The world burns more energy than ever — and now, finally, in different ways.
Since 2010, global primary energy consumption has risen by about a quarter. Coal demand keeps reaching new records, but its growth rate has slowed. Renewables have moved from a footnote to a fast-growing chapter. This is the helicopter view.
Four things that actually changed
First: the world did not stop burning fossil fuels. Coal, oil and gas all consume more energy today than in 2010 — and coal reached record-high consumption in both 2023 and 2024. Coal's growth rate has slowed sharply since the early 2010s, with China's trajectory now the critical variable for whether and when global demand peaks. Renewables grew faster than overall demand, but mostly on top of fossil consumption rather than displacing it. (A measurement note: headline charts on this site use the IEA's substitution method, which inflates renewables to make them comparable to fossil fuels in primary-energy terms. The Overview chart has a toggle so you can see the alternative direct-method view too — the framing affects how big renewables look.)
Second: the centre of gravity moved decisively east. China alone accounted for more than half of all global energy demand growth since 2010, and India is now the second-fastest-growing major consumer. The European Union, by contrast, consumes meaningfully less energy in 2024 than in 2010 — a decline driven primarily by structural forces: EU emissions trading (the ETS) pricing carbon into industrial decisions, successive efficiency directives, the offshoring of energy-intensive manufacturing, and the long phase-out of coal-fired generation. The 2022 gas crisis accelerated the trend but didn't cause it; industrial gas demand partly recovered in 2024.
Third: electricity is eating energy. Electricity generation grew roughly 25% faster than total primary energy over the period — a sign that more of the economy is electrifying, even as overall energy demand keeps climbing. Renewables now generate around 30% of global electricity; solar and wind are the fastest-growing sources by a wide margin.
Fourth: prices broke their pattern. The decade was bookended by two oil-price collapses (2014 and 2020) and one extraordinary natural gas shock (2022, in Europe), with Dutch TTF gas briefly trading above ten times its pre-crisis level. The price story matters more than the volume story for understanding the politics of energy.
How to read this site
Eight tabs. Overview is the helicopter view. Geography lets you click any region of the world to see its fuel mix. Europe drills into all 27 EU member states with electricity mix and retail prices. Sources goes fuel-by-fuel. Sectors shows where the energy actually goes. Prices tracks commodity volatility. AI & Data covers data-centre electricity demand — the fastest-growing slice of global power use. Future sketches three scenarios out to 2050.
Any term with a dashed underline has a quick explanation on hover. The Glossary button (top-right) opens a full reference for units and concepts.
Why 2010 to 2024 — and a few exceptions
The Atlas's spine is the fifteen-year window 2010 to 2024. There are three reasons.
2010 sits just after the 2008–09 global financial crisis. Starting the clock there avoids letting an unusually deep recession year distort growth-rate calculations, while still capturing the entire modern energy-transition era: the rise of US shale, the post-Fukushima nuclear retreat, the Paris Agreement, the renewables cost collapse, the 2014 oil crash, COVID, the 2022 European gas crisis, and the recent acceleration of AI-driven electricity demand.
2024 is the last full year for which the major data sources have published cleaned, audited annual figures: the Energy Institute Statistical Review of World Energy 2025 (released June 2025, covering through 2024) and Ember's Yearly Electricity Data. Going further would mean using preliminary or quarterly numbers, which would compromise the consistency of every chart.
Where newer data adds genuine value, it appears as a clearly-marked exception. The Geography tab includes a country-level electricity ranking for 2025 (Ember Global Electricity Review 2026, released April 2026). The Sources tab notes the 2025 milestone of coal slipping below one-third of global electricity. The Prices tab extends commodity benchmarks through 2025 and partial 2026, including the Iran conflict that began February 2026. The Future tab uses IEA WEO scenarios that run to 2050 — these are scenarios, not forecasts.
Why no further back than 2010? Because going back further (say, to 2000) would dilute the modern-transition narrative with a different era — pre-shale, pre-Paris, pre-renewables-cost-collapse — and the questions a 2026 reader actually wants answered are about the period we've been in, not the period that preceded it.
Where the world's energy is used.
