The Expensive Electron

Angola spent about $3.8 billion in 2022 holding down the price of fuel, which was 92 percent of what it spent that year on education and health together. Sonangol was buying the petrol at 583 kwanza a litre and selling it at 160.¹ So from 2023 the government started unwinding it and the pump price went from 160 kwanza to 300, which reads like a serious reform and was one.²

Then price it in dollars, which is the money the cars and the spares and the diesel are actually bought in. A litre cost about 33 US cents at the start of 2020, and it costs about 33 US cents now.³ The kwanza fell at the same speed the price rose. Six years, a politically expensive reform, and for the driver nothing moved.

T'Leva had already bet the other way. It is a Luanda ride-hailing company, and in 2020 it announced more than $20 million with Ledo Holding to put Chinese electric cars on the road. By August 2021 it had over 400 of them running with about 700 people on the payroll. The plan from there was 10,000 active electric vehicles by 2025, Benguela and Lubango and Huambo after Luanda, then every country in SADC.⁴

It got to about four percent of that. The Android listing is gone, and the iPhone app has 35 ratings on Angola's App Store with its last build shipped in 2023.⁵ The company is still running.

But the cars were never the problem. A driver in Luanda in 2021 was buying petrol at 160 kwanza that the state was paying 583 for. No electric vehicle competes with a subsidy that size, however well you run the business around it.

What has changed since is upstream. Manufacturing is mostly solved and so is logistics, up to a point, which means both of those have been commoditised. Chinese manufacturers exported a record 540,000 electric vehicles in July 2026, the biggest month BYD has ever had, while sales inside China fell 12 percent over the first seven months of the year.⁶ So there are a lot of cheap and genuinely good cars looking for somewhere to go, and the conclusion people draw from that is that poorer countries will skip the combustion era the way they skipped landlines.

I have argued before that this reading fails for physical AI, and it fails here for the same reason. It takes a cost story and calls it a technology story. You can get cheap cars and cheap transport now. What is not cheap in emerging markets is the policy to get those cars from the port into the hands of end users, and the local infrastructure for electrons to get into the cars, on top of the problems with financing. The solved problems only make the unsolved problems much more evident.

The entry condition

So the first thing I want to know about a market is what a litre costs there against what a kilowatt-hour costs, in the money the driver earns. Income tells you less than you would expect, and so does urbanisation, and so does the state of the grid.

Angola is a good example of why the crude version of that test misleads. Petrol there is still among the cheapest anywhere, and electricity is cheaper still, somewhere between two and a half and three US cents a kilowatt-hour.⁷ That is third lowest on the continent. So the electric case works there today because the state subsidises the electron harder than it subsidises the litre, not because petrol ever got expensive.

Tajikistan runs the same test from the other end, with the most expensive petrol in Central Asia and close to the cheapest electricity in the world. In 2022 it had 120 electric vehicles against about 445,000 cars, and in October 2022 it exempted them from every duty and tax for ten years. By 2025 it had 34,354 of them, which is around 5 percent of everything on the road.⁸

So the ratio is necessary but it is not sufficient, which is the part I keep having to explain. Nigeria removed its petrol subsidy in 2023 and passes that test on paper, and Nigeria is also where the argument falls apart.

What a kilowatt costs

Take the car that is actually being bought at this end of the market. A Wuling Bingo is rated at 9.58 kWh per 100 kilometres, so call it 12 in stop-start traffic with the air conditioning on.⁹ On Nigeria's Band A tariff at 209.50 naira a unit, that is about 2,500 naira to move the car 100 kilometres, against 9,600 for eight litres of petrol. Nearly four times cheaper, and on that arithmetic nobody would hesitate.

