Energy Reality
What actually powers the world, what the AI boom is doing to the grid, and the widening gap between climate pledges and what grids can physically deliver.
(down from 94% in 1965)
(renewables plus nuclear, 2024)
(about 4% today; the projection range is wide and methodology-sensitive)
A note on data freshness. Global energy data is structurally lagged. The Energy Institute publishes its world review each June with the prior year's figures, and other agencies (Ember, the International Energy Agency, the US Energy Information Administration) cross-check and harmonise the numbers over the months that follow. The series on this page run through 2024, the most recent year for which the global picture is complete. For slow-moving structural questions like the energy mix, the trend matters more than the latest single year, and the trend has been consistent for decades.
The raw mix
When most people hear "energy transition," they picture renewables replacing fossil fuels. The reality is closer to renewables being added on top of fossil fuels while overall energy use keeps rising. Coal, oil, and natural gas together still supply about four out of every five units of energy the world uses. That share has come down - from roughly 94% in 1965 to 81% today - but the pace has surprised almost everyone, including the agencies that write the official forecasts.
Why so slow? The world keeps using more energy. Cleaner sources have been added in large quantities; they just have not been enough to outrun the demand growth. Solar generation has roughly doubled every three years for the past decade, and wind has grown at a similar pace. But the developing world has been getting richer at the same time, and the first thing a growing economy does is use more electricity, more transport fuel, more industrial heat. Solar and wind have been racing against population growth, urbanisation, and rising living standards in countries with billions of people. They are winning on share, but only slowly.
There is also a physics reality often missed in the political conversation. Solar panels and wind turbines produce electricity. They do not directly produce the high-temperature heat used for steel, cement, glass, fertiliser, and most heavy industry. Roughly half of the world's total energy goes to things that are not electricity at all - transport fuel, industrial heat, and the raw oil and gas inputs used to make plastics, fertiliser, and a long list of everyday materials. Replacing fossil fuels in those applications is harder than replacing them in electricity generation, and the technology to do it at scale is mostly still being built.
Electricity is moving faster than the rest
Inside the slow overall picture, the electricity sector has been moving meaningfully faster. About 41% of the world's electricity now comes from sources that do not burn fossil fuels - roughly 30% from renewables (mostly hydro, wind, and solar) and roughly 10% from nuclear. The interesting feature of the chart below is the shape, not the level: the share was nearly flat from the mid-1980s through about 2015, then started climbing.
Two things explain that long flat stretch. Nuclear power, which had been growing through the 1970s and 1980s, plateaued after Chernobyl and went into a slow decline after Fukushima in 2011. Hydropower kept growing in absolute terms but stayed roughly steady as a share of a growing electricity total. Wind and solar were tiny. So even as wind and solar grew quickly, they were mostly replacing the nuclear and hydropower share rather than the fossil share.
The reason the curve has steepened over the last five years is that solar and wind are finally large enough in absolute terms to outpace both nuclear's slow decline and the world's rising electricity demand. The cost crossover is the engine here: solar and wind are now the cheapest forms of new electricity in most of the world - cheaper than building a new coal plant or a new gas plant. That happened around 2020 and has only widened since. Once new builds are cheaper, even countries with no climate policy at all start adding renewables, simply because they are the lowest-cost option for new generation capacity.
But two things complicate the picture. First, solar and wind are intermittent: the sun does not shine at night, and the wind does not blow on demand. Running a modern grid on them requires either large amounts of storage (mostly batteries, for now), backup capacity (typically gas plants that sit idle most of the time), or long-distance transmission lines that can move electricity from where the wind is blowing to where the demand is. All three are still being built, slowly. Second, replacing the existing fleet takes decades. A coal plant built in 2010 is expected to run until 2050 or beyond. Even if no new fossil plants were started tomorrow, the existing ones keep running, and the share of low-carbon electricity only accelerates after the existing fleet retires.
How national grids actually look
The "energy transition" is one phrase covering very different national stories. Each grid below tells a different version of the same set of choices about geography, history, cost, and politics. The shares are approximate and shift year to year, but the structural picture is steady.
