The Energy Transition, read whole
Two words carry this whole argument. Learn them first.
You do not need a background in energy to follow this. You need two ideas, and they are both simple.
Those two numbers, and how they move, are the entire story. Everything below is either evidence or arithmetic.
TWh is a terawatt hour: a trillion watt hours. One TWh runs roughly 100,000 American homes for a year.
We have never replaced an energy source. We have only ever added one.
That is Jean-Baptiste Fressoz's argument, and it is not political. It is a reading of the historical record by a historian of science who works at the French national research centre, and who has studied how the climate models themselves were built.
His book is called More and More and More: An All-Consuming History of Energy. The title is the thesis. When humanity found coal, we did not stop burning wood. When we found oil, we did not stop mining coal. We used more of everything, every time, and we still are.
He goes further. He says the phrase itself, "energy transition", was invented by atomic scientists after the Second World War, who were imagining a very slow shift over three or four centuries as fossil fuels ran out. It was a guess about the future. Then economists and policymakers pasted that guess onto the past, as though it had already happened three times, and used it to argue the fourth one was routine.
In his words: "This idea of energy transition is based on a false history. It projects a false history on an elusive future."
Every layer of this chart is still growing. That is the whole proof.
This is every source of energy humanity used, from 1965 to today. If transitions happened, old bands would shrink as new ones arrived. They do not shrink. They stack.
Watch the bottom band: wood and biomass. It is still the fourth largest source of energy on Earth, larger than solar and wind combined. Watch the top: coal, which was supposed to have been replaced by oil, is the second largest source and hit its all time high in 2025.
The details that make his case
Coal was built on wood. To prop up the roofs of its coal mines, Britain used more wood than it burned as fuel in the 18th century. Without wood there is no coal, and without coal there is no steam, and no steel. The new fuel did not replace the old one. It consumed more of it.
Oil runs on coal. Building a car in the 1930s took about seven tons of coal, nearly as much as the oil the car would burn in its whole life. Every ton of oil burned in interwar Britain needed about 2.5 tons of coal to make the steel infrastructure around it. Oil and coal are partners, not rivals.
Materials never become obsolete. Technologies do. Nobody uses a steam engine to cross an ocean. But the materials underneath them keep expanding. The only raw material to genuinely decline from being replaced is ship wool, displaced by nylon, and that is not an environmental victory. Asbestos fell because it was banned, not replaced.
The renewable flagship runs on imported wood. Drax in Britain is the country's single largest corporate carbon emitter. It burns over 8 million tons of wood chips a year, shipped from the United States and Canada, to make about 1.5 percent of Britain's energy. Two centuries after all those supposed transitions, Britain burns four times more wood for energy than it did in the 18th century. And Drax's emissions are not counted against the national carbon budget, because burning wood is classified as renewable.
Where the world's energy actually comes from
Here is the same story as shares, which is where the argument really bites. Fossil fuels went from about 80 percent of world energy in 1965 to about 81 percent in 2025. Six decades, every renewable technology invented, and the share barely moved.
A thing that has never once happened, on schedule, in sixty years. Betting against that is just reading the record.
We will not pretend this is going well, because it is not. Fressoz is right on the facts, and we agree with his gloom on the mechanism.
The honest position is this: every card in this deck would have to fall face up. Not most of them. All of them. The build has to keep compounding for decades. Governments have to hold policy steady across changes of power. Supply chains for copper, lithium, steel and rare earths have to expand, mostly in countries that are not friendly to each other. The fossil system has to keep supplying the energy to build its own replacement. Nobody has to invade anybody. No financial crisis. No drought that turns a region against its own government. The probability of all of that holding at once is not low. It is near zero, and pretending otherwise is how we got here.
The single strongest number in his favour: solar and wind together are 3.11 percent of world primary energy after twenty years of the fastest energy cost collapse in history. Coal, oil and gas are 80.85 percent. That is not a transition in progress. That is a rounding error with good public relations.
And the second strongest: in 2025, for every unit of new energy the world drew from fossil fuels, it drew 0.66 units from new solar and wind. Less than one. The additions are still losing.
But the gap is closing, and fast. Eight fold in seven years.
Here is where a reader who trusts only the last section would go wrong. The case for Fressoz rests on the shape of the system today. The case against him rests on how fast that shape is changing.
This chart shows the raw tonnage added every single year, fossil in dark and solar plus wind in green. The red line is the ratio between them, on its own right hand axis.
That red line is the number that matters. In 2018 it was 0.08. In 2022, 0.25. In 2024, 0.31. In 2025, 0.66. Solar and wind additions are closing on fossil additions eight times faster than seven years ago, and the trend is not flattening.
So the honest answer to "is he right" cannot come from the past, and it cannot come from extrapolating the past forward either. It comes from one question: what does it cost, in energy, to build this fast? Which brings us to the loop.