Except Band A means twenty hours of supply a day and most urban households see about four.¹⁰ So the electron a driver can actually reach is the one he makes himself. A petrol generator burns half a litre to seven tenths of a litre for every kilowatt-hour, which puts his electricity somewhere between 600 and 840 naira a unit against the 209 on the tariff sheet.¹¹ The same 100 kilometres now costs him 7,200 to 10,080 naira, and the fourfold advantage is gone. He did not buy a cheaper car. He bought a car whose fuel he has to manufacture at three times the retail price. Public fast charging in Lagos at about 35 US cents a unit only leaves him a fifth better off, which is not a reason to change anything.¹²

Ethiopia is the counter-case and it says something about scale. More than 140,000 electric vehicles have been imported, while the utility's first phase of public charging is 40 stations, 32 of them in Addis, against a national target of 2,230. The three that opened most recently are at Kotebe, Yelebu and Piassa, and electricity still reaches only about half the population.¹³ The state also wrote a separate charging tariff of 14 to 18 birr a unit against a residential rate starting under one birr, which is a government deciding on purpose what the transition should cost.¹⁴

What that means for one person is easier to see with bikes. Ampersand runs battery swapping for motorcycle taxis out of a depot on KK 6 Av in Kigali, and it says a petrol rider spends more than $11 a day on fuel and vehicle leasing while taking home as little as $1.60.¹⁵ Those are the company's own figures and should be read as such, but that is the shape of it. The electron takes most of what a man earns, and no tariff sheet will show you that.

Nobody to fix it

Fewer moving parts means fewer mechanical problems, which in theory is great. The difficulty is what happens when you do have a problem, because your entire labour force has been built around combustion vehicles and nobody quite knows what to do. A man who can rebuild a carburettor in an afternoon cannot open a battery pack or read a controller fault. So if you do not have a proper dealership with maintenance behind it, or a flagship shop, you are beating around the bush trying to work out what has happened.

The repair bill is the smaller half of what that costs. The bigger half is the revenue the driver never earns while the car sits, and in most rent-to-own structures the repair lands on him as well while the payment schedule carries on regardless.¹⁶ In Kenya riders lose up to 500 shillings on a single missed swap, which is about four dollars, so multiply that by the days a car waits for a part.¹⁷

There is another version of this that almost nobody puts a number on. These markets run on used vehicles, and an electric car whose battery is a third of its value has no resale market yet, partly because nobody local can certify how far that battery has degraded. No residual value means no financing structure, so the maintenance gap does not only idle cars. It caps what the whole asset class can be lent against.

Only the asset

Equity is not built in a vacuum, and in most of these markets there is no real equity, because the markets are informal and the bulk of the legibility work has never been done. I have written before about how a platform makes a transport company's vehicles visible enough for a bank to underwrite them, and the same logic runs down to the driver. What is real is the productive asset, and the asset stands in as a proxy for equity, hence the explosion of rent-to-own.

The capital has arrived on that basis. Moove raised $250 million in August 2026 at a $2.1 billion valuation led by Mubadala, and Watu posted net profit of 4.8 billion shillings for 2025 on revenue up 92.7 percent to 28.3 billion.¹⁸ Both are real businesses built on the same insight, that the asset is the only collateral there is. What those numbers do not tell you is what happened to the driver, because a financier's margin and a borrower's balance are separate questions and the second one is much harder to find published anywhere.

You can see how much of that legibility work is missing in an odd place. Try to establish what share of the cars in a city are working vehicles rather than private ones, and the registries cannot tell you. The Philippine split between private and for-hire is not cleanly published, and Colombia's public-service share covers cars while leaving out the motorcycles that are most of its fleet. Nigeria's own tables disagree with each other, and one of them counts a different population altogether.¹⁹ Three countries, three incompatible categories, three different years. You cannot compare them, and the fact that you cannot is the point, because this is what an illegible market looks like from the inside and it is why capital prices these assets the way it does.

