction-start deadline10 Nov 2026US–China truce and rare-earth suspension expire (to be verified)30 Sep 2027First CBAM certificate surrender2028EU ETS2 startsJan 2029Next US administrationAfter 2033US storage, geothermal and nuclear credits phase down

Discontinuities: the leaps and convergences that could break the trend lines

The biggest "leap" is probably not one invention. It is the convergence of things already demonstrated: very cheap solar, cheap batteries and mass electrification crossing thresholds that open new uses. Large-unit "miracle" technologies deserve a heavy delay discount.

The forecasting record says which way to lean.

• Learning curves (Wright's law) forecast modular technologies better than expert models do (Way et al.).

• RethinkX was right on costs but more than 10 years early on adoption. In 2017 it projected self-driving "transport-as-a-service" carrying about 95% of US passenger miles by around 2030 (RethinkX). Waymo, the largest robotaxi operator, ran about 400,000–500,000 paid rides a week in 2026 (Automotive World; InsideEVs).

• Slow-transition sceptics have been right that absolute fossil use keeps rising: every fuel hit a record in 2025 (EI).

• The "primary energy fallacy" reconciles the two. Electrification removes about two-thirds of the conversion losses in fossil systems, so fossil shares of primary energy overstate the remaining task.

Leap candidates. The ranking is the researchers' speculation; the evidence column reports the notes.

CandidateEvidence status, Oct 2026What must be trueMaterial impactRank (2046)Ultra-cheap solar + batteries + sodium-ionDemonstrated cost curves; tandem cell record 35.5% (pv magazine)Continuation, plus integration (grids, storage)Already under way1Robotaxis and fleet electrificationWaymo grew ~2× in under a year from a small baseApprovals in many jurisdictions; cost per mile below owning a car2030s2AI-optimised grids and operationsIncremental; little quantifiedUtility adoption2030s3Enhanced / superhot geothermalFirst commercial enhanced plant (Fervo); superhot drilling started (Quaise)Drilling >10 km and >374°C2035+4Synthetic fuels from cheap solarClaim stageElectrolysers following learning curves; solar ~USD 0.10–0.20/W2035+5Geologic hydrogenUS prospectivity map published (USGS); no commercial productionEconomic accumulations found2030s at best6Commercial fusionNo net electricity yetNet-electricity pilot around 2030Post-2046 capacity7Space solarMilliwatt-scale demonstration (TechCrunch)Launch costs ≪ USD 100/kg; km-scale arrays2040s+8Room-temperature superconductorsNone credibleA discoveryUnknown9

AI-accelerated science is real but narrow. An audit of Google DeepMind's GNoME materials database found "scant evidence" that the claimed compounds were novel, credible and useful (Cheetham & Seshadri). Discovery-to-gigawatt lags have historically run 10–20+ years, so AI discoveries made in 2026–2030 would mostly reach markets in the late 2030s–2040s [Low].

Negative discontinuities are already happening. Coral reefs have passed their tipping point, and 2026 has brought a Hormuz oil and LNG shock (to be verified against the IEA Oil Market Report and Reuters). Shocks to fossil supply tend to accelerate electrification among importers. Shocks to clean-tech supply mainly slow Western deployment.

Archetype worlds for scenario building (researcher synthesis).

WorldKey driversSignposts1. Electrotech AbundanceLearning curves continue; Chinese overcapacity; distributed adoptionTotal fossil use falls two years running; emerging-market solar imports keep doubling2. AI-driven SurgeHyperscaler capex; all sources builtElectricity demand growth above 4%/yr; turbine orders; firm-power deals3. Fragmented Fossil PersistenceTariffs, an LNG glut, policy reversalsLocal-content rules spread; LNG investment decisions surge; ~2.7–3°C4. Two-Speed Green Trade WarChina-linked Global South electrifies fast while the West goes slowerPrice gaps across the US/EU border; Chinese export records5. Climate-Shock-Forced TransitionDisasters drive emergency policyInsurers withdrawing; hydro failures; tests of solar geoengineering6. Stalled StagnationAI bust or financial crisis; high interest ratesAI capex retrenchment; real rates above 3%

Key uncertainties and stress-tests for any scenario

Seven uncertainties separate the futures. Any scenario should be stress-tested against each, using the ranges below rather than single points. The two axes that best distinguish the archetype worlds are geopolitical integration vs fragmentation, and high vs low demand growth (AI plus the Global South). Technology surprise and climate shock are overlays on top.

