What happened to the all-EV future?
Battery costs fell. Electric-vehicle sales grew. The all-EV future did not arrive.
In 2017 the loud forecast was simple: batteries keep collapsing, EVs take the fleet, oil demand falls. We thought the cost curve would slow as it hit materials, that cheap cars would electrify last, and that the oil market was much larger than the gasoline barrel.
EV adoption was not the investment conclusion.
Materials, cheap-car economics, fleet uptime, and the rest of the barrel.
The market was debating how fast EVs would grow — and what that meant for oil.
The loud 2017 claim was speed.
Musk forecast that more than half of new U.S. vehicle production would be electric inside a decade, with almost all of it autonomous in the same window.
The cost curve was the foundation.
BNEF expected battery costs to fall from about $273/kWh in 2016 to roughly $73 by 2030. More aggressive forecasts went further: Tony Seba extrapolated continued exponential declines to roughly $50/kWh by 2027 and about $30–35/kWh by 2030. The argument was not merely that batteries would improve; it was that extraordinary cost declines would continue doing much of the economic work.
That became an oil-demand forecast.
BNEF’s 2017 outlook did not stop at market share. It put a barrel figure on the page: EVs taking 54% of new sales and knocking out about 8 million barrels a day of transport fuel by 2040.
IEA’s Sustainable Development Scenario was peak-and-decline.
The goals-based case put oil demand at about 73 mb/d by 2040, with an EV stock approaching 875 million.
This was becoming an investment decision, not just a forecast.
By 2017, fossil-fuel divestment was spreading through endowments and institutional portfolios, and the movement accelerated through 2020. A decade of weak energy returns mattered enormously. Peak-demand forecasts added a second argument: long-lived oil and gas assets might become stranded before they paid out.
Our 2017 report took the other side.
The 2017 report expected EVs to grow and batteries to get cheaper. It did not expect an all-EV sales mix, and it did not expect the oil market to shrink.
Battery manufacturing could keep improving. Raw materials would not. The battery would stay expensive in a cheap car. Charging time would punish any vehicle that has to earn its keep. Most of the oil barrel does not live in a passenger tank.
The cost curve slowed as the battery became a raw-material problem.
Battery manufacturing could keep getting cheaper. Raw-material costs could not follow the same curve.
Our 2017 forecast had a different shape. Battery costs would keep falling, but the rate of decline would slow as raw materials became a larger share of the battery. We modeled a gradual glide toward roughly $90/kWh by 2030.
We forecast a gradual decline in battery costs, not an exponential collapse.
Our 2017 forecast was not a 2025 price target. We modeled battery costs declining gradually toward roughly $90/kWh by 2030 as manufacturing gains slowed and raw materials became a larger share of the cost.
BloombergNEF’s 2025 survey put BEV batteries at $99/kWh. That is still close to the gradual path we forecast. There was no exponential break toward $30 — even after lithium prices collapsed.
Batteries stayed expensive. As we forecasted in 2017, that cost was easiest to hide inside a higher-priced vehicle.
The battery was easiest to hide in an expensive car.
Our 2017 model split the market into economy cars below $30,000, mid-priced cars from $30,000–$60,000, and luxury cars above $60,000. The same battery produced very different economics across those price points.
We modeled three very different EV markets.
Economy cars below $30,000. Mid-priced cars from $30,000–$60,000. Luxury cars above $60,000 — all based on 2017 MSRP. The same battery that overwhelmed the economics of an economy car could disappear into the cost structure of a luxury car.
Economy cars — below $30,000 in 2017 — were nowhere close to parity.
In our model, an unsubsidized economy EV ran about 52 percent higher cost per mile than a comparable combustion car. A five-figure battery was simply too large a cost to hide in a sub-$30,000 vehicle.
Mid-priced cars — $30,000–$60,000 — were better, but still not at parity.
The battery mattered less as the vehicle price rose, but our 2017 model still put an unsubsidized mid-priced EV about 23 percent above the cost per mile of a comparable combustion car.
Luxury cars — above $60,000 — were already roughly cost-competitive.
At luxury price points, the battery became a manageable share of the vehicle cost, while expensive combustion cars offered unusually large fuel savings for an EV to capture. In our 2017 model, luxury EVs were already roughly cost-competitive.
Tesla started at the top of the market for a reason: luxury was not simply branding. It was where the battery was easiest to hide.
Roughly 80% of a car’s cost had nothing to do with fuel.
Cheaper fill-ups were being asked to do an enormous amount of work.
BNEF’s 2017 outlook put unsubsidized sticker parity in the late 2020s. But fuel was only one part of the economics. Roughly 80% of the cost of owning a car had nothing to do with fuel.
In addition to forecasting a slower decline in battery costs, we thought it was a mistake to assume cheaper fueling alone could transform what people bought and how they used cars in under a decade.
We modeled the costs that did not disappear when gasoline did.
Depreciation, repairs, resale value, battery replacement and downtime all remained part of the ownership calculation.
A long-lived electric drivetrain did not guarantee strong resale value.
EVs now sit alongside luxury combustion cars among the fastest-depreciating vehicles in the market. iSeeCars estimates the average EV loses 57.2% of its value over five years.
