EV batteries keep 81.6% of capacity after eight years, yet 330,000 US lease returns face a market that cannot verify battery health. The gap is the story.
The Reckoning Is Not About Demand
Twenty million cars, and three markets pretending to be one
Global electric car sales passed 20 million in 2025 and the International Energy Agency expects roughly 23 million in 2026, about 29% of the world market ✓ Established Fact [1][2]. The open question is no longer whether people will buy them. It is what happens to those cars in year four, on a residential feeder at seven in the evening, and in a used listing that cannot say what the battery is worth.
Start with the aggregate, because the aggregate is the part almost everyone gets right. Electric car sales grew 20% in 2025 to exceed 20 million units, and the International Energy Agency projects around 23 million in 2026 at roughly 28% of total car sales [1]. After a stronger than expected second quarter the agency raised that share forecast to 29% [2]. Those are the headline numbers, and they describe a market still expanding faster than almost any other durable-goods category in the world economy.
The aggregate also conceals a violent internal divergence. Worldwide electric car sales fell about 8% year on year between January and March 2026, to 3.9 million units, on weaker demand in China and the United States [2]. The second quarter then rose 35% on the first, with record quarterly sales in 50 countries [2]. A market that swings from minus 8 to plus 35 in six months is not a smooth adoption curve. It is a policy-timed market, and the timing is set by tax codes rather than by consumers.
Geography does the rest of the concealing. In China, electric cars accounted for nearly 55% of all car sales in 2025, and Chinese manufacturers supplied around 60% of global electric car sales [1]. European sales rose more than 30% to over 4 million vehicles, a share of roughly 28% [1]. The United States stayed below 10% [1]. Three markets, one aggregate, and almost nothing in common between them in price structure, charging density or policy direction.
The fastest growth is now outside all three. Electric car sales in emerging and developing economies outside China grew around 80% in a year [3]. Thailand rose 70% in 2025 to roughly 140,000 units, close to a quarter of new car sales, with Chinese brands supplying 88% of them [39]. Latin American sales grew 75%, and in Brazil more than 85% of new electric cars were imported from China [39]. India reached a record 2.3 million electric vehicles of all types, and India, Mexico and Brazil now each sit above Japan on electric share [39].
The category itself is also splitting. Plug-in hybrids built in China rose from 37% to 60% of total plug-in hybrid imports into the European Union between 2024 and 2026, because the tariffs applied to battery-electric imports do not reach them [29]. The December 2025 European decision then explicitly preserved plug-in hybrids, range extenders and mild hybrids beyond 2035 [30]. What a national statistic counts as electric is therefore drifting, and drifting toward vehicles carrying a smaller battery, a shorter electric range and an entirely different second-hand profile.
This matters for everything below, because adoption statistics measure only the first transaction. Every electric car sold in 2021 is now a four-year-old used car. Every car sold in 2018 leaves its eight-year warranty window within two years. And the systems that determine what those cars are worth — battery-health disclosure, public charging density, housing tenure, distribution-grid capacity — were not built alongside the sales. They are being built afterwards, unevenly, and in several jurisdictions they are being dismantled [31][33].
Every mass-market technology has a second decade in which the resale market, the repair market and the insurance market decide what the first decade was actually worth. Electric vehicles entered theirs in 2025. The evidence below shows the physical asset performing better than almost anyone forecast, and the informational and infrastructural systems around it performing considerably worse.
The reckoning in the title is therefore a second-order one. The technology works. The manufacturing works, at more than 20 million units a year [1]. What remains unresolved is everything downstream of the showroom: whether a second owner can price the asset, whether a second owner can charge it, and whether the local grid notices when several of them plug in on the same street at the same hour.
The Batteries Are Not the Problem
What 22,700 vehicles say about degradation
Telematics from more than 22,700 electric vehicles across 21 makes and models put average battery degradation at 2.3% a year, with the average pack retaining 81.6% of original capacity after eight years ✓ Established Fact [4][5]. The durability question that dominated a decade of coverage has largely been answered, and the answer is reassuring.
Geotab published the study in January 2026, drawing on aggregated telematics from commercial and mixed-use fleets rather than on owner-reported range [4]. State of health — remaining capacity measured against original rated capacity — degraded at an average of 2.3% a year across the sample [4]. Applied across an ordinary ownership period, that rate leaves a pack comfortably inside the range needed for daily commuting well past the point at which the rest of the vehicle begins to wear out.
