The great oil inflection

Who pays when petrol stops paying the refinery’s rent

First a warning that may sound churlish. I can't find the real numbers I want, they are hiding in plain sight but neither I or any AI tool are producing them. I spent a whole afternoon in a fight with Claude and Gemini who would not admit that the decline we saw in Norway fuel use will accelerate because we are five years further forward. So if I am honest, I personally believe the situation to be far worse than I can reasonably find. Norway is only now entering its steep stretch, down 9% in a single month, diesel falling faster than petrol. That stretch is what China reaches around 2030, and the modellers i suspect backed by petrodollars, are penciling in a gentle 2 to 3% a year through the 2030s and are probably significantly under-cooking it.

The oil market has just been through the largest supply disruption in its history, which is roughly what it goes through every five to ten years apparently! The Strait of Hormuz, the channel for about a fifth of the world’s traded oil, has been effectively shut since early March, whatever the rhetoric out of Washington. Traffic is 93% down at a measly 6 transits a day, that compares to traffic usually around the 100 transits a day mark. Dated Brent traded above $140 in the first week of April. Saudi Arabia says drone and missile strikes have taken about 600,000 barrels a day of its capacity offline. American pump prices rose by nearly a third in ten weeks. This week, with a 14 point draft agreement between Washington and Tehran on the table and a signature possibly days away, Brent slid back to about $86. The administration has talked the price down before, so next week may read differently.

That is the loud story of 2026, with energy prices up about a quarter on the year and Brent averaging $86 against $69 in 2025. The quieter story matters more. The sharpest squeezes were never going to be at the pump, although that’s what we see most obviously. They were in jet fuel and petrochemical feedstock for making plastics and soaps and paint, the parts of the barrel with nowhere else to come from, the parts that will slowly creep into consciousness. Now hold that thought.

Last year the world bought 20.7 million electric cars, more than a quarter of everything that left a showroom. China alone registered 16.5 million new energy vehicles, 48% of its market. From those two numbers a popular conclusion gets drawn: petrol is finished by 2035, and the war in the Gulf is the old world’s last tantrum, not so, not yet at least.

The conclusion is wrong, and the reason it gets read this way is the interesting part. Plenty of petrol will still be flowing in 2035 and well beyond. What ends sooner is harder to see. Petrol stops being the product that pays for the refinery. A refinery makes everything at once, so when its anchor product stops paying, the bill lands somewhere else. Where it lands tells you more about the next twenty years of oil than any date for the last petrol car.

Figure 1. Petrol refining falls from about 27 to 9 million barrels a day by 2050 on the base case. Source: BNEF New Energy Outlook 2025, with the range from OPEC and IEA Net Zero.

The wrong question

Start with the arithmetic that kills the 2035 story. It earns a proper burial and i am here to do just that.

Electric car sales are genuinely astonishing. From under 1% of the global market a decade ago to more than a quarter last year, whilst the steadfastly conservative and murkily funded IEA expecting roughly half of all new cars sold in 2035 to be electric on today’s policies alone, so you know its going to be a far bigger number once you remove teh petrodollar funded research. On that path the global electric fleet reaches as many as 510 million cars by 2035, six times what is on the road now.

Now do the other half of the sum. There are about 1.5 billion cars on the road, and on current growth there will be around 1.7 billion by 2035. A 510 million electric fleet is 30% of that. Seven in ten cars on the road in the IEA’s own central case will still burn petrol or diesel a decade from now. Cars are durable goods. The average car survives well past fifteen years in much of the world, and a petrol car sold this morning will still be queuing at a forecourt in 2040. The showroom is a leading indicator. The road is the market.

Figure 2. Sales share races ahead. Fleet share grinds. Source: IEA Global EV Outlook, baseline beyond 2025.