No two regions tell the same story. China consumes around eight times more energy than the entire African continent. The European Union is the only region using less energy today than fifteen years ago. Click any region on the map to see its fuel mix, its trajectory, and why it matters.
How the regions compare
Total 2024 primary energy consumption, ranked.
The biggest electricity consumers · country view, 2025
A complementary lens: this ranks individual countries by electricity demand (not total primary energy, and not by region). Toggle to per-capita to see how total scale and personal consumption diverge — China leads on total but sits mid-table per person.
Notes on the regional split
- European Union covers the 27 member states (UK shown but excluded from EU totals).
- North America covers the US, Canada and Mexico.
- Other Asia-Pacific covers Japan, South Korea, ASEAN, Australia/NZ, Central Asia, and other non-China/non-India Asia.
- Latin America covers South America plus Central America & the Caribbean.
- Slight overlap (~3%) exists at the continental level — the world total is treated as ground truth.
- The country ranking above is electricity demand for 2025 (Ember) — a different metric and a more recent year than the primary-energy regional data, included as a complementary view.
Twenty-seven countries. One grid. Twenty-seven different stories.
The EU is treated as a single energy market — but Germany pays four times what Hungary pays for residential electricity, and Sweden's electricity mix is barely recognisable next to Poland's. Click any country to see how its grid is powered and what its households and businesses actually pay.
The EU electricity transition, at a glance
Household electricity prices · cross-country comparison
Average residential rate including all taxes and VAT, second-half 2024. Hover for more detail.
On this H2 2024 snapshot, Germany is dearest and Ireland ranks 3rd (€0.37/kWh, behind Germany and Denmark). On the latest H2 2025 Eurostat data, Ireland rises to the top of the EU range (€0.4042/kWh) — see the Ireland case study on the Future tab. Cross-country rankings shift by reference period, so the two figures are consistent, not contradictory.
Why prices vary so much across the EU
The headline price gap between Germany (~€0.39/kWh) and Hungary (~€0.11/kWh) isn't primarily about generation costs. The biggest structural drivers are taxes, VAT and policy levies — which in some countries make up close to half the residential bill. Germany historically funded its renewables build-out with the EEG-Umlage, a surcharge added directly to household electricity bills. Denmark applies high electricity taxation as part of broader fiscal policy. By contrast, Hungary, Bulgaria and Malta regulate residential tariffs at levels below market cost, with the gap funded by the state.
The physical cost differences — generation mix, grid investment needs, interconnection — matter, but they're a smaller component than the policy architecture. Countries with cheap, abundant nuclear (France) or hydro plus nuclear (Sweden) have lower wholesale costs; countries reliant on imported gas, with grid bottlenecks (Italy, Ireland), have higher wholesale costs. But the retail bill a household actually pays is shaped at least as much by national tax policy as by what's flowing through the wires.
About the EU price data
Eurostat publishes per-country electricity and gas prices semi-annually. Anchor values for 2010, 2015, 2020, 2022 and 2024 are taken directly from Eurostat datasets nrg_pc_204 (household electricity), nrg_pc_205 (non-household electricity) and nrg_pc_202 (household gas). Values for intervening years are linearly interpolated — the shape captures the actual story (gradual rise 2010–2020, dramatic spike in 2022 for most countries, partial normalisation through 2024). Countries without significant residential piped gas (Cyprus, Malta, Finland) are not shown on gas charts.
The same sun. The same earth. Different ways of pulling energy out of them.
Coal is the heaviest of the climate's enemies and the cheapest of its fuels. Oil is unrivalled for moving things. Gas is the bridge that became contested. Nuclear is the quiet survivor. Renewables are no longer marginal. Click a fuel to see how it has fared.
The carbon weight of each fuel
Grams of CO₂ emitted per kilowatt-hour of energy delivered — lifecycle estimates, widely-cited median values from peer-reviewed studies.
What "carbon weight" means and how it's calculated
The chart below shows lifecycle carbon intensity — the total greenhouse-gas emissions, expressed as grams of CO₂-equivalent, attributable to one kilowatt-hour of useful energy from each fuel. Two things matter here.