The eight-year figure is the one that matters commercially, because eight years is the minimum federal warranty term in the United States [7]. At that mark the average vehicle in the sample retained 81.6% of original capacity [5]. Catastrophic failure is rarer still: fewer than 2% of electric vehicles built since 2016 have required a full pack replacement [7]. The failure mode that dominated early scepticism — a cliff-edge battery death somewhere past 100,000 miles — did not materialise at fleet scale.
Geotab measured state of health across more than 22,700 vehicles spanning 21 makes and models, finding average degradation of 2.3% a year [4][5]. Replacement data from Recurrent puts full-pack replacement below 2% of the post-2016 fleet [7]. Together these establish that battery longevity is no longer the binding constraint on electric vehicle ownership economics, which shifts the analytical burden onto the markets that price these vehicles rather than onto the chemistry inside them.
Charging behaviour, not age, now explains most of the variance. Vehicles relying heavily on direct-current fast charging above 100 kW degraded at up to 3.0% a year, against roughly 1.5% for vehicles charged mainly on alternating current or at lower power [4]. Vehicles using high-power chargers for more than 12% of sessions ran at 2.5% a year against 1.5% for those below that threshold [5]. Climate mattered less than expected: hot regions added about 0.4% a year, and the heaviest-use vehicles about 0.8% [4].
That finding is contested ⚖ Contested. A separate analysis of more than 160,000 data points found no statistically significant difference in range degradation between vehicles fast-charging more than 70% of the time and those fast-charging less than 30% of the time, including observations of 2012 through 2023 Tesla vehicles [6]. The disagreement is methodological rather than factual. Fleet telematics measure state of health directly and skew toward high-utilisation duty cycles; consumer range datasets measure usable range under varying ambient conditions. Both can be correct about their own samples.
Using the lowest charging power that still meets operational needs can make a measurable difference to long-term battery health.
— Charlotte Argue, Senior Manager for Sustainable Mobility, Geotab, January 2026Regulation has begun to codify the durability the data already shows. Federal rules oblige manufacturers to warrant traction batteries for eight years or 100,000 miles, whichever comes first [7]. California goes further: from the 2026 model year, batteries must retain at least 70% of range for ten years or 150,000 miles [7]. A warranty floor of that kind converts a technical property into a contractual one, which is precisely what a secondary market needs in order to price risk at all.
The controllable variable is state of charge rather than age. Geotab found degradation accelerating principally where vehicles spent more than 80% of their time at very high or very low states of charge [5]. That is a behavioural finding with a commercial implication: two identical cars of identical age can carry materially different packs depending on how their previous owners charged them. Charging history is, in principle, recordable and disclosable. It is not currently recorded or disclosed in any consumer market at the point of resale [37].
And yet the pack remains cheap at the factory and expensive at the kerb. BloombergNEF put average lithium-ion pack prices at 108 dollars per kilowatt-hour in 2025, down 8% on the year, with battery-electric vehicle packs averaging 99 dollars and cells 74 dollars [28]. Out of warranty, a replacement pack runs roughly 6,000 to 30,000 dollars installed, with refurbished packs and single-module repairs between 2,000 and 8,000 dollars on some models [7]. The distance between the wholesale cost of cells and the retail cost of a replacement is the hinge on which the entire used-vehicle question turns.
The Secondhand Cliff
Where the transition actually breaks
Three-year depreciation on United States electric vehicles ran at 38% to 42% of original list price in the first half of 2026, improved from earlier peaks of 49% to 55% but still faster than comparable petrol cars ◈ Strong Evidence [8]. Into that market, at least 243,000 leases expire this year, and possibly 330,000 [10].
The headline improvement is real. Three-year depreciation across the United States used market sat at roughly 38% to 42% of original list price in the first half of 2026, down from peaks of 49% to 55% recorded earlier [8]. Measured against the roughly 40% that an average new petrol car loses over three years, the gap has narrowed substantially [9]. Across five years, however, electric vehicles still lose about 10% to 15% more value than comparable combustion models [8].