The takeaway is that grids look very different country by country, mostly because of geography and history rather than ideology. A country with mountains and rivers can run on hydropower. A country that built nuclear in the 1970s runs on nuclear. A country that found cheap shale gas runs on gas. Solar and wind are now reshaping the bottom of every national mix, but they are layered on top of these older choices, not replacing them overnight.
The new demand floor: AI, electric cars, and the grid
For most of the last decade, electricity demand in the United States and Europe was roughly flat. Population growth was slow, energy efficiency was improving, and heavy industry was either declining or moving overseas. Forecasters got used to drawing a flat line. That assumption broke around 2022. Three things hit at once: artificial intelligence companies started building large data centers, electric vehicles started showing up in real numbers on the road, and households began replacing gas furnaces with electric heat pumps.
The data center surge is the most concentrated of the three. Training and running large artificial intelligence models requires unprecedented amounts of computing power, and computing power is just electricity in a different form. A single large AI training facility can use more electricity than a small city. The biggest technology companies are now competing for the same kind of grid connection that aluminium smelters and steel mills used to compete for. In some places, utility companies are warning customers that data centers are using up the available grid capacity, and new homes or businesses are facing connection delays of three to five years.
The implications go beyond the headline demand numbers. The grid was built to deliver electricity from large central power plants outward to homes and small businesses. A grid that has to handle steady high demand from data centers, mostly-overnight charging from electric vehicles, and rooftop solar panels feeding power back the other direction is a different machine. Upgrading it requires transmission lines, transformers, and substations, and almost none of these have been getting built fast enough. The bottleneck is no longer money or technology - it is the slow process of permitting, land use, and political fights over who has to host the new infrastructure. A new transmission line in the United States typically takes ten to fifteen years from first proposal to operating. The renewable build-out is currently being held back less by panel costs and more by the wires that would carry the electricity to where it is needed.
The China paradox
No discussion of global energy is honest without a section on China, because China alone accounts for roughly a third of the world's annual electricity generation and an even larger share of new generation being built each year. The widely-reported headline is that China's coal share has been falling, which is true: it has dropped from over 80% in 2007 to under 60% today. The widely-missed part is that this is a share figure. The total has roughly tripled.
What this means in practice is that China is adding solar and wind faster than any country in history while also adding new coal plants faster than any country in history. Both numbers can be true at once because Chinese electricity demand has been growing so quickly that even an unprecedented clean buildout has not been enough to cover it. The Chinese government's official approach has been to use coal as the reliability backbone (the dispatchable layer that runs whenever wind and solar are not enough) while pouring investment into the renewables that will, eventually, replace it.
Whether and how this transition completes is one of the largest open questions in global energy. China's emissions matter enormously for the world's climate trajectory, and China's manufacturing scale matters enormously for everyone else's costs - solar panels, batteries, electric vehicles, wind turbines, and the materials behind them are now mostly Chinese-made. Most of the cost decline in clean energy hardware over the last decade traces back, in some way, to Chinese factories scaling up. A reader who pictures the energy transition as a Western project misses where most of the actual building is happening.
The paths from here
There is no single answer to the energy question, and every advanced economy is some combination of the choices below. Each is listed with the question that actually matters: does the political and physical reality support this path, or is it a slogan?
Build out renewables faster
Continue scaling solar, wind, and battery storage. The cost economics keep improving each year, and most countries are already on this path in some form.
Will it happen? Yes, almost everywhere. The harder question is whether the build-out can keep up with rising electricity demand from data centers, electrification, and growing economies, and whether the supporting grid investment arrives in time. In many places it currently is not.
Build the grid that the buildout requires
Transmission lines, transformers, substations, the capacity to move electricity from where it is generated to where it is used. This is the actual chokepoint that gets less political attention than the headline generation numbers.
Will it happen? Slowly. Permitting and local opposition delay every major transmission line in the developed world by five to fifteen years. Until that changes, much of the new generation will end up in places where it cannot fully reach the demand. Reform of permitting law is now being discussed seriously in the US and Europe; whether the political will follows is another matter.
Restart nuclear
Build new reactors, including smaller modular designs intended for faster construction. China is doing this aggressively. Western countries are talking about it more seriously than they have in decades.
Will it happen? Slowly in the West, faster in China and a few other countries. Each new Western reactor in the last twenty years has cost two to three times its budget and run a decade behind schedule. Until the cost and timeline problem is solved, nuclear will remain a small share even of new builds in the developed world. The smaller modular designs are promising on paper but mostly still in the prototype stage.