This is why the forecasters who count barrels for a living put the peaks close and the zeroes nowhere. BloombergNEF has petrol demand peaking around 2030 and road diesel around 2033, that is close enough to be very accurate, with electric vehicles displacing some 19 mb/d of oil by mid-century. OPEC, talking its book, sees total demand rising to 113 mb/d by 2030 and roughly 123 mb/d in 2050. The two camps disagree violently about the slope after the peak. Neither has demand anywhere near zero in a working lifetime but it’s coming.

The transport sector explains why the road matters so much. Global demand runs at about 103 mb/d. Petrol is roughly 26% of it, and petrol with road diesel together just over half. Transport in all its forms takes more than 60%. Road fuel demand worldwide was essentially flat in 2024, an OECD decline of about 200 kb/d cancelled out by the same growth elsewhere. The plateau has arrived. The argument is about the far side.

For calibration, the entire global electric fleet displaced about 1.3 mb/d last year, a little over 1% of consumption, in a year! The IEA conservatively expects that figure to pass 5 mb/d by 2030 on stated policies, the difference between a rounding error and a recession in demand. One marker has already passed unnoticed. Global sales of combustion cars peaked in 2017 and, on the IEA’s figures, will not see that level again.

So the popular question, when does the tank run dry, has a dull answer: not for decades but the fall is rapid and steep. The useful question is different. When does petrol stop covering the fixed costs of the machine that makes it? Those two dates are a generation apart, and the second one is close.

The rent and the important piece of the decline puzzle

A refinery is not a petrol factory. It is a separation and conversion machine that takes one input and produces a slate of outputs in roughly fixed proportions: LPG and naphtha at the light end, petrol, jet fuel and diesel through the middle, fuel oil, bitumen and lubricant feedstock at the bottom. Operators can swing yields between neighbouring products, a few percentage points here and there, more with serious capital. What they cannot do is make one product without making the rest of the barrel show up in some form. Economists call it joint production. Refiners call it Tuesday.

The swing that does exist is expensive and directional. A site built around fluid catalytic cracking is a petrol machine. One built around hydrocracking leans toward diesel and jet. Moving between those worlds is a capital project measured in years and hundreds of millions of dollars, not a dial in the control room. So when demand for one product erodes, the first response is price, the second is exports, and the third is closure.

The commercial structure on top of that chemistry is lopsided. Petrol and road diesel are roughly half of what the world’s refineries sell and considerably more than half of the margin. The rest of the slate rides along. Jet fuel, bitumen and lubricants are passengers whose tickets have been subsidised by drivers for a hundred years. The fixed costs of the site, the crude units, the hydrocrackers, the people, the turnarounds, get recovered mostly across the road fuels, because that is where the volume and the pricing power live.

That is the rent, and the tenant is leaving slowly. When the anchor products sag, the adjustment is not graceful. Refineries carry heavy fixed costs and need utilisation somewhere above 80% before the accounting works. Below that, a site covers its cash costs and stops covering its capital, which is survivable until the next turnaround bill arrives. Take 10 to 15% out of light-end demand in a region and the marginal site does not throttle back. It closes, and its volumes re-tighten the survivors until the next sag claims the next site. The system sheds whole refineries, one at a time, each closure a small local crisis of jobs and fuel security.

By 2050, on BloombergNEF’s base case, the slate has flipped. Petrol falls to about a tenth of demand and road diesel to a sixth, while jet fuel doubles its share and petrochemicals hold theirs, so the captive middle of the barrel climbs toward a third and now the machine gets rebuilt around the products that used to ride along on the fumes for free.

Figure 3. As petrol and diesel shrink, the cost of running the refinery lands on the cuts that cannot leave. Volume shares: IEA, JODI, BNEF.

To me this shape is familiar from power markets. An output you are compelled to produce into a market that no longer needs and wants to pay for it is what negative electricity prices are made of and why we see energy prices crashing in the midday sun.. Petrol is heading for the same status, the “must run” output of a machine that is increasingly running for other reasons. It has a precedent. In the spring of 2020, refinery-gate petrol cracks across the Atlantic basin went negative for several weeks, and refiners paid the market to take petrol away because the alternative was tankage they did not have, prices crashed to lows not seen in decades. Lockdowns made that a curiosity. It will show up in duration before it shows up in averages. The annual average petrol crack will look respectable long after the share of the year spent below breakeven has started to grow because averages have an ability to flatter a dying product especially in the hands of sales teams. The duration curves will do the damage.