The average is the least interesting number in the dataset. The dispersion is enormous: the strongest performers hold value at under 30% loss in their early years, while the weakest shed more than 60%, and some luxury models lose over 70% of original list price within five years [9][8]. A category in which two vehicles of similar age, similar mileage and similar chemistry can differ by forty points of residual value does not have a pricing problem at the margin. It has a market that cannot identify what it is buying.
Supply is about to test that pricing machinery. Electric vehicle lease returns are forecast to jump more than 200% in 2026 as 2023 originations come due, with at least 243,000 leases expiring and estimates running as high as 330,000 vehicles, more than three times the 2025 volume [10]. Leasing was the channel through which a great many 2023 buyers accessed incentives, which means the incentive design of three years ago has become the supply schedule of this year.
The price response has so far confounded the forecast. United States used electric vehicle prices rose 5.1% in the first half of 2026 even as roughly 300,000 cars came off lease [11]. Europe went the other way: used electric prices in several European Union markets fell 20% to 35% from their 2023 highs [13]. J.D. Power attributes the European pressure to powertrain divergence, with high-penetration markets such as the Nordics and the Netherlands sitting alongside leasing-driven fleets in Germany, France and Belgium, creating oversupply risk in exactly the markets that electrified first [12].
Battery state of health is the single largest determinant of a used electric vehicle's value, yet no standard disclosure exists at the point of sale in most jurisdictions [37]. The European Union battery passport, which will carry state-of-health and cycle-count data, takes effect in 2027, and even then per-unit performance data is restricted under Annex XIII to owners, authorities and parties with a legitimate interest such as repairers and recyclers [37]. Until that information reaches buyers, residual values must price an unobservable risk, and unobservable risks are priced conservatively.
Insurance amplifies the same opacity. Electric vehicles average 3,159 dollars a year to insure in the United States against 2,218 dollars for petrol cars, a premium of about 42% [14]. Collision repair runs roughly 25% to 30% more per claim, and the battery alone can represent 30% to 40% of total vehicle value [15]. The Highway Loss Data Institute found that electric vehicle collision claims were written off as total losses at a rate about 50% higher than combustion vehicles of similar value [15].
Residual forecasts are set years before the cars arrive, which compounds the problem. A lease written in 2023 embedded an assumption about 2026 second-hand values made under a different incentive regime, a different tariff schedule and a different set of emissions targets [31][30]. Where that assumption proved generous, the lessor absorbs the difference and tightens terms on the next cohort; where it proved conservative, the buyer captures a windfall [12]. Neither outcome improves the information available to the person standing on the forecourt, and both widen the spread that dealers quote.
That mechanism deserves stating precisely, because it is frequently misreported as a safety finding. It is not. Vehicles are written off because the estimated cost of verifying and repairing a pack after an impact can exceed the insured value even when the car appears structurally intact [15]. The write-off is a decision made under uncertainty about the battery, not a judgement about the crash. The same missing information that depresses residual values also destroys otherwise repairable cars.
The write-off arithmetic follows directly from the pack price. A replacement pack at 6,000 to 30,000 dollars installed sits against a three-year-old vehicle that has already shed 38% to 42% of its list price [7][8]. On a mid-market car the pack alone can approach the residual value of the whole vehicle, which is why the battery represents 30% to 40% of insured value and why uncertainty about its condition after an impact so readily tips a repair estimate past the total-loss threshold [15]. Cells at 74 dollars per kilowatt-hour never reach the owner at anything close to that price [28].
A used combustion car can be assessed in an afternoon with a compression test and a service history. A used electric car cannot: its most valuable component reports its condition only to software the buyer does not control. Every actor in the chain — the second owner, the dealer, the insurer, the lender setting a residual — prices the same blind spot, and each adds a margin for it. The cumulative discount is larger than the underlying engineering risk.
This is the classic market for lemons, and it has the classic remedy. Sellers who know their pack is healthy cannot prove it, so they withdraw or accept the pooled price; buyers who cannot verify condition assume the worst and bid accordingly. Mandatory, transferable state-of-health certification at the point of sale would collapse most of the discount without a single engineering change. It is the cheapest available intervention in the entire transition, and no major market has yet implemented it for consumers [37].
Charging Is a Geography Problem
Not a plug problem, an estate problem
China runs about 10 electric vehicles per public charging point, Europe 13, and the United States nearly 24 ✓ Established Fact [16]. That ratio is not a measure of convenience. It is a measure of who can plausibly own the cars now coming off lease.
The three-way ratio is the cleanest single indicator of infrastructure adequacy available, and it separates the markets more sharply than adoption share does. China sits at roughly 10 electric vehicles per public charger, Europe at 13, and the United States at nearly 24 [16]. The American figure is not a transitional artefact of fast growth: American electric share remains below 10%, so the ratio reflects a charging estate that has fallen behind a comparatively small fleet [1][16].
Absolute stock tells the same story. China accounts for roughly 65% of the world's public charging stock and around 60% of the global electric light-duty fleet [18]. Europe is projected to add more than 219,000 public charging points during 2026, taking the regional total to 1.47 million [17]. The United States had 249,997 public charging ports as of June 2026 [21]. Analysts expect the divergence to widen, with China and Europe maintaining steady expansion while American momentum softens after the phase-out of purchase incentives [17].
The federal programme meant to close the American gap has moved slowly. The National Electric Vehicle Infrastructure formula programme allocated 5 billion dollars to states between 2022 and 2026, and by late 2025 at least 384 charging ports had been built through it [20]. Some 885 million dollars was apportioned for the 2026 fiscal year, and the final 2026 budget trimmed the programme by 503.8 million dollars [20]. Revised guidance in August 2025 relaxed the 50-mile corridor spacing rule and removed several non-statutory siting and labour requirements, while retaining a 97% average annual uptime standard per port [20].
Research from the Center for Sustainable Energy and EV Connect puts multifamily charging availability at roughly 5% of rental properties in 2026 [19]. Renters dependent on public direct-current fast charging pay 0.35 to 0.50 dollars per kilowatt-hour against a national home-charging average near 0.18 dollars, a gap worth 600 to 900 dollars a year on a mid-size vehicle driven 12,000 miles [19]. Charging access is therefore distributed by housing tenure rather than by income alone.
Density is only half of the measure; availability is the other half. The American federal programme requires each funded port to maintain at least 97% average annual uptime, a standard introduced precisely because reliability rather than count had become the binding complaint [20]. A charging estate of 249,997 ports with intermittent availability serves fewer drivers than a smaller estate that works, and no published national figure reconciles installed ports with functioning ones [21].
The behavioural consequence is measurable. In a 2026 multifamily resident survey, 62% said they planned to own or drive an electric vehicle within five years if charging were available at their building, against only 25% where it was not [19]. That is a thirty-seven point swing produced entirely by a socket. No purchase incentive in any market has ever moved stated intent that far.
China built its position the other way round, with public network expansion tracking fleet growth rather than trailing it. The country holds roughly 65% of world public charging stock against about 60% of the global electric light-duty fleet, a ratio close to parity and the reason its vehicle-per-charger figure sits near 10 [18][16]. Europe is converging on the same balance through sheer volume, adding more than 219,000 points during 2026 alone to reach 1.47 million [17].
Connect this to the previous section and the two problems become one. The vehicles coming off lease in 2026 are mostly three-year-old compact cars and crossovers priced, after depreciation, squarely in the range of second-hand buyers, who are disproportionately renters [10][8]. A used electric vehicle is affordable to precisely the population least likely to have anywhere to plug it in. That mismatch, not battery chemistry and not sticker price, is the structural ceiling on used electric vehicle demand.
Home charging at roughly 0.18 dollars per kilowatt-hour is the single largest running-cost advantage electric vehicles hold over petrol. Renters, excluded from it, pay three to four times as much for the same electricity and capture almost none of that advantage [19]. The result is ownership economics that are genuinely excellent for homeowners and merely adequate for everyone else — and it is everyone else who buys used cars.
Utility programmes are beginning to respond, with multifamily budgets expanding as the federal charger tax credit lapsed [19]. But the unit of analysis has shifted. The relevant question is no longer how many chargers a country has built; it is what proportion of its housing stock can charge overnight. On that measure the United States trails Europe and China by a margin no highway corridor programme can close.
The Grid Will Cope. The Feeder Might Not.
Why the load question is local rather than national
United States electricity demand is entering its strongest four-year growth since 2000, and the Energy Information Administration attributes up to 80% of growth to 2050 to electric vehicles and data centres combined ✓ Established Fact [22]. The national total is manageable. The street-level coincidence is the part that costs money.