The transition is powered by the thing it is replacing. Here is the arithmetic.
This is the part almost nobody accounts for, and it is where Fressoz's instinct turns out to be measurably correct. There is a name for it in the energy literature: energy cannibalism.
The idea is simple. To grow a solar and wind industry, you must spend energy building panels, turbines, steel, concrete and transmission lines. That energy has to come from somewhere, and today it comes overwhelmingly from fossil fuels. So a growing renewable industry consumes a slice of its own output just to keep growing, and the faster it grows, the bigger the slice.
The rule reduces to something you can write on a napkin.
The cannibalism tax
Tax on renewable output = growth rate × energy payback time. A solar panel takes roughly one to three years to return the energy used to make it. Grow the fleet at 19.5 percent a year, and the tax comes to between 20 and 59 percent of everything those panels generate. That portion cannot displace a single lump of coal. It is spent making more panels.
Measured against real numbers, solar and wind together have been growing at about 19.5 percent a year since 2010. Using published payback times of one to three years for solar and roughly five months for wind, the tax lands near one third of output. Which is precisely why fossil energy additions rose by 1,273 TWh in 2025 at the same time as renewable additions rose by 843 TWh. Both went up. That is not a contradiction in the data. That is the loop, visible.
So the tax is real, and it is not fatal. The tax is a percentage. Growth is a compounding quantity. A discounted compound still compounds. The tax slows the crossover. It does not stop it. What it does do is set a ceiling on how fast the build can go, and that ceiling falls as the system gets weaker.
The ceiling nobody talks about
Energy return and energy payback are two ways of saying the same thing. A high return means a short payback. A long payback means any given growth rate demands a bigger share of output. So there is an optimum speed, past which building faster wastes more energy than it delivers.
At an energy return of 12 to 1, which is roughly where the system sits today, solar and wind can hold about 20 percent annual growth. That is right about where they are. At 3 to 1, the ceiling falls to about 5 percent. And 3 to 1 is where the energy literature expects a fast transition to end up.
He is wrong that it cannot happen. He is right that it will not be the one being sold.
Put the loop into the model, let the growth rate fall when the ceiling makes it fall, and run it forward. This is what comes out. It is a model, built on the published numbers above, not a forecast.
Fossil energy peaks in the early 2030s and falls below its 2025 level around 2038. It halves by the early 2050s and approaches zero in the 2060s. Solar and wind keep compounding the whole time, just more slowly than today, because the energy return keeps falling underneath them.
Where we agree with him
Nothing has been replaced. The phrase is a guess about the future pasted onto the past. New supply has so far expanded consumption rather than displacing it. A third of renewable output is currently spent reproducing renewables, and that spend is powered by fossil fuel. All of that is his argument, and all of it holds up.
Where we dissent, and the data is why
He treats a pattern in history as a law. History is strong evidence about mechanisms and weak evidence about a genuinely new exponential process, and solar doubling every two and a half years for eleven straight years is new. The transition can happen. What the numbers say is that it arrives late, slow, and materially poorer than advertised, with the system's energy return near 3 to 1, which the literature says is below what an industrial society needs to run. That is not green growth. That is a lower energy world, reached the hard way.
So the final position is this. We accept his gloom because the probability of everything going right at once is near zero, and because the record up to this point honestly read gives no other answer. We dissent from his fatalism because the same resource limits that constrain his story constrain the optimists harder, and the ceiling that forbids endless exponential growth does not forbid a slow, modest, real one. He is right about the world we are in. He is wrong to say there is no way out of it, and wrong to imply that if there were, it would look like what the brochures promise.
He says the numbers rest on machines that do not exist. The assessment agrees with him, and then says something worse.
Fressoz is not only arguing with the history books. He is arguing with the assessment itself. It is worth setting out exactly what he accuses it of, because the accusation is specific, and so is the reply.
What he says about their numbers
He starts with who writes the mitigation chapters. Working Group III is, in his words, the economists and the modellers, and he says neoclassical economics has played a key role in that expertise. He names the two tools he believes do the work. The first is time discounting, which makes a cost landing after 2050 look small in present value, and which is why a machine nobody can build today can still come out cheap in a model. The second is the learning curve, the assumption that a technology gets cheaper as it is made, which turns a hope about the future into a number in a spreadsheet. His conclusion is that both bias the answer toward one grand technological fix and away from questions of sufficiency.
He then supplies a history to match. He says the 2001 assessment found that capturing carbon and burying it made electricity dearer than nuclear, so what was the point, and that a 2005 special report on carbon capture and storage turned the same technology into a green one, after which the literature on it poured out. The reason, he argues, is that oil and cement companies had pushed the technology since the early 1990s and had manufactured a scientific discipline around it, and that the assessment then endorsed what industry had already framed.