Rent the battery

If you have to build the missing layer anyway, then build it as something you own, and build it through productive assets first. Batteries are the most expensive part of the car, so make them swappable and charge for that instead of selling the vehicle with the battery in it. Packs are down to about $84 a kilowatt-hour in China and roughly $105 as a global average. So on a car selling for nine to twelve thousand dollars a 40 kWh pack is somewhere between 28 and 37 percent of the whole thing.²⁰ Third-party estimates put the saving at $500 to $1,000 off the purchase.²¹

You can knock 30 to 40 percent off the down payment that way, which gives access to far more people much quicker. It also moves the most expensive failure in the car off the driver's balance sheet and onto the network's, which matters more than the discount does.

There is an objection to all of this and it is Acemoglu's. Technology does not distribute its own gains, they go to whoever holds power over its direction, and historically they got shared only when people organised and forced the sharing.²² What I am describing has an operator owning the swap network, setting the specification and taking a margin on every swap. That is what it is, and I am not going to dress it up.

What makes it defensible is the split. The driver owns the car and rents the battery, and that is honest, and the trade is fair because it is productivity acceleration, money cycling faster into people's pockets, which ends up pushing consumption. It stops being fair the moment the contract hands the risk back to him, which is exactly what the repair clause does today.

The mechanics are further along than most people think. CATL passed its two thousandth standardised swap station at the end of June 2026 and is adding more than 200 a month, on packs built to work across brands and vehicle classes. The first production car on that standard went to a taxi fleet, a thousand Changan Oshan 520s.²³ Spiro runs 2,500 swap stations across six African countries.²⁴ It is easy for bikes and it can be done for cars, but read Spiro carefully. A motorcycle pack is a few kilowatt-hours and a man carries it in one hand, while a 40 kWh car pack is a different physical and capital problem. Sliding between the two is how people talk themselves into projects that do not work, and Better Place raised around $850 million on swapping and died of that exact gap.

Two things decide whether it happens, and the first is permission. In Nigeria, generating under a megawatt for your own use is a registration, but selling a kilowatt-hour to anyone else needs a licence whatever the size.²⁵ In Côte d'Ivoire it sits inside a concession granted in 1990 and extended in 2020.²⁶ A swap network is a company selling electricity to third parties, so getting permission to do it is harder than sourcing the batteries.

The second is density. You need fleet density first and interoperability at some point after it, and whoever gets to market density first is who leads the interoperability charge and moves a city in that direction. Interoperability is what creates the scale. Think about the app store, where any developer can publish but Apple still keeps part of the revenue and the right to decide who gets in.

The art and the science

States need to iterate fast on what works. Ethiopia works, Tajikistan works, and Tajikistan's five-year programme funds maintenance capacity and battery recycling rather than only the tariff, which is the whole ecosystem written into policy by the one government that got it right.⁸ The economics of EVs work, but the disruption needs the state either to step out of the way or to remove the structural barriers to building that future.

The rest of it is generation, because electrons do not generate themselves and charging infrastructure will not build itself either. Couple the transition to renewable energy growth, give people the chance to produce and sell electricity at scale, and simplify the process of bringing the surplus to the grid. South Africa stopped raising its generation licensing threshold and removed it altogether in January 2023.²⁷ Kenya gazetted net metering in 2024 and then found the utility could not credit exported power, so it went into a pilot.²⁸ One of those is a state getting out of the way, and the other wrote the rule and never built the system that would pay people for what they export. It is not one answer, it is several, and the puzzle only fits as a whole.

Six years after T'Leva ordered its first Chinese cars, the money in Luanda has gone somewhere else. Anda raised $3.4 million in November 2025 to put motorcycle taxis on a drive-to-own footing, and it does not own a fleet, it finances riders.²⁹ The productive asset sits with the person using it.

Building in this environment is as much an art as a science. The art is understanding the policy winds and adapting very quickly to them, and the science is getting rigorous about what makes the math work in each market and then pursuing that aggressively. I am bullish on how Addis is doing this, and I would like to see it across twenty more cities.