Take the aggregate first. Global electricity consumption passes 29,000 terawatt-hours in 2026 [22]. Electric vehicle adoption is expected to add more than 100 terawatt-hours to United States demand through 2030 [22]. Measured against a national system, that increment is real but unremarkable: smaller than the data-centre build-out with which it shares the headline, and spread across a decade of planning cycles.
The difficulty is not energy. It is coincidence. Uncontrolled charging overlaps with the existing evening peak, concentrating new load into the hours when distribution assets are already most stressed [23]. A commercial building drawing 250 kilowatts that adds twenty chargers can double or triple its total load, forcing substation and feeder reinforcement alongside transformer replacement [24]. The cost of electrification lands on the last hundred metres of wire, not on generation.
California has quantified it. Research on the state's distribution system found that 67% of feeders will require capacity upgrades by 2045, amounting to 25 gigawatts of reinforcement at a cost between 6 and 20 billion dollars [24]. That is the bill for continuing to charge the way people charge now, and the range is wide because the outcome depends almost entirely on when vehicles draw power rather than on how much they draw.
Which is why the mitigation is unusually cheap. Managed charging programmes cut distribution upgrade costs by roughly 30%, and some utilities report that managed charging reduces electric vehicle peak demand by 50% to 70%, effectively doubling the number of vehicles an existing network can support [24][25]. The California Public Advocates Office calculated that shifting electric vehicle load away from peak hours could save between 5 and 18 billion dollars in distribution costs by 2040 [25].
| Risk | Severity | Assessment |
|---|---|---|
| Residual-value opacity | Battery state of health is unobservable at the point of sale until at least 2027, and the resulting discount applies to every used electric vehicle regardless of actual condition [37][8]. | |
| Midstream mineral concentration | China processes over 90% of the world's graphite and more than two-thirds of lithium and cobalt refining capacity, a chokepoint that mining diversification cannot address [27]. | |
| Charging access by housing tenure | Roughly 5% of United States rental properties offer charging, which caps used electric vehicle demand among the buyers most likely to want the cars [19]. | |
| Policy reversal | The purchase credit, the California waiver and the European 2035 target all changed within fifteen months, making residual forecasting a political rather than a technical exercise [31][33][30]. | |
| Distribution-feeder coincidence | Uncontrolled evening charging drives local reinforcement costs, but managed charging demonstrably removes most of the exposure at low cost [24][25]. |
The comparison with data centres is politically consequential rather than technically illuminating. Both loads appear in the same Energy Information Administration projection, and together they account for up to 80% of United States electricity growth to 2050 [22]. But a data centre arrives as a single interconnection request with a named counterparty and a negotiated tariff. Electric vehicles arrive as several hundred separate household decisions distributed across a residential feeder, with no counterparty to negotiate with and no single point at which the cost can be allocated [23].
Vehicle-to-grid work points the same way. A study of the German power system projected up to 6% lower distribution grid investment to 2040 under grid-friendly vehicle-to-grid operation, compared with uncontrolled and purely cost-optimised smart charging [26]. Work on a region of northern France found that unidirectional and bidirectional flexibility could cut 2040 peak loads by 6% and 9% respectively, saving around a quarter of annual grid reinforcement costs relative to a low-flexibility case [26].
That asymmetry explains why a cheap mitigation has been slow to arrive. The commercial case is concentrated: a building drawing 250 kilowatts that adds twenty chargers can double or triple its load and will negotiate the reinforcement directly [24]. The residential case is diffuse, and the reinforcement cost is socialised across a rate base that includes households with no vehicle at all [23]. Managed charging resolves the engineering question. It does not by itself resolve who pays for the wire.
The policy conclusion is narrow and unglamorous. Whether electric vehicles strain the grid is contested ⚖ Contested only because the answer depends on tariff design rather than on vehicle numbers. Time-of-use pricing, default managed-charging enrolment and interoperable telematics are the entire intervention. They cost less than the reinforcement they avoid, they require no new technology, and they sit within the authority of regulators who have already approved them elsewhere [25].
The Dependency Beneath the Car
Minerals, midstream, and the price of a pack
Battery pack prices fell 8% in 2025 to 108 dollars per kilowatt-hour, the lowest on record ✓ Established Fact [28]. The cost curve that made electric vehicles competitive runs through refining capacity that one country controls almost entirely [27].
The price data first. The 2025 BloombergNEF survey put average lithium-ion pack prices at 108 dollars per kilowatt-hour, an 8% fall on the year and the largest drop since 2017, with battery-electric vehicle packs averaging 99 dollars and cells 74 dollars [28]. Prices fell despite rising metal costs, driven by cell manufacturing overcapacity, intense competition and the continuing shift to lower-cost lithium iron phosphate chemistry [28]. A further 3% decline is forecast for 2026, to just under 105 dollars [28].
Now the structure underneath the price. China produced 79% of the world's natural graphite in 2024, some 1.27 million short tons, and processes over 90% of global graphite supply [27]. Chinese companies accounted for more than two-thirds of world cobalt and lithium processing capacity, and own roughly 80% of cobalt production in the Democratic Republic of the Congo, where more than half of global cobalt originates [27]. Mining is moderately distributed. Refining is not.
Trade flows confirm the concentration. China imported almost 12 million short tons of raw and processed battery minerals in 2023, some 44% of interregional trade, and exported nearly 11 million short tons of battery materials, packs and components, or 58% of interregional trade [27]. A single jurisdiction standing on both sides of a supply chain at those proportions is not a market participant. It is the clearing house of that market.
United States Energy Information Administration data establishes both figures, alongside Chinese production of 79% of natural graphite in 2024 [27]. The concentration sits in midstream refining rather than in mining, which is why diversifying extraction does not resolve the exposure: ore mined in Australia or the Congo still travels to Chinese refineries before it reaches a cell plant anywhere in the world [27].
That position has already been used as an instrument. China introduced export restrictions on graphite products tied to electrode manufacturing in 2023 [40]. In November 2025 it temporarily eased licensing requirements for graphite shipments to the United States, with the suspension valid through 27 November 2026 [40]. A dependency that can be suspended and restored on an announced calendar is a dependency with a term structure, and every cell plant outside China now plans against it.
The commercial consequence is visible in Europe. Chinese brands recorded 550,000 car sales in the European Union in 2025 [29]. Cars built in China accounted for 17% of European Union battery-electric sales in the first quarter of 2026, down from a peak of 22% in 2024, when tariffs of 17% for BYD and 35% for SAIC were introduced on top of the standard 10% duty [29]. Chinese battery-electric models nonetheless remain about 21% cheaper on average than European equivalents, and plug-in hybrids made in China, which the tariffs do not cover, rose from 37% to 60% of total plug-in hybrid imports between 2024 and 2026 [29].
The response from BYD demonstrates how quickly the dependency reshapes itself. The company overtook Tesla in European registrations during the first half of 2026, recording 174,144 against 170,351, achieved not by absorbing the 17% duty on its battery-electric imports but by pivoting toward hybrids the measure does not reach [42][29]. A tariff calibrated to one powertrain redirected production into another within two model cycles, leaving the underlying cell supply relationship untouched.
None of this alters the environmental arithmetic, which has become steadily more favourable. The 2025 update from the International Council on Clean Transportation puts battery-electric lifecycle emissions at 63 grams of carbon dioxide equivalent per kilometre on the projected 2025 to 2044 European grid, 73% below the 235 grams estimated for petrol vehicles, and 52 grams on renewable electricity [36]. Production emissions remain roughly 40% higher than for combustion vehicles, but are offset after about 17,000 kilometres of use, within the first year or two of ownership [36].
End-of-life capacity is arriving faster than the systems built to absorb it. More than 120 gigawatt-hours of electric vehicle batteries reach end of vehicle life worldwide in 2026, with over half of those packs retaining 70% to 80% usable capacity [38]. Second-life deployment is scaling from roughly 25 to 30 gigawatt-hours in 2025 toward 330 to 350 gigawatt-hours by 2030 [38]. The European Union has made recovery a legal obligation carrying explicit efficiency and material-recovery targets rather than a voluntary commitment [38].
The recycling economics are lopsided in a way that shapes those flows. Recovery rates reach 95% to 99% for nickel and cobalt and 85% to 95% for lithium, but lithium iron phosphate black mass contains neither nickel nor cobalt, which is why lithium iron phosphate packs are steered toward second-life deployment far longer than nickel-manganese-cobalt packs are [43]. The recycling market is projected at roughly 6.9 billion dollars in 2026, rising to 37.5 billion by 2035, with Redwood Materials alone targeting 100 gigawatt-hours of cathode material output, enough for about a million vehicles a year [44]. Recovery capacity is being built for a chemistry the cost curve is displacing [28].
The Policy Reversal
Four governments changed the rules in fifteen months
Between June 2025 and January 2026 the United States repealed its purchase credit and its state-level mandate authority, the European Union softened its 2035 target, and China halved its purchase tax ✓ Established Fact [31][33][30][32]. No residual-value model survives that intact.
The American reversal came in two parts. The One Big Beautiful Bill Act, signed on 4 July 2025, terminated the Section 30D clean vehicle credit of up to 7,500 dollars for vehicles acquired after 30 September 2025, roughly seven years ahead of its scheduled expiry in 2032 [31]. A narrow binding-contract exception preserved eligibility for buyers who had contracted and made a qualifying payment before the deadline [31]. Electric vehicles reached a record 12% of new light-duty sales in September 2025 as buyers moved ahead of the cut-off [41].
The second part removed the states' authority rather than the buyers' subsidy. Congressional Review Act resolutions overturning the Environmental Protection Agency waivers underpinning Advanced Clean Cars II passed the House on 1 May and the Senate on 22 May 2025, and were signed on 12 June [33]. The effect is not a pause but a prohibition: California and the eleven states that had adopted the rule cannot enforce it [33]. The Government Accountability Office had advised in March that the waivers were not rules subject to the Act, an objection that did not change the outcome [33].
Europe moved in December. The European Commission announced on 16 December 2025 that manufacturers will face a 90% carbon dioxide reduction requirement from 2035 rather than the 100% previously legislated, with the remaining 10% compensated through European low-carbon steel or sustainable fuels [30]. Plug-in hybrids, range extenders, mild hybrids and combustion vehicles may therefore continue to be produced beyond 2035 [30]. The proposal passes to Parliament and Council, with negotiations beginning under the Cypriot presidency in January 2026 [30].
This is a clear signal that other technologies than battery electric vehicles can be put on the market after 2035.
— Apostolos Tzitzikostas, European Commissioner for Transport, December 2025China adjusted in the opposite register, tapering rather than reversing. From 1 January 2026 the new energy vehicle purchase tax moved from full exemption to a 50% reduction, with the per-vehicle cap falling from 30,000 yuan to 15,000 yuan, and the arrangement extends through 2027 [32]. Dealerships reported order volumes up nearly 60% on typical monthly levels in the weeks before the change [32]. A scheduled, pre-announced, partial withdrawal produces a demand pull-forward; an abrupt repeal produces a cliff.
The European national picture shows how far outcomes diverge under broadly similar targets. Norway reached 97.6% battery-electric share of new cars in the first half of 2026, with a record 98.6% in April [34]. The zero-emission vehicle mandate in the United Kingdom rose to 33% for 2026, but actual share reached 26.2% in April and 23.1% year to date [34]. Two European markets, one mandate design each, and a gap of more than seventy points in delivery.
The case that the transition is stalling
Global electric car sales fell about 8% year on year in the first quarter of 2026, to 3.9 million units, on weaker Chinese and American demand [2].
At least 18 manufacturers are cancelling, delaying or scaling back electric plans in the United States, against roughly 65 billion dollars of cumulative writedowns by early 2026 [35].
The federal purchase credit ended on 30 September 2025 and the California waiver was revoked on 12 June 2025, ending mandate authority across twelve states [31][33].
The 2035 requirement fell from a 100% reduction to 90%, explicitly preserving plug-in hybrids and combustion vehicles beyond that date [30].
The case that it is compounding
Sales exceeded 20 million in 2025 and are forecast near 23 million in 2026, about 29% of the world market [1][2].
Pack prices fell 8% to 108 dollars per kilowatt-hour in 2025 and are forecast just under 105 dollars in 2026 [28].
Norway reached 97.6% battery-electric share in the first half of 2026, and China nearly 55% of domestic car sales in 2025 [34][1].
Emerging economies outside China grew around 80%, with Thailand near a quarter of new car sales and India at a record 2.3 million electric vehicles [3][39].
United States used electric vehicle prices rose 5.1% in the first half of 2026 despite roughly 300,000 vehicles coming off lease [11].
Manufacturers have priced the volatility rather than the technology. At least 18 automakers are cancelling, delaying or scaling back electric plans in the United States, and the industry absorbed roughly 65 billion dollars in writedowns and losses by early 2026, including 19.5 billion at Ford, 27 billion at Stellantis and 7.6 billion at General Motors [35]. Those charges are not a verdict on batteries. They are a verdict on the forecastability of the policy environment the batteries were built for.
The contrast in design is the transferable lesson. China announced a partial, capped, multi-year taper with an effective date set months ahead, and absorbed the resulting pull-forward as a known cost [32]. The United States announced outright termination roughly three months ahead, producing a record 12% monthly share in September 2025 followed by the demand vacuum every analyst had predicted [41][31]. Neither approach is more generous than the other in aggregate outlay. They differ only in whether the market could plan, and residual values are a direct function of whether the market could plan.
What the Evidence Actually Tells Us
Three bottlenecks, one of them solvable this year
The physical asset outperformed expectations and the informational systems around it underperformed ◈ Strong Evidence [5][8]. That inversion — durable hardware, illegible markets — is the finding that organises everything above.
Take the claims in order of evidential strength. Battery durability is settled: 2.3% average annual degradation, 81.6% of capacity at eight years, under 2% full replacement across the post-2016 fleet [4][5][7]. Lifecycle emissions are settled: 63 grams against 235 for petrol on the projected European grid, with the manufacturing deficit repaid inside 17,000 kilometres [36]. Neither is where the remaining risk sits.
The first genuine bottleneck is informational. A used electric vehicle's most valuable component reports its condition to nobody the buyer can question, and the regulation that changes this arrives in 2027 with access restricted to parties holding a legitimate interest [37]. Until then, forty points of residual-value dispersion between similar cars is not market inefficiency waiting to be arbitraged away. It is the rational price of missing information [9][8].
The second is tenurial. Roughly 5% of United States rental properties offer charging, while 84% to 94% of single-family electric vehicle owners charge at home [19]. Renters pay three to four times more per kilowatt-hour for the same energy [19]. Because used vehicles flow disproportionately to renters, the charging estate determines the size of the market into which the lease returns of 2026 must clear [10][19].
The third is geopolitical and the least tractable. Over 90% of graphite processing, more than two-thirds of lithium and cobalt refining, and 58% of internationally traded processed battery minerals run through a single jurisdiction that has already demonstrated a willingness to regulate the flow [27][40]. This is a midstream problem, which is exactly why it is immune to the mining diversification that dominates the policy conversation.
The most consequential finding in this report costs almost nothing to act on. Mandatory, transferable state-of-health certification at the point of resale would convert the single largest source of used electric vehicle discount from an unpriceable unknown into an ordinary disclosed attribute, in the way that mileage and accident history already are. Every other bottleneck identified here needs capital or diplomacy. This one needs a form.
The grid question, by contrast, has been solved analytically and merely awaits implementation. Managed charging cuts distribution upgrade costs by about 30% and can reduce electric vehicle peak demand by 50% to 70%, with California modelling 5 to 18 billion dollars of avoided distribution cost by 2040 [24][25]. The instruments — time-of-use tariffs, default enrolment, interoperable telematics — already exist and are already approved somewhere.
None of these bottlenecks is technological, and that is the substance of the finding. Disclosure is a legislative act. Charging at multifamily buildings is a building-code and utility-tariff act. Managed charging is a rate-design act [25]. Only the midstream mineral exposure requires capital on a decade horizon, and even there the constraint is refining capacity rather than geology [27]. A transition described for fifteen years as an engineering problem has become, almost entirely, an institutional one.
What remains is the volatility itself. Four major jurisdictions changed the economics of electric vehicle ownership within fifteen months, in both directions [31][33][30][32]. Residual values are forecasts about the future policy environment as much as about the future condition of a battery, which is why they move faster than any physical property of the vehicle. The reckoning is not that electric vehicles failed to deliver. It is that the markets asked to price them were built for a product whose value does not depend on next year's legislation.