Tesla's repeated new-car price cuts made the mechanism especially visible: the battery did not suddenly deteriorate. The market price of the used car changed. That was our 2017 point — depreciation is a supply-and-demand outcome, not an engineering-life estimate.
For a revenue-producing vehicle, downtime is a real cost.
Forecasters saw fleet operators as natural early adopters of EVs.
High mileage meant more fuel savings, lower maintenance costs looked attractive, and autonomy was expected to arrive quickly enough to lift utilization further.
We thought charging time and resale value were much bigger obstacles.
In 2017, charging a 150–200 mile battery to 80% could take roughly 40 minutes. We estimated a heavily utilized fleet EV could lose 80–100 minutes of productive time each day to charging.
For most fleet operators, lower uptime and uncertain residual values cut directly against the reason the vehicle existed: to stay on the road and earn money.
A van that sleeps at the depot is not a rental car on Saturday.
Hertz cut 30,000 EVs after customer demand, damage costs and poor residual values hurt the economics.
Depot-charged delivery vans are different. They return to the same place every night and can charge during planned downtime.
We did not forecast that EVs would fail. We forecast that oil demand would keep growing anyway.
That was a consequential call in 2017. Peak-oil-demand forecasts were moving rapidly into the investment mainstream, and fossil-fuel divestment was accelerating. The IEA Sustainable Development Scenario put 2040 oil demand near 73 million barrels a day.
Our argument was not that electrification would be irrelevant. It was that EVs would slow gasoline-demand growth without shrinking the much larger oil market. Nine years later, we can test that forecast against what actually happened.
In 2017, the contrarian forecast was not about EV market share. It was that oil demand would keep growing.
EV share was the input. The size of the future oil market was the investment conclusion.
We did not publish a 36-page EV report because we wanted to win a debate about cars. In August 2017, with peak-demand forecasts gaining credibility and divestment accelerating, we forecast that global oil demand would still grow to roughly 106 million barrels a day by 2040.
That was the bet: EVs could become important without making the oil market smaller.
Some markets actually delivered the EV future.
China — along with Norway and several European countries — delivered on some, if not most, of the aggressive EV projections we analyzed in 2017. Gasoline demand collapsed. Overall oil demand did not. Depending on the market, it remained roughly flat or continued to grow.
Part of the reason is the counterfactual. EVs did not replace a frozen fleet of 2017 gasoline cars. They displaced cars that were themselves becoming more efficient, including hybrids. If the next buyer would otherwise have chosen a 50-mpg small car or hybrid, electrification removes much less gasoline than a comparison with the old fleet suggests.
Gasoline is not the barrel.
Aviation, petrochemicals, LPG, diesel. Passenger gasoline can fall hard without taking the whole barrel with it. China became the clearest large-scale example: transport electrified far faster than we forecast, while petrochemical and other oil demand kept the total market growing.
Growth declined… Oil demand did not.
Our 2017 forecast had oil-demand growth slowing from roughly 1.8% to roughly 0.8%. We did not think that was the same thing as demand destruction. The 2017 forecast still had the global oil market around 106 million barrels a day by 2040.
The call was not that EVs would fail. It was that oil demand would keep growing anyway.
Our August 2017 forecast put global oil demand near 106 million barrels a day in 2040. The IEA Sustainable Development Scenario put 2040 demand near 73 million barrels a day.
By late 2025, observed liquids demand was already about 104.6 million barrels a day. That is a checkpoint, not the end of a 2040 forecast. But it makes the scale of the original disagreement hard to miss.
The 2017 narrative changed capital allocation. Nine years later, the investment outcome looked very different.
Poor energy returns mattered too.
Climate policy supplied one argument. Performance supplied another. For every year from 2011 through 2022, rolling three-year energy returns lagged the S&P 500.
The deepest three-year underperformance overlapped with the heaviest years of divestment announcements from 2014 through 2021.
The cost of divestment became visible in 2022.
In calendar 2022 the S&P Energy sector rose 65.7 percent. The S&P 500 fell 18.2 percent. Institutions that had removed energy from their portfolios missed the strongest sector in the index.
Trailing three-year energy returns still led the S&P through year-end 2025, and energy was another 15 percent ahead year-to-date through May 2026. On June 1, Princeton’s endowment walked back the exit from publicly traded oil and gas.
Nine years later
Electric vehicles are now a meaningful automotive technology. That is not the same thing as the all-EV future investors were being asked to underwrite in 2017.
We missed pieces of the adoption path. We did not miss the size of the energy market.
The investment mistake was treating the passenger car as a proxy for the oil market.
As Recurrent wrote in 2017, EVs’ impact on oil demand was dramatically less than feared. Demand-destruction fears led to lower energy capex and refinery shutdowns, and the effects of those investment decisions continue to impact energy markets today.
2017 market-narrative examples: Elon Musk, National Governors Association, 15 July 2017; BloombergNEF 2017 battery-cost forecast (about $273/kWh in 2016 to $73/kWh by 2030) and Electric Vehicle Outlook 2017 (Colin McKerracher); Tony Seba / RethinkX transportation framework on exponential lithium-ion battery-cost declines. IEA Sustainable Development Scenario: about 73 mb/d of oil demand by 2040. 2025 battery costs: BloombergNEF survey. China fuel-demand context: IEA, March 2025. Divestment: divestmentdatabase.org and Recurrent Research, updated May 2026.