UncertaintyPlausible rangeSignpost to watchStress testConfidence in rangeClean-tech deployment paceSolar frozen at ~540 GW/yr (IEA CPS) vs SolarPower Europe's ~864 GW/yr by 2030Annual solar and storage additions; China after its auction reformDoes the plan still work if solar additions stall at the 2025 level until 2035?MediumAI / data-centre demand2035: 700–1,700 TWh; 2046: 1,500–5,000 TWh (derived)Hyperscaler capex; cancellations; chip and transformer lead timesRerun regional grids (US, Ireland) at the Lift-Off level and at a capex bustMedium / LowGrid build and permittingUS 536 miles/yr vs ~5,000 needed; global grid spend 400bnvs>600 bn neededTransformer lead times; queue completion rates; reconductoring uptakeWhat if grid spending stays flat in real terms?HighOil and gas path2050 oil 75–124 mb/d across outlooks; Hormuz bimodal (USD 60s or USD 100–200)IEA Oil Market Report; Chinese crude imports; LNG investment decisionsTest both a 2027 oil glut and an extended closure (to be verified)Medium / LowTrade and mineralsExtension vs full re-escalation on 10 Nov 2026; copper deficit 25–30% by 2035Export licences; copper prices; non-Chinese magnet capacityWestern storage at USD 219/kWh vs USD 73/kWh in ChinaMediumCost of capital in emerging economies8.5–11% vs ~5% weighted average cost of capitalConcessional finance flows; sovereign ratingsHalve and double emerging-market capital costs; check universal access by 2035HighClimate damage and feedbacksDamages from the IPCC's "few percent" to the retracted-and-revised 6–31% (revision not peer-reviewed); AMOC 2037–2109 vs no collapseNGFS Phase VI; hydro droughts; insured-loss trend (+5–7%/yr real)Add hydro and thermal derating plus a USD 300 bn+ catastrophe yearLow–Medium

Firm-power learning. No Western first-of-a-kind nuclear or SMR has yet shown unit-to-unit cost decline. Treat Western nuclear above about 3 GW of SMRs by 2035 as an upside case only.

Policy durability. DNV estimates US policy reversal delays US emissions cuts by about five years (DNV). Test a further US reversal after January 2029, and also a restoration of credits.

Measurement risk. Several core numbers differ by 10–60% between sources: solar AC vs DC, data-centre TWh, access counts, and CCS tonnes. Every scenario input should carry its definition.

Implications for the vision

The system that gives the most economic benefit to the most people by 2046 is an electrified, solar-and-battery-led system built on much larger grids and cheaper capital for the Global South. It is firmed regionally by hydro, geothermal and nuclear where institutions can deliver them, with gas as a declining backstop. AI load should be connected flexibly so it helps pay for grids rather than locking in fossil plants.

The case is strong on several dimensions.

• Cost: solar is the cheapest new electricity almost everywhere.

• Firm clean power is affordable: India buys solar with 4-hour storage at about 3.3 US¢/kWh.

• Health: fossil air pollution kills 2.5–5.1 million people a year.

• Security: 80% of people live in fossil-importing countries.

• Access: universal electricity costs less than 1% of today's energy investment.

The trade-offs, in plain language.

• Speed versus fairness. Cheap rooftop solar lets the better-off leave the grid first. In Pakistan, 88% of high-spending households have solar against 38% of low-spending ones, which raises bills for those left behind. Tariffs that charge for grid services rather than per kWh, plus pay-as-you-go finance, extend the benefit downward.

• Cheap kit versus supply-chain dependence. Chinese panels, cells and minerals make the transition affordable. Trade walls make it 2–3 times costlier in the US (storage USD 219 vs USD 73/kWh). A two-speed world is likely, and Global South buyers benefit from the cheap bloc.

• Building fast versus local consent. Seven in ten Americans oppose a nearby data centre. Grid lines take 5–15 years. Reconductoring, flexible loads and co-located storage avoid most fights over new land.

• AI growth versus emissions. Without flexible connection and fast clean firm supply, AI demand is met with gas turbines and retained coal. With them, AI capital can finance grids.

• Big bets versus modular bets. History favours modular, factory-made technology. Keep nuclear, geothermal and fusion options alive, but do not plan on them arriving early.

Priorities that follow from the evidence.

1. Treat grid expansion, permitting and transformer supply as the top constraint, ahead of generation.

2. Cut capital costs in emerging economies: local-currency lending, guarantees and concessional finance (about USD 28 bn/yr of concessional money could unlock about USD 90 bn/yr of private investment in Africa (IEA)).