Stretch the materials supply chain
Mine and refine more lithium, copper, nickel, and rare earth metals to support batteries, transmission lines, and wind turbines. The clean energy build-out is much more materials-intensive than fossil fuel infrastructure was.
Will it happen? The mining and refining is real and growing. The deeper question is who controls it. Today, China refines around 80% of the world's rare earths and a large share of lithium and processed solar materials. That dependency has become a serious geopolitical conversation, and it will reshape industrial policy in the United States and Europe over the next decade.
AI demand reshapes the grid
Build new generation capacity specifically to power data centers, place AI training facilities near cheap and reliable electricity, and restart retired plants where it makes sense. AI's electricity demand is now large enough to be a planning input for utilities.
Will it happen? Already happening. Microsoft has signed an agreement to restart a retired nuclear plant in Pennsylvania to power its data centers. Amazon has bought a nuclear-powered data center campus. Smaller companies are building their data centers right next to hydropower dams and geothermal plants where electricity is cheap and reliable. The 2030 grid will look noticeably different from the 2020 grid because of this, and the change is being driven by tech-company demand more than by climate policy.
Energy as foreign policy
Sanctions on Russian and Iranian oil, liquefied natural gas exports as a diplomatic tool, pipeline routes that double as alliances. Energy and security policy have merged in ways they had not since the 1970s.
Will it happen? Already happening, and intensifying. The Russian invasion of Ukraine in 2022 reshaped European gas supply almost overnight. Sanctions on Iranian oil have become routine. The United States is now the world's largest exporter of liquefied natural gas. Expect more of this rather than less, and expect energy markets to remain unusually political for the foreseeable future.
Use less energy in the first place
Better building insulation, more efficient appliances, lighter cars, smarter industrial processes, and behavioural changes. Reduces the absolute demand growth so the build-out has a smaller target to cover.
Will it happen? Slowly and unevenly. Efficiency works and has been working for decades - the energy used per dollar of economic output has been falling steadily in most developed countries. But it has been outpaced by rising overall consumption. New efficiency gains help at the margin and rarely shrink the absolute totals on their own.
The realistic forecast is, again, a mix. Renewables will keep growing faster than any other source. Coal will keep being used in much of Asia for another twenty to thirty years. Natural gas will be the dominant new fossil capacity in the West because it pairs cleanly with intermittent renewables. Nuclear will grow, slowly. And data center demand will quietly become one of the most important single drivers of grid policy in the United States and Europe.
Where energy analysts disagree
The framing above is the mainstream reading, but it is not the only credible one. Several careful voices push back from different directions, and the data does not rule them out. A reader who hears only one of these views is getting an incomplete picture.
The transition is faster than the official forecasts say
The International Energy Agency has consistently underestimated solar growth for two decades. Each year's forecast has been followed by reality that beats it by a wide margin, and the gap has widened rather than narrowed. By the time the long-running models catch up, solar may already be the largest single source of electricity globally.
Held by: the analysts at Ember Climate, the Rocky Mountain Institute (a clean-energy research group), Bloomberg's energy team, and a growing fraction of the academic energy economics literature. The data on solar costs supports their case: each successive cost decline has come earlier than even the optimistic scenarios predicted.
There is no real "transition" - only an addition
The historian Vaclav Smil has argued for years that previous so-called energy transitions did not actually replace older sources. Oil did not replace coal at the global level; the world simply uses more of both. The same may be true now: solar and wind are being added on top of fossil fuels, and the world keeps burning more of everything in absolute terms. The story of replacement is wishful thinking dressed up as analysis.
Held by: Vaclav Smil and the broader energy-realism school. The data on absolute global coal use, which has been roughly flat or rising over the past decade, supports the claim more than the share-of-mix data does.
Renewables cannot run a modern grid without large amounts of backup
Solar and wind are intermittent. The grid has to deliver electricity on demand, every second of every day. Until storage costs drop several more times, you need fossil-fueled backup capacity sitting idle most of the time, and the cost of that backup is what most cheerful analysis under-prices.