First: lifecycle, not just combustion. The figure includes the emissions from extracting the fuel (mining coal, drilling for gas, refining oil), transporting it, building the power plant or solar farm itself, and decommissioning it at end of life. For fossil fuels, combustion dominates everything else — most of coal's 820 g/kWh is from burning it. For renewables and nuclear, there's no combustion at all, so what's measured is the embedded carbon in manufacturing panels, turbines, reactors, concrete, steel and copper, divided across the energy each unit produces over its lifetime.
Second: CO₂-equivalent, not just CO₂. Methane leaks (especially from natural gas extraction), nitrous oxide, and other gases are converted into their CO₂-warming equivalent over 100 years (the GWP-100 framework) and added in. This is why gas isn't half of coal despite being the cleaner-burning fuel — methane leaks across the gas supply chain partly offset the combustion advantage.
The numbers shown are median lifecycle values from the IPCC's Sixth Assessment Report (AR6, Working Group III, Annex III) and US National Renewable Energy Laboratory meta-analyses. Real-world values vary widely by deployment — a solar farm in cloudy northern Europe will have a higher per-kWh figure than one in Spain, simply because it generates less electricity over its lifetime to amortise the panel-manufacturing emissions. These are central estimates, not precise figures.
An economy is a machine for converting energy into useful work.
Roughly a third for industry. A bit less for moving things. A bit less again for keeping buildings warm, cool and lit. A sliver for agriculture. The proportions have barely moved in fifteen years — but what's powering each sector has.
Caveat on sector data
The IEA's free-access sector splits are released with a ~2 year lag and at lower granularity than fuel data. The figures above are anchored to published IEA values and smoothed across the period — directionally robust but not precise to the percentage point. Sector fuel mixes shown are most-recent-year approximations.
The decade of price shocks.
Energy prices over fifteen years tell a story that no single chart of volumes can. Three moments — the 2014 oil crash, the 2020 COVID collapse, and the 2022 European gas crisis — reshaped both markets and politics. Here are the benchmarks that mattered.
What the price data says
Four things stand out. Oil remains the single most internationally-traded commodity, and its price cycles still drive macroeconomic outcomes worldwide. Adjusted for inflation, the 2011–14 peak (above $130 in 2024 dollars) was substantially higher than the 2022 Ukraine spike (~$108 in 2024 dollars) — a striking reframing of what most readers think of as a recent oil price record. The 2014 crash (engineered by US shale taking market share) and the 2020 COVID demand collapse were both deflationary forces. 2025 was the cheapest year in nominal terms since 2020, helped by sluggish demand and OPEC+ output increases.
Then came February 2026. The US-Israel military operation against Iran on February 28, followed by Iranian retaliation against US bases and the de facto closure of the Strait of Hormuz (through which roughly 20% of global oil supply normally flows), drove Brent from around $71 in late February to a monthly average of $117 in April and a daily peak of $138 on April 7. The year-to-date marker shown on the chart sits around $97, reflecting the high volatility and partial-year averaging. This event is comparable in severity to the 2022 Ukraine shock and is still unfolding.
The natural gas story remains the most consequential of the period. Until 2022, European TTF prices typically traded at 2-3× Henry Hub. After Russia's invasion of Ukraine and the resulting gas cut-off, that gap blew out to more than 10× at the peak, with European industrial consumers — particularly chemicals, fertilisers and ceramics — facing existential cost pressure. The gap narrowed through 2024–25 but remains structurally wider than it was pre-2022.
For coal, the 2022 spike was the largest in modern memory, driven by European utilities scrambling to replace gas and Asian demand staying robust. Prices have normalised but coal's economics have become more volatile than the long preceding stable era of 2015-2020.
What's missing from this commodity view
For retail electricity prices, head to the Europe tab — country-by-country data is most reliable for EU members. Retail prices outside the EU vary so widely by subsidy structure and currency that any survey-level chart would mislead more than inform. Notable points: US electricity prices rose far less than European; many Middle Eastern and African retail prices are heavily subsidised and don't reflect underlying costs. Carbon prices (EU ETS) are a separate story not shown here.
A few hundred buildings now drive global electricity demand growth.
Until recently, data centres were a rounding error in global electricity. AI has changed that. They now consume more electricity than the United Kingdom — and the IEA's base case has that figure more than doubling by 2030. The implications run from grid investment to where new power plants get built.
Where the data centres are — and where the growth is
US has 45% of current capacity; China is closing fast. Europe is third. Together, the US and China account for nearly 80% of projected growth to 2030.
Local concentration is the political story
Globally, data centres are 1.5% of electricity demand — manageable. Locally, they can be 25% or more, and that's what makes them a grid-planning problem. Ireland's national grid operator is constraining new data-centre connections; US states are racing to extend transmission lines; new gas-fired peaking plants are being permitted specifically to serve hyperscale campuses.
Data-centre share of metered electricity demand in selected places (CSO Ireland, EirGrid, Virginia State Corporation Commission). Last cell shows the share of US data-centre capacity concentrated in just five regional clusters.
Share of national electricity demand growth from data centres, 2024–2030
In some advanced economies, data centres explain half or more of all new electricity demand to 2030.
How data centres are powered today
Roughly 35% gas, 27% renewables, 18% coal, 12% nuclear — though the renewables figure reflects reported corporate procurement (including virtual PPAs), not necessarily the physical grid mix in real time.
How AI buys its power
The headline "data centres are 27% renewable" rests on a piece of accounting that's worth understanding. Standard renewable Power Purchase Agreements (PPAs) are financial contracts, not physical electricity supply. A hyperscaler signing a 200 MW solar PPA in Texas hedges its electricity costs against that solar farm's output, and gets to claim 200 MW of renewable procurement on its sustainability report. But the data centre itself may sit on a different grid entirely — and even where the geography matches, the timing doesn't: solar generates during the day, the data centre runs 24 hours.
For years the industry treated this as good enough. Annualised paper accounting let Google, Microsoft, Amazon and Meta declare themselves "100% renewable" while their actual grid power remained mostly fossil-fuelled at night and during low-wind periods. The accounting wasn't dishonest — it correctly described how much new renewable capacity these contracts added to the world — but it didn't match the marketing language.
That gap is now closing. Google announced a goal of 24/7 Carbon-Free Energy in 2020: every hour of consumption matched by carbon-free generation on the same regional grid. Microsoft and Amazon have followed with similar commitments. Hitting 24/7 CFE is much harder than buying annual PPAs and typically requires a portfolio mixing solar, wind, batteries, and firm clean power — which in practice means nuclear, geothermal, and hydro. This is why hyperscalers have signed deals to restart Three Mile Island Unit 1 (Microsoft), buy power from existing nuclear plants (Amazon, Constellation), and back small modular reactor developers (Google, Amazon, Meta).
The short version: AI demand growth is forcing the cleanest part of the tech industry to graduate from accounting-grade renewable claims to physically-matched clean power. That's a meaningful shift, even if the absolute numbers remain small relative to total demand growth.
What this means for the energy transition
The net effect on emissions depends on three things: (1) how fast renewable build-out and 24/7 CFE deployment can keep pace with AI demand growth, (2) whether grid bottlenecks force new fossil capacity that would otherwise have retired — gas-fired peaking plants are already being permitted specifically to serve hyperscale campuses, and (3) whether efficiency gains in chips and cooling offset compute demand. Historically efficiency has roughly tracked compute growth; AI's curve may finally outrun it.
The uncertainty range matters
The IEA's "base case" of 945 TWh by 2030 is one of several credible projections. Some industry analysts see 1,200+ TWh by 2026 (i.e. tomorrow); others see efficiency gains slowing the curve. The honest answer is that we are in early innings of a major new demand source whose trajectory we cannot yet pin down to within ±30%.
Five futures, ranked by how much each rests on evidence.
The world consumed 624 EJ of primary energy in 2024, with fossil fuels supplying ~87% — a share that has barely moved in a decade. This tab presents five credible trajectories for the next 25 years, ranked by how much of each rests on observed behaviour versus modelled assumption versus policy aspiration. Scenarios are stress tests, not forecasts.
The 2024 baseline
Six measured numbers that anchor everything downstream. All Observed.