3. Fund clean cooking and productive-use appliances, the most neglected gaps.

4. Cut methane now, the cheapest near-term warming lever.

5. Pay for just-transition support in coal regions.

6. Diversify mineral refining and build recycling.

The inspiring version is grounded in data. Solar generation doubles about every three years (Ember). Ember estimates that electrified technologies are about three times more efficient than fossil systems at delivering the same services. The physical potential for renewables exceeds current demand more than tenfold in 92% of countries (Ember). A world of abundant, cheap, clean, locally produced electricity is a credible 20-year destination. Whether it arrives in time, and for whom, is a choice about grids, finance and fairness.

Source library and data atlas

SourceBest forCadence / latest editionURLIEA World Energy OutlookScenario baselines (CPS/STEPS/NZE), access pathwayAnnual (Oct–Nov); WEO 2025, Nov 2025https://www.iea.org/reports/world-energy-outlook-2025/executive-summaryIEA Global Energy ReviewPrior-year outturn: demand, CO2Annual (spring); GER 2026, Apr 2026https://www.iea.org/reports/global-energy-review-2026/key-findingsIEA ElectricityPower demand/supply to 2030, queues, gridsAnnual (Feb); Electricity 2026https://www.iea.org/reports/electricity-2026/executive-summaryIEA RenewablesRenewable capacity forecasts, heatAnnual (Oct); Renewables 2025https://www.iea.org/reports/renewables-2025/executive-summaryIEA World Energy InvestmentClean vs fossil capital flowsAnnual (May–Jun); WEI 2026https://www.iea.org/reports/world-energy-investment-2026IEA Global EV OutlookEV sales, fleet, oil displacement, chargingAnnual (May); GEVO 2026https://www.iea.org/reports/global-ev-outlook-2026/executive-summaryIEA Energy and AIData-centre demand cases and supply mix2025 report + Jun 2026 updatehttps://www.iea.org/reports/energy-and-ai/executive-summaryIEA Global Critical Minerals OutlookRefining concentration, supply gapsAnnual; GCMO 2026, Jul 2026https://www.iea.org/reports/global-critical-minerals-outlook-2026/executive-summaryIEA Global Methane TrackerFossil methane, abatement costAnnual; 2025https://www.iea.org/reports/global-methane-tracker-2025/key-findingsIEA Global Hydrogen ReviewHydrogen pipeline vs final investment decisionsAnnual; 2025https://www.iea.org/reports/global-hydrogen-review-2025/executive-summaryIEA World Energy EmploymentEnergy jobs by sector and scenarioAnnual (Dec); 2025https://www.iea.org/reports/world-energy-employment-2025/executive-summaryEmber Global Electricity ReviewGeneration by source and country; monthly dataAnnual (Apr) + monthly; GER 2026https://ember-energy.org/latest-insights/global-electricity-review-2026/Energy Institute Statistical ReviewPrimary energy by fuel, long historyAnnual (Jun); 75th ed., Jun 2026https://www.energyinst.org/statistical-reviewGlobal Carbon BudgetFossil CO2, carbon budgetAnnual (Nov); GCB 2025https://globalcarbonbudget.org/BNEF New Energy OutlookEconomics-only (ETS) and net-zero scenariosAnnual; NEO 2026, May 2026https://assets.bbhub.io/professional/sites/44/BNEF_20260515_NEO26-Exec-Sum.pdfBNEF battery price and LCOE surveysCost curves for packs, storage, generationAnnual (Dec / Feb)https://about.bnef.com/insights/clean-energy/lithium-ion-battery-pack-prices-fall-to-108-per-kilowatt-hour-despite-rising-metal-prices-bloombergnef/BNEF Electric Vehicle OutlookEV sales and fleet forecastsAnnual (Jun); EVO 2026https://assets.bbhub.io/professional/sites/44/EVO2026-Executive-Summary.pdfDNV Energy Transition OutlookSingle best-estimate forecast to 2060Annual (Oct); ETO 2025 (2026 not yet located)https://www.dnv.com/eto2025/IRENA Renewable Power Generation CostsGlobal LCOE and installed cost (AC)Annual (Jul); 2024 datahttps://www.irena.org/-/media/Files/IRENA/Agency/Publication/2025/Jul/IRENA_TEC_RPGC_in_2024_Summary_2025.pdfITRPVPV technology roadmap, learning rate, silver useAnnual; 17th ed., 2026https://www.pv-tech.org/itrpv-2026-solar-industry-maintains-historic-learning-curve-despite-market-turbulence/SolarPower Europe Global Market OutlookSolar installations (DC) and forecastsAnnual (Jun); GMO 2026–2030https://www.solarpowereurope.org/press-releases/new-report-global-solar-market-hits-new-record-of-664-gw-installation-in-2025-as-global-solar-fleet-passes-3-twGWEC Global Wind ReportWind installations, offshoreAnnual (Apr); GWR 2026https://www.gwec.net/reports/globaloffshorewindreportLazard LCOE+US unsubsidised costs incl. nuclear and storageAnnual (Jun–Jul); 2026https://www.lazard.com/media/kcfconhf/lazards-lcoeplus_vf.pdfLBNL Queued UpUS interconnection queues and completion ratesAnnual + update; 2025 ed. / 2026 updatehttps://emp.lbl.gov/queuesGrid Strategies load growth reportUS utility load forecastsAnnual (Nov); 2025https://gridstrategiesllc.com/wp-content/uploads/Grid-Strategies-National-Load-Growth-Report-2025.pdfWorld Nuclear AssociationReactor pipeline, costsContinuous; Outlook Sep 2026https://world-nuclear.org/information-library/current-and-future-generation/plans-for-new-reactors-worldwideSEforALL / Tracking SDG7Electricity and clean-cooking accessAnnual (Jun); 2025 (2023 data)https://www.seforall.org/system/files/2025-12/SEforALL%20SDG7_TRACKING_2025-data_analysis.pdfGOGLA Off-Grid Solar Market ReportOff-grid kits, pay-as-you-go, productive useAnnual; 2026 (2025 data)https://gogla.org/wp-content/uploads/2026/06/GOGLA_Global-Off-Grid-Solar-Market-Report_2026_DEF3.pdfUNEP Emissions Gap / Adaptation GapPolicy warming gap; adaptation financeAnnual (Oct–Nov); 2025https://unep.org/resources/emissions-gap-report-2025Climate Action TrackerWarming implied by policies and pledges~Twice yearly; Nov 2025https://climateactiontracker.org/global/temperatures/Forster et al., Indicators of Global Climate ChangeHuman-induced warming, carbon budgetAnnual (Jun); ESSD 2026https://essd.copernicus.org/articles/18/3889/2026/Copernicus C3S bulletinsMonthly global temperatureMonthlyhttps://climate.copernicus.eu/climate-bulletinsLancet CountdownHeat, pollution and health indicatorsAnnual (Oct); 2025https://lancetcountdown.org/2025-report/State of Global AirAir-pollution deathsAnnual; 2025https://www.stateofglobalair.org/news-events/2025/new-state-global-air-2025-report-shows-nearly-nine-ten-global-air-pollution-deathsIMF fossil fuel subsidiesExplicit and implicit subsidiesPeriodic; WP 2025/270, Dec 2025https://www.imf.org/en/publications/wp/issues/2025/12/20/underpriced-and-overused-fossil-fuel-subsidies-data-2025-update-572729Swiss Re sigmaInsured and economic catastrophe lossesAnnual + H1; sigma 1/2026https://www.swissre.com/institute/research/sigma-research/sigma-2026-01-natcat-2025-wildfire-storm-risk.htmlGlobal Tipping Points ReportTipping thresholds; positive tipping pointsIrregular; 2025https://wwfes.awsassets.panda.org/downloads/gtp_summary_report_2025-v4-pages-lo-res--1-.pdfNGFS scenariosCentral-bank macro-financial climate scenariosPhase V Nov 2024; Phase VI expected 2026https://www.ngfs.net/system/files/2025-11/NGFS%20scenarios%20narratives%20and%20key%20findings.pdfRhodium/MIT Clean Investment MonitorUS clean investment and cancellationsQuarterly; Q2 2026https://rhg.com/research/clean-investment-monitor-us-q2-2026-update/OPEC World Oil OutlookProducer-side oil viewAnnual (Jun–Jul); WOO 2026 (via secondary summary)https://energynewsbeat.co/crude-oil/world-oil-outlook-2050-opecs-woo-2026-robust-growth-to-124-mb-d-no-peak-in-sight/ExxonMobil Global OutlookCompany view of oil and gas to 2050Annual; 2026 editionhttps://corporate.exxonmobil.com/sustainability-and-reports/global-outlookIEA Oil (medium-term report); IEA Oil Market Report and Reuters for verificationOil demand plateau and spare capacity; the monthly OMR is needed to verify the 2026 Hormuz claimsAnnual Oil 2025 (fetched); OMR monthly (not fetched)https://www.iea.org/reports/oil-2025/executive-summary