The live experiment

The world’s largest car market is already running it. Sinopec, which sells more petrol in China than anyone, has called the national peak: 2023, at about 3.7 mb/d. Demand slipped to 3.6 mb/d in 2024 and kept sliding. By May 2025, Chinese gasoline consumption had fallen back to the level of May 2022, a month in which Shanghai was locked down. The Oxford Institute for Energy Studies (an organisation heavily embedded in fossil energy) puts the base case at 3.0 mb/d by 2030, a fifth below the peak in seven years. Behind it sits the fastest fleet transition ever attempted: new energy vehicles took 47.9% of new vehicle sales in 2025 on CAAM’s count.

Figure 4. The peak is behind China. The decline is a glide. Sources: Sinopec, OIES base case, JODI, CAAM.

Even at half the showroom, the fleet maths still rules. China ended 2024 with about 31 million NEVs against 353 million petrol vehicles on the road, and the Bank of Italy puts the petrol displaced by Chinese electric cars in 2024 at around 0.43 mb/d. Demand is declining at 2 to 3% a year. The most aggressive electrification programme anywhere produces a glide, and a glide is what ruins refining economics, because it strands capacity continuously rather than all at once. Diesel is further down the same path, from 4.7 mb/d in April 2023 to about 4.0 mb/d two years later, eaten by LNG trucks and a property slump. Put the three main transport fuels together and Chinese demand was about 8.1 mb/d last year, already 2.5% below 2021 on the IEA’s numbers.

Watch what Chinese refiners are doing, because it is the preview. The state majors are rebuilding yields around chemicals, with crude-to-chemicals complexes designed to turn 40% or more of each barrel into petrochemical feedstock rather than fuel. The independent teapot refineries of Shandong, gasoline-heavy and margin-starved, are being squeezed toward consolidation, and Beijing’s export quotas decide year by year how much of the surplus is allowed to wash into Singapore. In the world’s biggest market, petrol is already becoming the by-product of a chemicals business. The pricing system has been noticed. Asian gasoline cracks have already had spells trading below naphtha, an inversion that tells a refiner the blending pool is the wrong place for its molecules.

The stranded slate

Europe reaches the breaking point first, for a reason that is almost tidy. Its refining system was built to export petrol, and the export markets are vanishing faster than its own demand.

European refineries have long produced far more petrol than Europeans buy, around 1.33 mb/d of exports, with roughly a third going to West Africa in a trade worth some $17bn a year. Then the Dangote refinery in Nigeria reached full operation: 650 kb/d of crude capacity, up to 300 kb/d of petrol, parked directly on top of Europe’s best customer. West African imports of European petrol fell by a third in a single year, to about 285 kb/d in the first half of 2025. The other big outlet, the US East Coast, is shrinking as American demand plateaus. The sink is draining, and the petrol has nowhere to go.

There is more to Dangote than one plant’s ramp up. The ambition in Lagos is to turn the Atlantic basin’s petrol trade around, south to north rather than the reverse, and every cargo it places is a cargo a European site no longer ships. Demand decline arrives at 2% a year. A mega-refinery arrives all at once.

Figure 5. Cause on the left, effect on the right. Sources: Kpler, FGE, JODI, Argus Media, company statements.

The closures have started. Grangemouth, Scotland’s only refinery and Britain’s oldest, stopped processing crude in the second quarter of 2025. It had been losing around $500,000 a day on 13% of the country’s refining capacity, and it is now an import terminal. Shell closed Wesseling. BP is shutting a third of Gelsenkirchen. Lindsey followed its owner into insolvency. Eni converted Livorno. That is 573 kb/d gone in two years, about 4% of Europe’s capacity, on top of roughly thirty closures since 2000. Kpler and FGE both reckon another 300 to 400 kb/d is at risk before the decade is out. The cull is selective. Simple coastal plants with gasoline-heavy yields go first, which is what the joint-production logic predicts.

The strategic sting is that Europe closes petrol-long refineries while staying short of diesel and jet fuel, deepening its dependence on mega-refineries east of Suez. The spring made the cost of that arrangement vivid. When Hormuz shut, the products that bit first were jet fuel and petrochemical feedstock, while politicians talked about pump prices. Europe is removing the machines that make the captive products and keeping its exposure to the strait that supplies them. Energy security debates in Brussels and London are about to rediscover the refinery, just as the refineries leave.

The captive barrel

So who pays the rent when petrol stops? Well, we can sort the surviving demand by one question: can it leave the refining system?

Figure 6. The captive corner inherits the fixed costs. Sources: IEA, JODI. Placement is the author’s judgement.

Petrochemical feedstock is the biggest growth product and has the best exit. Ethane from American and Middle Eastern gas bypasses the refinery entirely, and crude oil to chemicals plants are built to skip the fuels business so to be fair that growth is real. The captivity is not.

Marine fuel has partial exits in LNG and methanol, slow but genuine.

Jet fuel has none. Sustainable aviation fuel does not change the arithmetic this decade nor the next at the current speed of progress, clean H2 has virtually stopped everywhere. The EU mandate starts at 2% this year and reaches 6% in 2030, and most of those volumes are co-processed inside refineries anyway. Nothing else carries the energy density that gets a wide-body across the Pacific. Jet fuel is captive and growing, and for a century it has been cheap relative to what it takes to make, because drivers were covering the building. That arrangement is ending. The mechanism is undramatic: jet cracks settle structurally wider, and flying gets quietly dearer even in a world of cheap crude. Bitumen and lubricants, captive and unglamorous, follow the same logic. Roads will pay a little more rent too.

The last payer is the crude producer. A barrel is worth what its products are worth. For a hundred years that meant a premium for light, sweet crudes rich in gasoline fractions. As the gasoline machine winds down, the premium migrates toward the grades that suit chemicals and middle distillates, and the quality spreads that organised the tanker map start to reshuffle. Producers whose geology made them rich under the old slate will find the slate has opinions. The marginal bid for crude has been a refinery chasing gasoline for a century, and as that bid fades, crude gets durably cheaper than the old demand curve would have priced it. Comfortable for consumers, corrosive for producer budgets built on $80 assumptions. Cheap crude does not save the refinery either. Feedstock cost was never the problem. The value of the main output is.

Nowhere in this does scarcity appear. The rent certainly does not vanish, it will just move, from drivers to flyers and from forecourts to runways and air travelers will pay for the next 10 years until that too disappears because clean H2 finally arrives. The petrol era ends the way it began, with the price system quietly deciding which fraction of the barrel deserves the capital.

Full circle

It has happened before, in the other direction.

In 1900 the prize was kerosene. Lamp fuel was more than half of what a refinery sold, the product on which Standard Oil was built and the cargo behind the phrase oil for the lamps of China. Petrol was the dangerous light fraction nobody wanted. Refiners sold it for pennies as a solvent, and when they could not sell it they ran it into rivers at night. Then the grid came for the lamp and the Model T came for the buggy. Edison switched on Pearl Street in 1882. Ford shipped the Model T in 1908. By 1920 petrol had taken kerosene’s crown, and the refiners who had treated it as effluent were spending fortunes on cracking units to make more of it. Within a generation the barrel inverted. The machine barely changed its address. The buyer changed everything.

The buyer is changing again. This time the new prize is the sky and the steam cracker, and the residue is the fuel that built the twentieth century. The war in the Gulf will be settled and Hormuz will reopen. Prices will go back to being set by tankers and politics. The quieter inversion will outlast all of it, working through fleet maths and refinery accounts at the speed of depreciation.

Petrol entered the world as a by-product. It will leave the same way.