Then the part that carries the most weight. After Paris, he says, researchers went and read the Working Group III scenarios and found that all of them carried huge amounts of negative emissions. His question is the sharp one: what is the meaning of an agreement that relies on technologies which do not exist.
What the assessment actually says
On that last point he is not misreading. He is quoting its conclusions back at it, and it concurs.
The report states, at high confidence, that the deployment of carbon dioxide removal to counterbalance hard-to-abate residual emissions is unavoidable if net zero carbon dioxide or greenhouse gas emissions are to be achieved. Unavoidable is its word, not his. In the pathways that hold warming to 1.5 degrees with no or limited overshoot, it puts cumulative removal by bioenergy with capture at 30 to 780 gigatonnes of carbon dioxide over the century, plus a further 0 to 310 from direct air capture. Those are the machines the agreement leans on.
Then comes the sentence that settles it. The report states, again at high confidence, that afforestation, reforestation, improved forest management, agroforestry and soil carbon sequestration are currently the only widely practised carbon dioxide removal methods. The two technologies carrying the numbers above, bioenergy with capture and direct air capture, are not on that list. Fressoz says there is not the beginning of that industry. The assessment says the same thing in its own wording.
The number neither side is using
There is a figure inside the Technical Summary that neither the doomers nor the promoters quote much, and it reframes the dispute.
Assuming existing plant runs on roughly historic patterns of use and retirement, the assessment estimates future carbon dioxide from existing fossil fuel infrastructure alone at 660 gigatonnes, range 460 to 890. Add the plant already planned and it rises to 850, range 600 to 1100. Against that, the total carbon dioxide the world may still emit between 2020 and reaching net zero on a 1.5 degree path is 510, range 330 to 710.
Read those sentences together and the finding is not that the transition is difficult. It is that the budget is already spoken for. The plant standing today, living out a normal life, commits more carbon dioxide than the 1.5 degree target has left, and the plant already on drawing boards commits more still. The honest caveat is that the ranges overlap, so this is likely over rather than certainly over, and the 2 degree budget of 890 is only just consumed by what is already planned. The direction is not in dispute.
Where the objection still fails
Everything above supports him, and it would be easy to stop there. Two things stop us.
The first is that he is attacking a claim the assessment does not make. He writes as though the models assert that the transition is underway and the technology exists. They do not. The same pages that depend on removal state plainly that removal is unproven, that scaling it depends on solving feasibility and sustainability constraints, and that low carbon electricity deployment is currently insufficient to meet the goals, with solar and wind still holding under 10 percent of generation. This is not a body of work hiding its weakness. It publishes the weakness he says is buried, which means a critique of its optimism has to reckon with its pessimism too.
The second is that his causal story runs the wrong way. He treats the 2005 report as the moment industry captured the science. The report is real, and it did assess capture and storage as a mitigation option. But the capture he describes did not deliver what he predicts. Two decades on, the assessment's own finding is that the only removal practised at scale is trees and soil. If industry captured the models, it did not capture the ground.
And the budget arithmetic cuts against his fatalism hardest of all. If the standing plant has already committed the 1.5 degree budget, the binding constraint is not whether sunlight can run a factory. It is what happens to the machines already built. That is an argument about retirement dates and about not adding more, which is a fight that can be won inside a decade, rather than an argument about whether a new energy system is possible at all.
What this changes about the argument
It moves the question off the ground he chose. He wants the debate to be whether an energy transition can happen. On the evidence assembled here it can, slowly and at a lower energy return than the brochures promise, which is already a hard enough answer. But the assessment he mistrusts supplies a harder fact than his own thesis does: the carbon is spoken for by plant that already exists. That makes the near term decisive and the far horizon almost beside the point, and it makes 2050 the wrong place to look. The next ten years are not mainly about building the replacement. They are about not making the problem worse.
If the timeline is forty years, your land is the only part of it you control.
Strip out the politics and the disagreement narrows to one practical question. Should a household treat grid decarbonisation as something happening to it, or as something it does itself?
The model above says the transition is real and slow, and that its end state is a world with less energy per person than today, not more. Both readings of the evidence lead to the same place. If the transition arrives on schedule and cheaply, home energy independence was redundant. If it arrives slowly and painfully, home energy independence was the only version of it that was ever a promise instead of a plan.
Either way, households buy time and land buys duration.
There is one caution that follows from Fressoz rather than against him. A solar array behind a grid that is still 80 percent fossil is a real reduction in the fossil energy you buy, and a real increase in your resilience. What does not survive his critique is any argument that installing panels licenses you to consume more. His whole point is that new supply has historically expanded consumption rather than replaced anything. Do not build the array so you can run more. Build it so you need less.
Six talks, in his own words. Start with the short one.
These are embedded from YouTube so the speakers and the channels get the watch. Nothing is re-hosted here. If the topic is new to you, watch the 19 minute talk first, then the interview this article is built on.
All links verified live on 2026-10-09. Titles, channels, durations and view counts read directly from YouTube.
Every number on this page, and where it came from.
We do not print an energy figure we cannot point at. Here are the sources behind each claim, and the places where we are estimating rather than quoting.
| Figure | Value | Source |
|---|---|---|
| World primary energy, 2025 | 177,865 TWh | Our World in Data, global primary energy, Energy Institute series, substitution method |
| Fossil share of primary energy | 80.85% | Same dataset, computed |
| Solar and wind share | 3.11% | Same dataset, computed |
| Solar growth, 2010 to 2025 | 34.2% a year | Same dataset, compound rate |
| Solar and wind growth, 2010 to 2025 | 19.5% a year | Same dataset, compound rate |
| New solar and wind per unit of new fossil, 2025 | 0.66 | Same dataset, year on year additions |
| Solar energy payback time | 1 to 3 years | Solar Energy Materials and Solar Cells, 2013 (334 citations) |
| Solar lifetime energy return | 9 to 17 to 1 | Solar Energy, 2000 (119 citations) |
| Wind energy payback time | 0.44 to 1.9 years | Environmental Science and Technology, 2019; 4,161 sites across northwest Europe |
| Cannibalism rule and optimum growth rate | growth below 1/(2 x payback) | Pearce, energy payback for energy systems ensembles during growth, ISSST 2010 |
| System energy return on a fast transition | 12:1 down to about 3:1 | Energy Strategy Reviews, 2019 (220 citations), dynamic EROI and material requirements |
| Coal mine pit props, 18th and 19th century Britain | 4.5 million m³ timber | Fressoz, More and More and More, as presented in the interview below |
| Carbon dioxide removal | unavoidable | IPCC AR6 Working Group III, Summary for Policymakers, statement C.11 (high confidence) |
| Only widely practised removal methods | trees and soil | Same report, statement C.11.1 (high confidence) |
| Cumulative capture from bioenergy, 1.5°C pathways, 2020 to 2100 | 30 to 780 GtCO₂ | Same report, statement C.3.5 |
| Emissions cut needed by 2030 for 1.5°C | 43% (34 to 60%) | Same report, statement C.1.1 |
| Carbon budget left from 2020 on a 1.5°C path | 510 GtCO₂ (330 to 710) | IPCC AR6 Working Group III, Technical Summary |
| Committed by existing fossil plant | 660 GtCO₂ (460 to 890) | Same Technical Summary |
| Committed by existing and planned plant | 850 GtCO₂ (600 to 1100) | Same Technical Summary |
| Drax wood chip consumption | over 8 million tons a year | Fressoz, in the same interview |
What is measured, and what is modelled
Measured and sourced: every energy quantity, share, growth rate and ratio above. These are computed directly from the published dataset and can be checked line by line.
Published but contested: the energy payback times and energy returns. Analysts disagree on where to draw the system boundary, and a wider boundary lowers the number. We used the range, not a single figure, and we picked the conservative end where we had to choose.
Our own construction: the trajectory chart and the tax percentage. These apply the published rules to the published parameters. They are a model, not a forecast, and the peak and crossover years depend on the assumptions holding for forty years, which they will not exactly. Treat the shape as the finding, not the dates. The direction is what the argument turns on, and that direction held in every version we tested.
Not engaged by the source: the interview never addresses the growth rate argument. We searched the full transcript for the vocabulary it requires, exponential, doubling, growth rate, logistic, S curve. Zero hits. He is making a claim about history and the strongest objection to it is a claim about rates, and that objection is never raised or answered. That gap is why this page exists.
The book, and the interview
- Jean-Baptiste Fressoz, More and More and More: An All-Consuming History of Energy. Penguin, 2025. The argument in full, with the archival work behind it.
- Fressoz interviewed by Rachel Donald, "The Energy Transition is a Pipe Dream", Planet: Critical, published 20 March 2025. 52 minutes. The source for this article. Video embedded above.
- Nate Hagens, "The Unpopular Reality about Energy Transitions", The Great Simplification episode 162. A longer and more technical conversation.
- Vaclav Smil, Energy and Civilization: A History. The standard reference on how slow energy transitions actually are, and the strongest independent support for Fressoz's history.
- IPCC, Sixth Assessment Report, Working Group III, Summary for Policymakers and Technical Summary. Every removal figure and every committed emissions figure above comes from here. Free to read at ipcc.ch.
- IPCC, Special Report on Carbon Dioxide Capture and Storage, 2005. The report Fressoz puts at the centre of his account. Free as well.
- Our World in Data, global primary energy. The dataset behind every chart on this page. Free, downloadable, and worth reading directly.