The transition is real on the renewables-build axis and on the China/India coal-still-growing axis simultaneously. Both can be true. Most disagreements about "the transition" are disagreements about which of these trends extrapolates and which inflects.
Five trajectories
Total primary energy demand, exajoules · 2010 – 2050. Isolate a line by clicking its pill; click again to restore. The "Hide" pills toggle off a whole label category.
Where the demand comes from
The growth is not evenly spread. Four stories carry most of it.
India · The growth story
India's energy demand has roughly doubled since 2010 and IEA projections suggest it could double again by 2040. The population is younger and more urbanising than China's, the air-conditioning rollout is only beginning, and the middle class is expanding fast. Coal still provides over 70% of electricity, but solar capacity is the world's third-largest and growing.
China · Mix shift, not volume growth
China is already the world's largest energy consumer and its annual demand growth is slowing. The story is no longer about how much, but about what. China added more solar in 2024 than the rest of the world combined; coal demand may have already peaked. The interesting question is how fast renewables displace fossil generation, not how much total energy China uses.
United States · An electricity story, not an energy story
Total US primary energy in 2024 is about the same as in 2010. What's changing is the electricity picture: AI/data-centre demand, electrification of vehicles and heating, and reshored manufacturing are now driving electricity demand growth for the first time in nearly two decades. Most of that growth needs to be served by some combination of natural gas, nuclear, and renewables — exactly which mix is fiercely contested.
SE Asia + Middle East · The under-discussed growth
Indonesia, Vietnam, the Philippines, Thailand — plus Saudi Arabia, UAE, Iran — together account for more projected energy demand growth than the US or Europe. Cooling demand from air-conditioning, population growth, and industrialisation drive the picture. Coal use is still rising across much of Southeast Asia; the Middle East is solar-shifting but from a near-zero base.
The five scenarios
Each card carries a label, a probability band with reasoning, and a falsifier — the evidence that would change the view.
IEA CPS · Current Policies Modelled — IEA WEO 2025
- Fossil share stays above 75% through 2040
- China coal peaks late-2020s but declines slowly
- India coal grows through 2040
- Renewables triple — but from a small share
IEA STEPS · Stated Policies Modelled — IEA WEO 2025
- Assumes meaningful follow-through on announced targets
- Coal peaks globally by 2027
- Solar and wind reach ~40% of generation by 2050
- Gas demand plateaus rather than grows
IEA NZE · Net Zero by 2050 Aspirational — no precedent for transition this fast
- Coal phase-out by 2040 (advanced economies), 2050 globally
- Oil demand falls 75% from peak
- Annual clean-energy investment triples by 2030
- Behavioural change contributes ~15% of emissions cuts
BP / Energy Institute & OPEC range Modelled — industry sources
- Oil: BP 95–105 mb/d in 2030, OPEC 110+ mb/d in 2030
- Gas demand grows in all industry scenarios through 2035
- Coal declines slowly in OECD, persists in emerging Asia
Slow Transition · Realist case Extrapolation from observed rates & political behaviour
- Physical constraints (Vaclav Smil): modern civilisation rests on four pillars — cement, steel, plastics, ammonia — with no scalable substitute today. Solar and wind have scaled fast in electricity (~20% of final energy); the remaining 80% (heat, transport fuel, industrial feedstock) is harder.
- Political economy (Roger Pielke Jr.): when climate policy conflicts with economic growth, growth wins — the "iron law", observed across democracies for 30+ years. Voters accept emissions cuts only when they don't notice the cost.
- Normative case (Alex Epstein): fossil fuels deliver affordability, reliability and scalability alternatives cannot yet match. Cost-benefit that prices in human flourishing argues for slower decarbonisation.
What it costs, and who pays
Most transition discussions price the generator (LCOE) and stop. Real-world costs split across system costs (grid, storage, backup), household bills, industrial competitiveness, and public spending — and each falls on a different group.
A · Generator cost (LCOE) Observed — Lazard 2024 LCOE+
- Solar PV (utility): $29–$92/MWh
- Wind (onshore): $27–$73/MWh
- Combined-cycle gas: $45–$108/MWh
- Solar + 4-hr storage: $60–$210/MWh
- Wind + 4-hr storage: $45–$133/MWh
LCOE captures generator cost only. It excludes grid integration, balancing, and the cost of running thermal plant at low utilisation as backup.
B · System cost (what LCOE misses) Observed + Modelled
Observed System costs rise non-linearly above ~40% variable-renewable share. The UK, Germany and Australia all show rising consumer bills despite falling LCOE.
Modelled — utility studies At 80%+ renewable share, system costs (storage, transmission, curtailment, firming) add $30–$80/MWh on top of LCOE, depending on grid characteristics.
C · Household bills Observed — Eurostat H2 2025, DESNZ
- EU-27 average household electricity: ~€0.29/kWh
- Ireland: €0.4042/kWh (40% above EU avg, ~€480/yr premium)
- Germany: ~€0.39/kWh
- France: ~€0.25/kWh
- UK: ~£0.28/kWh
High bills cluster around: gas-marginal pricing (electricity priced at the gas plant clearing the market), renewables financing passed through, grid expansion, and policy levies.
D · Industrial & public spending Observed + Modelled
Observed UK industrial electricity for large users H1 2025: 25.33p/kWh — 125% above the EU-14 median (DESNZ). UK Steel reports £845m excess electricity cost vs French competitors since 2016/17.
Modelled — CCC 7th Carbon Budget, 2025 Net UK cost of net-zero by 2050: ~£4bn/yr (~0.2% GDP) — ~£26bn/yr gross investment offset by ~£22bn/yr operating savings.
Modelled — OBR July 2025 Total net UK resource cost 2025–2050: £116bn.
Two case studies: Ireland & UK
Same structure for each: headline, three drivers, what would actually move bills, verified-figure footer.
Ireland Observed unless marked
Reference-period note: the Europe tab's cross-country chart uses the earlier H2 2024 semester, where Germany is dearest and Ireland ranks 3rd (€0.37/kWh, behind Germany and Denmark). Irish prices climbed into H2 2025 to the top of the EU range. Both figures are Eurostat (nrg_pc_204); the ranking depends on the semester and consumption band, so the two tabs describe different snapshots, not a contradiction.
Drivers
- Gas-marginal pricing. Irish wholesale power clears at the gas-fired plant — gas sets the marginal price ~70% of hours. Even with renewables supplying 35%+ of generation, consumers pay the gas-priced clearing rate.
- Grid isolation. Limited interconnection (Moyle, East-West, Greenlink, Celtic) means Ireland can't import cheap continental nuclear or hydro at scale.
- Data-centre demand. 32% of national electricity demand (2023). Data centres pay ~0.7c/kWh wholesale-adjacent rates while households pay 7.6c/kWh on the same grid.
What would actually move bills
Gas-marginal-pricing reform (the cleanest lever), faster interconnection, or shifting policy costs from electricity to general taxation. Renewable build-out alone doesn't move bills while gas sets the price.
United Kingdom Observed unless marked
Drivers
- Carbon costs in electricity. UK ETS prices pass fully through to industrial users; continental competitors face lower effective rates.
- Network charges. UK transmission and distribution charges on industrial bills are among the highest in the OECD.
- Policy costs (CfD, Renewables Obligation, Capacity Market) loaded onto electricity rather than spread.
What would actually move bills
Moving policy costs off industrial electricity (the Dieter Helm proposal); a border carbon adjustment to level the field; or industrial flight to cheaper jurisdictions (already underway in steel, chemicals, fertiliser).
Neither pure narrative survives both countries. The "transition is cheap" framing struggles with Ireland's €480/year household premium and the UK's industrial flight. The "transition is unaffordable" framing struggles with the fact that the generators (LCOE) are now genuinely cheap, and the costs sit in system-design choices, not the technology. Both countries show the same lesson: the bill arrives not where LCOE predicts, but where policy and grid design place it.
A note on labels and probabilities
Every forward-looking number on this page carries Observed, Modelled (with source and year) or Aspirational. Probability bands are honest reads from historical build rates, prior transition precedents, and documented political behaviour — not formal forecasts. The gap between CPS and NZE is politically, not technically, defined.