Held by: many practising utility engineers, parts of the Manhattan Institute, and the natural gas industry. The argument has weakened year by year as battery costs have fallen, but the basic physics has not gone away. Texas in February 2021 and parts of Europe in summer 2022 are real-world reminders that grid reliability is harder than it looks.
The bottleneck is materials, not money or political will
To shift to electricity and remove fossil fuels on the timeline most countries have committed to, the world has to mine and refine more copper, lithium, nickel, and rare earth metals in the next twenty-five years than it has mined in all previous human history combined. There simply are not enough mines, and opening a new one takes ten to fifteen years even when nothing goes wrong. Money does not solve this; physics and geology do.
Held by: Daniel Yergin and the analysts at S&P Global Commodity Insights, plus the International Energy Agency's own materials team. Their position is not that the build-out is impossible, but that the official timelines have not honestly accounted for the physical supply chain underneath them.
Energy abundance is achievable, and the politics has been the real problem
A combination of next-generation nuclear, advanced geothermal, and aggressively-built solar and storage could, in principle, deliver cheaper and more reliable electricity than today, in roughly two decades. The reason it has not happened is permitting, regulation, and cultural opposition to large infrastructure - not technology and not money.
Held by: the Breakthrough Institute, Mark Mills, and a cluster of analysts who hold that abundance is the right framing rather than scarcity or restraint. Their critics counter that "if only we built more" understates how genuinely hard large infrastructure is to deliver on time in a democracy with property rights.
None of this dismisses the structural concerns earlier in this piece. It says the speed and shape of the transition depend on choices that have not yet been made. The reading that fits both the mainstream concerns and these counter-arguments: the energy story is moving in a constructive direction, but slower than the climate timeline most countries have officially committed to. Renewables are winning on cost and will keep growing. Fossil use is plateauing in the developed world but still rising globally. Whether the world's energy use peaks this decade depends on choices about grid investment, mining, nuclear, and permitting that have not yet been made.
What this means for you
Energy decisions touch your daily life through electricity bills, gas at the pump, heating and cooling costs, and the value of the assets you own. The direction is fairly clear; the timing varies by country and by region.
If you are choosing where to live
Electricity prices vary by a factor of three or more across US states, and by a factor of five across European countries. Reliability also varies; some grids brown out during heat waves and cold snaps, others almost never. Before signing a long lease or buying a home, look up the local utility's recent rate history and any planned price increases. A house in a region with cheap and reliable electricity is structurally easier to live in for the next thirty years than one in a stressed grid.
If you own a home
Heat pumps are now the most efficient way to heat and cool most homes, and they run on electricity rather than gas. If your gas furnace or air conditioner is more than ten years old, the next replacement is worth comparing against a heat pump on total operating cost, not just the upfront price. Solar panels still pay for themselves in most US states within seven to twelve years, depending on local rates and how much sun your roof gets. Battery storage adds resilience during outages but pays back more slowly.
If you drive
Electric vehicles have crossed price-and-range competitiveness with gas cars in most segments. Charging infrastructure remains uneven; if you live in an apartment without home charging, the math is harder than for a house with a driveway. The used-car market for electric vehicles is still volatile because battery health is hard to assess from outside, so be careful with older used electric cars and ask for a battery report before buying.
If you invest
The companies that build and operate the grid - utilities, transformer manufacturers, transmission specialists, copper miners - are positioned for steady demand growth over the next two decades. Pure-play oil and gas companies remain profitable but face slowly shrinking long-term demand for their core product. Lithium and rare-earth mining is volatile but structurally important. None of this is investment advice; it is observing where the demand curves point. Concentration risk is real - the world's clean-energy supply chain currently runs through China at multiple points, and that fact is increasingly priced into Western industrial policy.
If you work in technology
Electricity is becoming the new constraint on what artificial intelligence can do. The companies that solve the grid-and-power side of AI infrastructure will quietly capture as much value as the model labs themselves. The bottleneck is no longer chips alone - it is power, cooling, and grid connections. If you are early in your career, the energy-and-tech intersection is one of the deepest pools of opportunity over the next decade, and the smartest people in technology have started to notice.
The mechanics behind this
The energy story sits on top of three deeper mechanisms that show up across this site. If the analysis above depends on ideas you want to understand first, these fundamentals make the conversation more legible:
