Advanced Battery Technologies

Advanced battery technology means the chemistries and cell architectures being developed to store electrical energy at lower cost, higher energy density, longer life or greater safety than the lithium-ion cell that currently does almost all of the work. In practice it means four families: incremental lithium-ion, which is where essentially all of the delivered progress has come from; solid-state, which is where essentially all of the attention has gone; sodium-ion, which is real, shipping and very small; and the long tail of announced chemistries that reach a press release and no further.

A unit convention, stated first, because almost every disagreement in this subject is an incomparability rather than a disagreement. A battery number is uninterpretable without three qualifiers. The first is level: a figure quoted for a cell, for a pack, and for an installed system differ by roughly a factor of two at each step, and quoting one against another is the characteristic error of the field. The second is which density: gravimetric energy density in watt-hours per kilogramme and volumetric in watt-hours per litre move independently, and a chemistry can gain on one while losing on the other. The third is the duty behind the number: cycle life is meaningless without the depth of discharge, the charge and discharge rate and the temperature at which it was measured, and a price per kilowatt-hour is meaningless without the segment, the chemistry, the geography and the year. This brief attaches those conditions to every figure it gives, and where a source omits them it says so.

Where this brief stops, and the boundary is load-bearing rather than administrative. This is a brief about cells: what is inside them, what they cost per kilowatt-hour of cell or pack, how many cycles they survive under stated conditions, why they fail, and which chemistries are winning. Storage as a system service — how many hours a grid buys, what a complete installed plant costs per kilowatt-hour of system energy, how it is dispatched and what it earns — belongs to energy storage revolutions, Energy Systems slot 9. The arbitration rule where the two meet is the unit of account. If the question is what a kilowatt-hour of cell or pack costs and why, it is here. If the question is how many hours anybody bought and what the plant cost, it is there.

Established The framing this brief was commissioned to test was that a breakthrough chemistry is imminent, and the evidence points the other way with unusual clarity. Between 2010 and 2025 the volume-weighted average lithium-ion pack price fell from US$1,474 per kWh to US$108 per kWh in constant 2025 dollars, a 93% real decline. Over the same fifteen years the number of all-solid-state cells in customer vehicles anywhere in the world went from zero to zero. The incumbent got cheap faster than any challenger got real, and the reason the announcements did not notice is that they were optimising energy density, which stopped being the binding constraint around 2022.

One consequence of that reading is worth flagging at the outset because it inverts the usual organisation of the subject. The most consequential battery development of 2025 and 2026 is not a new chemistry at all. It is that lithium iron phosphate — invented in 1996, and dismissed for two decades as too low-energy to matter — crossed from under half to over 55% of global electric-vehicle battery deployments, took roughly 90% of grid storage, and drove stationary-storage pack prices below electric-vehicle pack prices for the first time in the history of the industry.

Established The measured historical record is unusually good here, and it is the right place to start because it is the baseline every breakthrough claim is implicitly measured against. One peer-reviewed study harmonised more than ninety data series into 1,716 individual cell records covering 1991 to 2018. At the 98th percentile of the commercial distribution — the best cells actually on sale, not laboratory records — cell-level gravimetric energy density rose from about 80 Wh/kg to over 250 Wh/kg, and cell-level volumetric from about 200 Wh/L to over 700 Wh/L. Real price per kilowatt-hour fell 13.1% a year across all cell types, or 17.1% a year once the density gains are folded into the service delivered, on a learning rate of 20.4 to 26.6% per doubling of cumulative production. Those are among the fastest sustained improvement rates ever recorded for a manufactured good.

Established The cost curve has continued and is now visibly flattening. BloombergNEF's 2025 survey, published 9 December 2025 and assessing 320 data points, puts the global volume-weighted average pack price at US$108 per kWh, down 8% in real terms on 2024, and the cell-only average at US$74 per kWh, down 5%. The series behind it runs US$1,474 in 2010, US$139 in 2023, US$115 in 2024 and US$108 in 2025, all in constant 2025 dollars. BNEF's own projection for 2026 is just under US$105 per kWh, a decline of roughly 3%, against recent years that ran at 8 to 20%. Frontier That flattening is a forecast by a commercial analyst house rather than an observation, and it is flagged accordingly.

Established The segment split matters more than the headline, and the headline is the number everyone quotes. US$108 per kWh is a volume-weighted average across electric vehicles of several types, buses and stationary storage. It is not the electric-vehicle pack price, which is US$99. Stationary-storage packs came in at US$70, down 45% year on year; two- and three-wheelers at US$133. By chemistry, LFP packs were US$81 against NMC at US$128. By geography, China US$84, North America US$121 and Europe US$131 — a 56% spread concealed inside a single global figure. The lowest observed values in 2025 were a US$36 per kWh LFP cell and a US$50 per kWh LFP pack, both in stationary storage.

Established And the density curve has decoupled from the cost curve. The best shipping-class cells announced for 2026 sit near 280 Wh/kg at cell level; the commercial top of distribution measured for 2018 was over 250 Wh/kg at cell level. Frontier Taken as endpoints, that implies gravimetric improvement of roughly 1.4% a year since 2018, against roughly 4.3% a year over 1991 to 2018 — a clear deceleration, while cost went on falling at 8 to 20% a year. This is an inference from two well-sourced endpoints and not a published finding. No authoritative post-2018 replication of the harmonised-series methodology could be located; the arithmetic converting the 1991–2018 endpoints into an annual rate is this brief's, and the endpoints are sourced. It is recorded as an inference precisely because the field's habit of quoting endpoint comparisons without saying whose arithmetic they are is one of the things this brief is arguing against.

Established Against that baseline, the state of solid-state as of mid-2026 is a single fact. No all-solid-state electric vehicle is on sale anywhere in the world. The International Energy Agency's Global EV Outlook 2026 places solid-state in the prototype and testing phase; trade reporting in June 2026 states directly that no all-solid-state electric vehicles are currently on sale. What has shipped is semi-solid — SAIC's MG4 in 2025 — which retains a liquid or gel electrolyte and is a different thing.

Established The most valuable single piece of evidence on solid-state timing is an interested party speaking against its own commercial interest. In June 2026 the chairman of CATL, the world's largest cell manufacturer, with more than ten billion yuan sunk into a sulfide demonstration line, placed all-solid-state chemistry at technology readiness level 4 of 9 — laboratory validation and prototype engineering. He put large-scale commercialisation, defined as one million vehicles, as unattainable before 2030, with initial use confined to vehicles above ¥250,000. China's Ministry of Industry and Information Technology equipment development centre puts mass production at least three to five years away. Ouyang Minggao of the Chinese Academy of Engineering puts all-solid-state at five to ten years from reaching 1% market share. A technology at readiness level 4 with a five-to-ten-year path to one percent is not imminent by any ordinary use of the word.

Established The cost gap is the part of the solid-state case that the promotional literature does not quote, and it is the decisive one. All-solid-state cells currently cost ¥1.6 to ¥2.2 per watt-hour against mainstream LFP at ¥0.39 to ¥0.5 per watt-hour — a penalty of 3.2 to 5.6 times, at cell level. Lithium sulfide accounts for 50 to 64% of all-solid-state cell cost at roughly ¥2 million per tonne. A 70 kWh all-solid pack costs ¥70,000 to ¥80,000 more than the conventional pack it would replace. And sulfide electrolyte handling requires a dew point below −60 °C, so existing liquid-electrolyte production lines cannot be converted. Meanwhile the incumbent it must beat got 8% cheaper in 2025 alone. The gap the challenger has to close is widening in absolute terms even while the challenger's own cost falls.

Frontier The sharpest comparison available is on the metric solid-state exists to improve, and it goes the wrong way. QuantumScape's QSE-5 — the flagship of the Western solid-state sector, fifteen years and over three billion dollars in — is a 5 Ah cell measuring 84.5 by 65.6 by 4.6 mm, rated at 301 Wh/kg gravimetric and 844 Wh/L volumetric at cell level, charging 10 to 80% in 12.2 minutes, discharging at up to 10C and operating to −30 °C. Set 301 Wh/kg against the 2018 commercial top of distribution for liquid-electrolyte cells at over 250 Wh/kg: the breakthrough cell is roughly 20% ahead of where mainstream lithium-ion already was eight years earlier, and behind CATL's announced 350 Wh/kg condensed-matter cell. On volumetric density the comparison is closer — 844 against over 700 Wh/L in 2018 — and volumetric is the axis on which the QSE-5 is genuinely strong, which is a distinction the coverage of it consistently loses.

Established Then there is the market's revealed preference, which is the cleanest evidence in the whole brief. Given a free choice between a higher-density chemistry and a cheaper, longer-cycling one, the global industry chose the lower-density one. Lithium iron phosphate went from nearly half of the electric-vehicle battery market in 2024 to over 55% in 2025, growing 48% in a single year, in a market that deployed 1.2 TWh of electric-vehicle batteries. In China, more than 80% of electric vehicles sold from January to November 2025 used LFP. The trade-off it accepted is quantified: LFP is roughly 20% lower gravimetric and 33% lower volumetric energy density than NMC at cell level, and more than 40% cheaper per kilowatt-hour.

Established And the sacrifice was one metric, not several, which is stronger than the usual framing. LFP's cycle life is not a trade-off against NMC; it is better. Its tolerance of charging to 100% state of charge without the degradation penalty that confines NMC to roughly 80% in practice is not a trade-off; it is better. Its cost is better. The only axis on which it is worse is energy density — the headline one — and even that was substantially engineered back at pack level rather than cell level, through cell-to-pack architecture: CATL's Qilin CTP 3.0, announced June 2022 and in series production, reached 72% volume utilisation and 255 Wh/kg at pack level for ternary systems and 160 Wh/kg at pack level for LFP. Packaging closed part of the chemistry gap.

Established Finally, the failure record, which is one-directional. Since 2024 the casualty list runs: Northvolt, which raised over fourteen billion dollars and filed Chapter 11 in November 2024 and Swedish bankruptcy on 12 March 2025; Li-Cycle, into Canadian creditor protection on 14 May 2025 having spent US$485 million on a plant that never operated, sold to Glencore for about US$40 million; Natron Energy, the leading United States sodium-ion developer, which ceased operations in late August 2025 and cancelled a US$1.4 billion, 28 GWh plant; Ample, in bankruptcy in December 2025; 24M Technologies, shut down in March 2026 after promising a thousand-mile battery; Morrow Batteries, bankrupt in May 2026 having reached 1 GWh a year against a 42 GWh plan; Freyr, which cancelled 34 GWh and pivoted to solar; KORE Power, which cancelled 12 GWh; and iM3NY in Chapter 11. Not one of them was defeated by a rival breakthrough chemistry. Every one was defeated by cheap LFP, or by the capital markets' reaction to it.

The genuine frontier in this subject is not where the announcements put it. It divides into three parts: a solid-state interface and manufacturing frontier that is real and slow; a measurement frontier that is under-reported and may be invalidating a share of the published progress; and a set of claims that sound open but are not open, they are simply unsupported.

Established The dendrite problem is now mechanistically understood, and understanding it did not solve it. Molecular-dynamics work published in February 2025 shows that lithium dendrite penetration of a ceramic electrolyte is a fracture-mechanical failure rather than a purely electrochemical one: dynamically generated lithium deposits accumulate internal stress until the solid electrolyte fractures at the dendrite tip. Classical Griffith theory applies, but the fracture toughness has to be corrected for local lithium-ion concentration. In polycrystalline electrolytes, dendrites deflect toward and then propagate along grain boundaries in mixed Mode I and Mode II fracture. The authors frame the work as contributing knowledge rather than offering a solution, and this brief carries that framing intact.

Established The size of the gap is quantified and it is not small. Lithium dendrites form in lithium lanthanum zirconium oxide garnets at current densities exceeding 0.3 to 1 mA/cm²; one time-limited protocol found a critical current density as low as 0.3 mA/cm². Practically relevant operation requires more than 1 mA/cm². That is the whole engineering problem stated in one comparison, and it has been stated in roughly that form since 2023.

Established The under-reported finding, and the one this brief regards as the most consequential in the pack: the metric used to claim dendrite progress is not standardised. There is no field-wide agreed electrochemical protocol for measuring critical current density, and identical garnet surfaces yield different critical current densities depending on which protocol is used. The direct implication is that an unknown share of the published dendrite-suppression literature is measuring its own methodology rather than a materials improvement. This is not a marginal caveat about experimental hygiene. It means that the primary quantitative claim by which solid-state progress is reported to the outside world cannot currently be compared between laboratories, and this brief will not rank dendrite-suppression claims against each other for exactly that reason.

Established Stack pressure is the second real frontier, and until recently the numbers were disqualifying. Capacity loss in lithium-metal solid-state cells had conventionally been prevented only by stack pressures above 100 MPa — a pressure that is impractical and hazardous inside a vehicle pack, and which quietly disqualifies most laboratory cell results from being read as cell results at all. Frontier Genuine progress was published in October 2025: stable lithium-metal cycling below 1 MPa, at room temperature, at 5 mAh/cm² areal capacity and above 1 mA/cm², without interlayers, achieved by tuning the cathode chemomechanics rather than by mechanical compression. That is a two-order-of-magnitude reduction in the enabling pressure at practically relevant areal capacity, which is the most important solid-state result in the pack. It is single-laboratory and at cell scale, and it has not been reproduced in a multi-layer pouch format.

Established The third real frontier is manufacturing, and it is the one nobody publishes a roadmap for. Work on scaling sulfidic solid-state cells from laboratory to pilot identifies three barriers: solid-electrolyte precursors cost two orders of magnitude more than liquid electrolytes; processing remains substantially manual; and there is a “severe lack of knowledge” about what determines and maintains pressure uniformity at module-to-pack level. A cell that works in a small die under uniform load has no demonstrated path to a multi-layer pouch under a bolted pack, and the field does not currently know what would establish one.

Established An interface mechanism named by the manufacturer that is trying hardest to solve it. CATL's chairman identified solid–solid interface defects arising from warm isostatic pressing at roughly 6,000 atmospheres, where differing compaction densities cause structural misalignment, raising internal resistance and accelerating degradation. That a manufacturer with a demonstration line names a specific unsolved failure mechanism, in public, is better evidence about the state of the art than any performance claim from the same company.

Frontier Sodium-ion's frontier is not technical, and that is what makes it hard. Nothing unsolved stands between sodium-ion and deployment: CATL's Naxtra cell, launched 21 April 2025, is rated at 175 Wh/kg gravimetric at cell level with more than 10,000 cycles, 90% usable power at −40 °C and an operating range of −40 to +70 °C. The frontier is that sodium-ion's cost advantage is measured against a moving target. LFP cells reached US$36 per kWh in 2025. Every month LFP gets cheaper, the argument for sodium as a hedge against lithium prices gets weaker, and sodium-ion manufacturing capacity is just over 1% of lithium-ion capacity, which means it has no scale from which to answer.

Handwave What merely sounds open: the six-hundred-watt-hour cell. Chery claims a 600 Wh/kg cell with 1,500 km CLTC range in the Exeed Liefeng, deploying into ride-hailing and rental fleets in 2026 with mass production in 2027. The system is described as in-situ polymerised — that is, semi-solid, not all-solid-state. A 600 Wh/kg cell would be roughly 70% above CATL's most aggressive announced laboratory figure and roughly double any shipping cell. No mechanism is given and no third-party measurement exists. It is quoted here precisely because it is unbelievable: it is the purest available example of optimising the metric the press reports rather than the metric the buyer pays for.

Handwave Also merely sounding open: the five-hundred-watt-hour laboratory sample. CATL's “exceeds 500 Wh/kg” solid-state laboratory figure and its chairman's technology-readiness-level-4 assessment come from the same company two months apart, and its April 2026 product day announced no solid-state product at all. The brief states both and treats the readiness assessment as the more informative one, because it is the one that costs the company something to say.

What this brief could not establish, stated as gaps rather than smoothed over. No authoritative post-2018 replication of the harmonised cell-improvement methodology could be found, so the post-2018 density deceleration rests on endpoint arithmetic rather than a published series. No government or peer-reviewed source gives a defensible pack-level volumetric time series in watt-hours per litre; vendor volume-utilisation percentages are the only hard datum, which is why this brief quotes pack-level Wh/kg and declines to quote pack-level Wh/L. No credible 2025 or 2026 global battery-metal recovery-rate figure exists; the most recent authoritative number is from 2023. And no source successfully consulted contained a Samsung SDI-specific all-solid-state milestone for 2025 or 2026 — the company is named here as a participant whose position this brief could not verify, and no number is attached to it. Several sources that would have been useful were unreachable: an IEEE Spectrum piece on Toyota's 2027 target, two Joule papers on the parameters distinguishing coin cells from pouch cells and on lithium–sulfur pouch-cell evaluation, an Advanced Materials 2025 review of low-pressure all-solid-state batteries, and an ACS Energy Letters review of critical-current-density measurement. They are named because a reader should know what was not read.

Established The first bottleneck is that the incumbent is a moving target, and it binds on everything else. Any challenger chemistry must beat LFP at US$81 per kWh at pack level and US$36 per kWh at the observed cell floor, on cycle life that is already better than NMC's, at a scale of 1.59 TWh of annual lithium-ion demand. All-solid-state currently sits 3.2 to 5.6 times above LFP on cell cost. The incumbent got 8% cheaper in 2025 and is forecast to get roughly 3% cheaper again in 2026. Every other bottleneck in this list is a bottleneck because it delays the challenger past the point where the gap can be closed.

Established The second is lithium sulfide, which is half the cost of a solid-state cell. At roughly ¥2 million per tonne it accounts for 50 to 64% of all-solid-state cell cost. That is a materials-supply and process-chemistry problem with no announced solution, and it means the solid-state cost curve is dominated by a single input rather than by manufacturing learning — which is the same structural situation that keeps other frontier materials expensive regardless of volume.

Established The third is that solid-state manufacturing cannot reuse the existing industrial base. Sulfide electrolyte handling needs a dew point below −60 °C. Existing liquid-electrolyte lines cannot be converted; the capacity has to be built new, into a market with visible overcapacity in conventional cells. Processing is still substantially manual at pilot scale, and precursor costs run two orders of magnitude above liquid electrolytes.

Established The fourth is pressure uniformity at pack level, and it is the least solved. The conventional stack-pressure requirement was above 100 MPa, and even the October 2025 result that brought it below 1 MPa is single-laboratory and cell-scale. The scale-up literature identifies a “severe lack of knowledge” about what sets and maintains pressure uniformity from module to pack. This is the bottleneck that separates a working cell from a working vehicle pack, and no source consulted for this brief describes a demonstrated route across it.

Established The fifth is measurement, and it binds in an unusual way: it makes progress unverifiable rather than impossible. There is no agreed protocol for measuring critical current density in garnet electrolytes, and identical surfaces give different values under different protocols. Laboratory cells run flooded with electrolyte in a way that structurally hides the capacity-plunge failure mode that lean commercial cells exhibit. Together these mean that the two headline claims a solid-state developer can make — dendrite suppression and cycle life — are both measured in ways that do not transfer to a commercial cell. An investor or a procurement officer cannot currently distinguish materials progress from protocol choice from published data.

Frontier The sixth is capital, and in 2025 and 2026 it became the binding constraint on Western manufacturing regardless of chemistry. Northvolt burned roughly US$100 million a month, raised over US$14 billion including a US$5 billion debt package in January 2024, lost a US$2 billion BMW contract in June 2024 over delivery delays, laid off 1,600 people and failed. Morrow Batteries reached commercial production in January 2026 at 1 GWh a year against a 42 GWh 2028 plan, shipped its first cells in April 2026 and filed for bankruptcy weeks later, its board citing oversupply and price pressure, rising capital costs, industrialisation delays and a restrained investment market. In the United States, the removal of key funding and incentives mid-build compounded a pipeline already shedding gigawatt-hours. None of this is a chemistry constraint and all of it determines which chemistries get built.

Established The seventh is recycling feedstock, and it is arithmetic rather than chemistry. If every announced project proceeded, global 2030 recycling capacity would be roughly seven times the available feedstock. Manufacturing scrap is two-thirds of 2030 feedstock; end-of-life electric-vehicle batteries do not exceed 90% of feedstock until around 2050. In Europe the mismatch is documented precisely: roughly 270 kt a year of pretreatment and 290 kt a year of refining capacity announced, rising to roughly 520 kt and 820 kt by 2030, against return volumes of about 100 kt a year now and a 2030 forecast of up to 270 kt a year — itself a downgrade of roughly 40% on the previous year's forecast, because cell production expanded more slowly than planned. Recyclers are not failing at chemistry. They are failing at arithmetic on input volumes.

Established What is not a bottleneck, stated plainly because the announcements say otherwise. Cell gravimetric energy density is not the binding constraint for road transport and has not been since roughly 2022; the market demonstrated as much by moving to a chemistry 20% worse on that axis. Volumetric density is not binding either, having been substantially recovered at pack level by cell-to-pack architecture. Fast charging is not binding at the cell level: shipping LFP and announced solid-state cells both do 10 to 80% in the low teens of minutes. The binding constraints are cost per kilowatt-hour, cycle and calendar life under real duty, cold performance, safety certification, and manufacturability at hundred-gigawatt-hour scale.

Established Nothing in this brief waits on a physics result, and that is the correct starting point. Incremental lithium-ion depends on electrode process engineering, coating precision and cell-to-pack architecture, all of which are being delivered continuously and none of which is a discovery. Sodium-ion depends on nothing unsolved at all: it is shipping, certified and installed, and what it depends on is lithium staying expensive enough to be worth hedging against, which is a market condition rather than a technical one and is currently moving against it.

Established Solid-state depends on four things in series, and each is industrial. Lithium sulfide at a fraction of its current price, since it is 50 to 64% of cell cost. Dry-room manufacturing at below −60 °C dew point, built new because existing lines cannot be converted. A solution to pressure uniformity from module to pack, which the scale-up literature identifies as an acknowledged knowledge gap. And a measurement regime that lets a buyer distinguish a materials improvement from a protocol choice. The conjunction is why credible internal assessments put readiness at level 4 of 9.

Frontier Everything here now depends on capital formation in a way it did not five years ago. The Western manufacturing failures of 2024 to 2026 were not technical. Northvolt raised over US$14 billion and failed on production ramp and burn rate; Morrow reached commercial production and failed on oversupply and price pressure weeks after its first shipment; a United States policy reversal removed funding and incentives mid-build. A chemistry that cannot be financed to hundred-gigawatt-hour scale in the West will be commercialised in China or not at all, and global nameplate manufacturing capacity is already above 4 TWh with China holding more than 80% of it against 6 to 7% each for the European Union and the United States.

Frontier Recycling depends on feedstock that does not exist yet, and this is the one dependency running backwards. Announced 2030 global capacity is roughly seven times available feedstock. Manufacturing scrap is two-thirds of 2030 feedstock, so recycling capacity is currently a derivative of cell production volume rather than of vehicle retirement — and when cell production expanded slower than planned, the European feedstock forecast for 2030 was cut by roughly 40%.

Established What depends on this brief. Electric road transport depends on it entirely and is priced by it. Energy storage revolutions depends on it for the cost floor beneath every two-to-four-hour lithium-ion system, though the duration question it addresses is not answered by any cell improvement. Planetary-scale energy systems and energy corridors both treat storage cost as an input to whether transmission or storage is the cheaper answer to the same problem.

Established The most valuable experiment in this subject has already run to completion, and it produced a negative result that almost nobody has absorbed. Nio's 150 kWh WeLion semi-solid pack is the only case in the world of a solid-state-adjacent product completing the entire journey. It was announced in January 2021, first cells followed in November 2022, delivery in July 2023, series production in April 2024 — and production was discontinued after only a few hundred units, confirmed on 17 November 2025. It worked: roughly 1,000 km on the CLTC cycle. It was killed by cost and indifference. A co-founder said the pack cost about as much as an entire ET5, a car starting at ¥298,000; it was offered on subscription only; and the chief executive said 97% of users preferred the 75 kWh pack over the 100 kWh one given the availability of battery swapping. The experiment tested whether customers would pay for energy density, and the answer was no.

Established A second completed experiment, on the Western side, tested whether a superior chemistry could be commercialised into a falling-price market. Natron Energy was the leading United States sodium-ion developer. It ceased operations in late August 2025, permanently closing facilities in Michigan and California and cancelling a US$1.4 billion, 28 GWh North Carolina plant, with certification delays preventing order fulfilment. The analyst readings are the finding: that targeting sodium-ion meant a much smaller pool of demand, and that there is limited value in fifty-thousand-cycle performance if other metrics are worse, because the bar keeps going up. 24M Technologies ran the same experiment on semi-solid thick-electrode manufacturing from 2010, reached a valuation over a billion dollars on a thousand-mile-battery target, and shut down in March 2026.

Frontier The decisive live experiment for solid-state is dated and public: whether Toyota ships in 2027 or 2028. The enabling test is not the vehicle but the electrolyte plant. Idemitsu Kosan broke ground on its solid-electrolyte facility on 30 January 2026, for completion at the end of 2027, at a capacity of “several hundred tonnes” a year. That is the readout: a plant finishing months before the deadline, at a capacity consistent with limited batches and nothing more. If the deadline holds and the volume is small, the correct conclusion is that the schedule was met by shrinking the scope, which is the honest version of the Toyota story.

Frontier QuantumScape's experiment is a manufacturing one and its readout is a line, not a cell. B1 samples — the first cells combining the QSE-5 design with the Cobra separator process — shipped on 22 October 2025. As of the first quarter of 2026 the Eagle Line installation was complete and startup operations had commenced, with the company still optimising equipment uptime, line throughput, control systems and process stability. Customer billings of US$11 million are development and ecosystem payments rather than product revenue; net loss was US$100.8 million against liquidity of US$904.7 million. No commercial launch date is given; the near-term milestone is field testing with a partner, and the commercial relationship is described as moving toward licensing. The recipients of the B1 samples were not named, and the named demonstrator is a race motorcycle rather than a road car. The experiment to watch is not the next cell specification. It is whether the line produces at yield.

Frontier The sodium-ion experiment has a date at the end of 2026 and a vendor that contradicts itself about where it already is. CATL described Naxtra as the world's first mass-producible sodium-ion battery in April 2025, said in December 2025 that it had begun large-scale production with deployment at scale across four sectors in 2026, and then in its own April 2026 release put full-scale mass production at the end of 2026 with no energy density disclosed. Those statements cannot all describe the same state of affairs, and this brief does not resolve the contradiction; it records that pilot and early-series volumes are real and that volume production was not complete as of August 2026. The first actual vehicle — a CATL and Changan passenger car announced on 5 February 2026 — has a stated pure-electric range above 400 km against the 500 km advertised for the cell, with market launch in mid-2026 and no production volumes disclosed.

Frontier The experiment the field most needs is one nobody is funding: a standardised critical-current-density protocol. Until identical garnet surfaces give the same number in different laboratories, published dendrite-suppression progress cannot be aggregated, and the low-pressure cycling result of October 2025 cannot be properly compared to what preceded it. Replicating that result — below 1 MPa, at 5 mAh/cm² and above 1 mA/cm² — in a multi-layer pouch cell at a second institution would be the single most informative experiment available in solid-state today, and it is a cheaper experiment than any of the gigafactories in the failure record.

Established The single most useful thing to understand about battery numbers is the level at which each one is quoted, because the steps between levels are large and systematic. A cell is the electrochemical unit. A pack is cells plus interconnects, cooling, structure, sensing and management electronics, and it is heavier and larger than the sum of its cells. A system is a pack plus power conversion, containers, fire suppression, civil works and connection. Energy density falls at each step and cost per kilowatt-hour rises. The 2025 survey figures make the size of the step visible directly: US$74 per kWh at cell level and US$108 at pack level, a 46% increment for the same stored energy.

Set out with their scopes attached, the figures in circulation for 2025 stop contradicting each other:

Established The same discipline applies to energy density and is applied less often. The 2018 commercial top of distribution — over 250 Wh/kg and over 700 Wh/L — is a cell figure at the 98th percentile of what was on sale. CATL's Qilin CTP 3.0 figures of 255 Wh/kg for ternary and 160 Wh/kg for LFP are pack figures. Those two numbers being similar is a coincidence of arithmetic, not a comparison: a pack that matches a 2018 cell on gravimetric density is a considerable engineering achievement, and a reader who takes it as a cell-level comparison concludes the opposite. The IEA benchmark of 205 Wh/kg for LFP and up to 175 Wh/kg for sodium-ion are cell figures. QuantumScape's 301 Wh/kg and 844 Wh/L are cell figures for a 5 Ah pouch. CATL's announced 280 Wh/kg third-generation Qilin and 350 Wh/kg, 760 Wh/L condensed-matter cells are cell figures.

Frontier Cell-to-pack architecture is the delivered engineering advance of the period, and it is packaging rather than chemistry. Removing module structure and integrating cells directly into the pack raised volume utilisation to 72% in the Qilin design. That is what allowed a chemistry with 33% lower volumetric density at cell level to be competitive at vehicle level, and it is why the LFP transition happened when it did rather than five years earlier. It is also why pack-level and cell-level trends have diverged: pack-level improvement over the last four years has come substantially from structure, and cell-level improvement has been slow.

Established Three laboratory-to-commercial gaps are quantified well enough to be engineering facts rather than folklore. The first is electrolyte volume. Commercial cells run electrolyte-to-capacity ratios below 2 µL/mAh, which is a remarkably low electrolyte content, and reducing electrolyte from laboratory-standard levels to lean levels shortened cycles-to-capacity-plunge by 59.5% and 82.3% at the two lean levels tested. Standard laboratory testing with excess electrolyte structurally cannot detect the capacity-plunge point at all, so a flooded coin cell and a commercial cell with identical nominal capacity and identical reported cycle life are not comparable objects. The second is stack pressure: any solid-state result reported at high stack pressure and low areal capacity is a materials measurement rather than a cell result, and the conventional requirement was above 100 MPa. The third is areal loading and processing: precursor cost two orders of magnitude above liquid electrolytes, manual processing, and no established knowledge of pressure uniformity from module to pack.

Established Sodium-ion's product range shows why a single headline figure is the wrong summary. The vendor specifications compiled for grid-oriented sodium cells sit well below the 175 Wh/kg electric-vehicle headline, and their cycle lives are quoted under conditions that differ from one another, so the cycle column is not internally comparable:

These are vendor figures compiled by trade press, not independent measurements, and the retention or rate condition attached to each cycle count differs between rows — 80% retention in one, a 0.5P rate in another, 1P in a third, unstated elsewhere. A ranking built from that column would be an artefact of the conditions rather than a comparison of the products. What the table does establish reliably is the spread: grid-oriented sodium cells run 95 to 160 Wh/kg gravimetric at cell level, so anyone quoting 175 Wh/kg for storage is quoting the electric-vehicle product.

The nearest neighbour is energy storage revolutions, and the two briefs are best read together because each is unpersuasive alone. This brief establishes that the cell got cheap and that the cheap chemistry is the low-density one. That brief establishes that the resulting industry built a two-and-three-quarter-hour revolution, and that the long-duration problem which determines whether a high-renewables grid survives a windless fortnight is untouched by anything in this brief. A cell can win decisively on cost per kilowatt-hour and leave the grid question entirely unanswered, which is why the boundary between the two is drawn at the unit of account rather than at the technology.

Energy corridors is adjacent in the substitution sense: transmission and storage solve overlapping problems, and the falling cost of the cell shifts the economic boundary between them without settling it. Planetary-scale energy systems treats storage as one enabling condition among several, and takes its cost inputs from here.

Circular economies is adjacent in the strong sense for one specific reason: battery recycling is the best-documented instance in the corpus of an industry building processing capacity against a feedstock stream that arithmetic says will not arrive, with announced 2030 capacity at roughly seven times available feedstock and a wave of Western closures to show for it. Anyone arguing about circularity in general should look at what happened to the people who tried it here.

Critical-minerals supply is adjacent in the weak sense that matters for reading the news: lithium, nickel and cobalt prices set part of the cell cost, but the 2025 price survey found pack prices falling 8% in real terms despite rising metals prices, which is a result about manufacturing rather than about mining. The mineral question that does bind is the one inside a solid-state cell rather than a conventional one: lithium sulfide at roughly ¥2 million a tonne is half the cost of an all-solid-state cell, which makes the challenger chemistry far more exposed to a single input price than the incumbent it is trying to displace.

Established The defining institutional fact about this field is that its price signal is a commercial product. The figure every government, analyst and journalist quotes for battery cost is a proprietary annual survey by an analyst house owned by a financial-data company. It is volume-weighted across end uses, it is not a transaction index, its segment, geography and chemistry breakdowns are published only to clients, and its respondent count is not disclosed — the only public methodology figures are “volume-weighted” and “320 data points assessed”. There is no public statistical agency anywhere producing a battery price index. An industry of 1.59 TWh a year, absorbing tens of billions in public subsidy across three continents, has no publicly auditable price series, and every policy document that cites one is citing a commercial survey.

Established The second institutional absence is a standards body for the metric the field claims progress on. No agreed protocol exists for measuring critical current density in solid electrolytes, and identical surfaces yield different values under different protocols. There is no equivalent of a metrology institute arbitrating solid-state performance claims, which means procurement officers, investors and grant panels are comparing numbers that are not comparable and have no institution to appeal to. The absence is cheap to fix relative to the sums being allocated on the strength of those numbers, and nobody has fixed it.

Established Certification is a real institutional gate and it has killed at least one company. The leading United States sodium-ion developer ceased operations in late August 2025 with certification delays preventing order fulfilment among the causes. On the other side, Chinese national standards work and Japanese ministry certification — Toyota obtained its in September 2024 — function as enabling institutions where they move quickly. Certification throughput is not usually counted as a technology constraint. In this sector it demonstrably is one.

Established Who buys, and it is not who the announcements address. The purchasers that set specifications are vehicle manufacturers and, increasingly, grid developers — and grid duty prices cycle life, cost per kilowatt-hour and footprint while treating mass and volume as nearly irrelevant. That is why roughly 90% of grid deployments are LFP, why stationary-storage packs fell 45% in a year to US$70 while electric-vehicle packs fell to US$99, and why North American stationary storage rose to 26% of battery demand from 16%, with China's stationary share reaching 45% of demand in December 2025 alone. A buyer indifferent to energy density now accounts for a quarter of the market and is growing at twice the rate of the vehicle segment. The chemistry roadmap follows that buyer, whatever the announcements say.

Frontier Who funds, and the geography of the answer is now the central industrial-policy fact. Global nameplate cell manufacturing capacity exceeds 4 TWh, with China above 80% and the European Union and the United States at 6 to 7% each. Western capacity has been contracting through insolvency and cancellation rather than expanding: over US$14 billion raised and lost at one Swedish manufacturer, 34 GWh cancelled at another developer, 12 GWh at a third, a Norwegian manufacturer bankrupt weeks after its first commercial shipment, and a United States policy reversal removing funding and incentives mid-build. The concentration was produced by cost structure and capital patience, not by a technology gap.

Established Who does not exist: an institution that owns the feedstock arithmetic. Recycling capacity is being announced against volumes that will not materialise — roughly seven times available feedstock globally by 2030, with European return volumes at about 100 kt a year now and a 2030 forecast cut by roughly 40% in a single year because cell production expanded slower than planned. Individual firms cannot see the aggregate, and no regulator publishes it as a planning constraint, so each entrant builds as though the feedstock were theirs. Eramet, BASF at Tarragona, Umicore, Northvolt Revolt and Li-Cycle are what that looks like in practice, and the two largest failures alone removed roughly 100 kt a year from announced 2030 European capacity. This is the clearest case in the corpus of an industry that would be better off with a shared, published, boring number.

Frontier What would change the institutional picture, stated as a testable condition. A published, independent, transaction-based cell and pack price index, and an agreed critical-current-density protocol with an institution behind it. Neither is expensive. Both would make the difference between a field where claims can be compared and a field where they cannot, and the absence of both is why a technology at readiness level 4 can be discussed in public as though it were imminent.

The primary ethical question in this field is evidence quality, and it is unusually acute because almost every headline number originates with a party that benefits from it. Established Cell specifications come from manufacturers. Roadmaps come from manufacturers. Range demonstrations come from manufacturers. The independent checks available — intergovernmental statistics, peer-reviewed measurement literature and insolvency filings — are slower, less specific and much less quoted.

Established The most valuable evidence in this brief is an interested party speaking against its own interest, and that is a statement about how thin the independent record is. The single most informative datum on solid-state timing is the chairman of the world's largest cell manufacturer placing the technology at readiness level 4 of 9 and calling one million vehicles unattainable before 2030, two months after his own company's product day announced no solid-state product. That this is the strongest available source, rather than an independent assessment, is itself the finding. This brief marks interested parties in its reading list and relies on intergovernmental and peer-reviewed sources for every comparison it draws between chemistries.

Established There is a research-integrity problem here that is structural rather than fraudulent, and it deserves to be named as such. Identical garnet surfaces yield different critical current densities under different measurement protocols, and no field-wide protocol exists. Laboratory cells flooded with electrolyte structurally cannot detect the failure mode that lean commercial cells exhibit, and running lean shortens cycles to capacity plunge by 59.5 to 82.3%. Neither of these is misconduct. Both mean that a substantial share of published progress may not be progress, and that the researchers producing it may not be in a position to know. The ethical obligation falls less on individual authors than on the reviewers, funders and journalists who treat an unstandardised metric as though it were a standardised one.

Established Public money is exposed in this sector at a scale that has not been publicly reconciled. A United States conditional loan commitment of US$850 million supported a 12 GWh project that was cancelled. A conditional loan commitment of US$375 million, later raised to US$475 million, supported a recycling hub on which US$485 million was spent and which never operated, whose parent could not meet the equity requirement and which was ultimately sold for about US$40 million. Norwegian state loans of NOK 550 million and grants of NOK 202 million — on one of the two conflicting accounts — supported a cell maker that reached 1 GWh a year against a 42 GWh plan before failing. And a United States policy reversal removed funding and incentives from projects that had already begun construction, which imposes its own costs on firms that behaved as the policy asked them to. None of this establishes that the spending was wrong. It establishes that it has not been audited in public against what was delivered.

Frontier An opportunity-cost question that is rarely posed as one. Fifteen years and more than three billion dollars at a single solid-state developer produced a cell roughly 20% ahead of where mainstream lithium-ion already was in 2018. Over the same period incremental process engineering and packaging on an unfashionable 1996 chemistry took roughly a fifth off the delivered cost of stored energy every year. If the objective is decarbonisation per dollar, it is not obvious that the allocation between those two activities has been the right one, and the question is not usually asked because the second activity does not generate announcements.

What this brief does not resolve, stated as an obligation rather than an omission. It does not resolve whether the post-2018 slowdown in cell-level gravimetric density is real, because no published replication of the harmonised methodology exists and the deceleration is an endpoint inference. It does not resolve CATL's contradictory statements about whether its sodium-ion cell is in large-scale production or reaches full-scale production at the end of 2026; it states both. It does not resolve the two trade sources that date Morrow Batteries' bankruptcy filing to 6 May 2026 and to 9 May 2026 respectively, and that give its funding as NOK 5.1 billion all in and as NOK 3.3 billion of equity plus NOK 550 million of state loans and NOK 202 million of grants; both are low-provenance outlets and a Norwegian court filing would settle it. And it attaches no number to Samsung SDI's solid-state position, because no source it could reach contained one.

Established The revolution that actually happened is not a chemistry, and mistaking which one happened has consequences for what gets funded next. What transformed energy storage between 2010 and 2026 was manufacturing scale, cell-to-pack packaging and a Chinese cost structure — applied to lithium iron phosphate, a chemistry invented in 1996 and dismissed for two decades as too low-energy to matter. Battery storage added 108 GW in 2025, forty percent more than in 2024 and eleven times the installed base of 2021, using the lowest-energy-density mainstream lithium chemistry available. The civilisational lesson is not about batteries. It is that a mature technology being manufactured better can outrun a superior technology being invented, and that the press and the capital markets are systematically bad at telling those two stories apart.

Established The general principle this case illustrates is that industries optimise the metric the buyer pays for, and commentary optimises the metric that is easy to report. Energy density is easy to report: one number, larger is better, no conditions required. Cost per kilowatt-hour under stated chemistry, segment and geography; cycle life at stated depth of discharge and rate; calendar life; cold performance; certification throughput; manufacturability at hundred-gigawatt-hour scale — these are what determined every purchasing decision of the last five years, and none of them fits in a headline. The gap between the two lists is where the entire promotional literature of this field lives.

Frontier What genuinely turns on this at civilisational scale is the decarbonisation of road transport and the two-to-four-hour firming of variable renewables, and both are already happening on the existing chemistry. Global lithium-ion demand reached 1.59 TWh in 2025, up 29%, with electric vehicles at 75% and stationary storage growing at 51% a year against the vehicle segment's 26%. Nameplate manufacturing capacity is above 4 TWh. None of that required a breakthrough and none of it is waiting for one.

Speculative What a genuine breakthrough would change is narrower than it is usually made to sound. A cell at 500 Wh/kg gravimetric and LFP's cost would not much alter grid storage, where mass is nearly irrelevant and roughly 90% of deployments already chose the low-density option. It would matter for aviation, for heavy long-haul, and for any application where mass is the binding constraint rather than cost. That is a real prize and it is a different prize from the one the announcements describe, which is usually a car that already exists doing what it already does with a smaller battery.

Established A quieter significance, which the failure record makes visible. Between 2024 and 2026 the West lost or cancelled a large fraction of its announced cell and recycling capacity — over fourteen billion dollars at Northvolt alone, plus 34 GWh at Freyr, 12 GWh at KORE, 28 GWh at Natron, and a recycling plant on which US$485 million was spent and which never operated. Manufacturing capacity is now above 80% Chinese against 6 to 7% each for Europe and the United States. That concentration was not produced by a technological gap. It was produced by cost structure, capital patience and industrial policy, which is a more uncomfortable finding than a technological one because it cannot be fixed by a discovery.

Established What already happened, because the timeline in this subject usually starts at the announcements rather than at the delivered record. Cell-level gravimetric energy density tripled and volumetric more than tripled between 1991 and 2018. Pack prices fell from US$1,474 per kWh in 2010 to US$139 in 2023, US$115 in 2024 and US$108 in 2025, all in constant 2025 dollars. Cell-to-pack architecture arrived in series production in 2022 at 72% volume utilisation. LFP passed half of the global electric-vehicle battery market during 2024 and reached over 55% in 2025. Grid storage went to roughly 90% LFP, up from well below 50% five years earlier, and stationary-storage pack prices fell below electric-vehicle pack prices for the first time in 2025. The first mass-production sodium-ion passenger vehicle was announced on 5 February 2026. Northvolt filed Chapter 11 in November 2024 and Swedish bankruptcy on 12 March 2025; Lyten completed the acquisition of its Swedish units on 27 February 2026.

Handwave 2027 to 2028: Toyota's all-solid-state target, in limited batches. The date has been stable in public since September 2023 and was restated in October 2025 and January 2026, which is better than this company's reputation suggests. What moved is the scope: from solid-state in hybrids by 2025, to electric vehicles in 2027 or 2028, to 2027 or 2028 in limited batches, with the enabling electrolyte plant completing at the end of 2027 at several hundred tonnes a year. Flagged handwave not because the date has slipped repeatedly but because the deliverable behind it has shrunk while the date held.

Handwave 2027: CATL and BYD all-solid-state small-batch production; 2030 for scale. CATL's reported roadmap is semi-solid mass production in 2026, all-solid-state small-batch in 2027 at roughly 5 GWh targeting readiness levels 7 to 8, and scaled mass production around 2030. BYD states limited batches in 2027 and mass production around 2030 on the sulfide route, with its chief scientist saying the technology is at a critical stage with hurdles remaining and that solid-state should complement rather than replace liquid lithium-ion. SAIC states 2027; Changan states prototypes at the end of 2026 and production in 2027. These are announced roadmaps from interested parties, and they sit against the same company's assessment of readiness level 4 of 9 in June 2026.

Handwave 2027: Chery mass production of a claimed 600 Wh/kg semi-solid cell, after 2026 fleet deployment. Recorded as a schedule attached to a specification this brief regards as unsupported.

Established The two independent assessments that bound the whole schedule. China's Ministry of Industry and Information Technology equipment development centre puts all-solid-state at least three to five years from mass production, as of 2026. Ouyang Minggao of the Chinese Academy of Engineering puts it five to ten years from reaching 1% market share. Neither is a vendor. Both are later than every vendor roadmap above, and the divergence is itself the finding.

Frontier End of 2026: CATL's stated full-scale mass production of sodium-ion, and the resolution of its own contradictory production statements.

Speculative Late 2020s: Mercedes-Benz and Factorial state production “by the end of the decade” for the quasi-solid FEST cell. The demonstrator covered 1,205 km on one charge from Stuttgart to Malmö in September 2025, in a lightly modified EQS, at an average speed that was not disclosed, carrying roughly 25% more energy than the standard pack at similar mass and volume. The pack requires pneumatic actuators to accommodate cell volume change during cycling, which is a substantial unresolved packaging problem, and this brief treats the production date as a reasoned intention rather than a schedule.

• Depends on Nothing on this map. No constraint in this brief is a physics result and none is a result in another brief here. Every binding constraint is a price, a process, a measurement convention or a source of capital, which is recorded on the row below. That is the cleanest statement of this topic's position: a breakthrough chemistry would be welcome, and its absence is holding nothing up.
• Requires (not on this map) Lithium sulfide at a fraction of roughly ¥2 million a tonne, since it is half the cost of a solid-state cell. Dry-room manufacturing below −60 °C dew point, built new rather than converted. Pressure uniformity from module to pack, an acknowledged knowledge gap rather than an engineering task with a known method. A standardised critical-current-density protocol, without which dendrite-suppression progress cannot be compared between laboratories — the only scientific requirement on this list and a metrological one. Capital willing to finance hundred-gigawatt-hour cell manufacturing through a period of overcapacity. Recycling feedstock, where announced 2030 capacity is roughly seven times the volume that will exist to feed it. And certification throughput: sodium-ion's leading Western developer was killed in part by certification delays preventing order fulfilment.
• Enables Electric road transport, and the two-to-four-hour battery that has become the fastest-growing power technology in the world. The system-level consequences of that — duration, dispatch, revenue — are argued in energy storage revolutions rather than here, and the enabling edge stops at the pack.
• Adjacent Energy storage revolutions, the sibling slot and the consumer of everything priced here; energy corridors, since transmission and storage are partial substitutes and the choice between them is economic; planetary-scale energy systems, where storage is an enabling condition rather than the subject; and circular economies, where battery recycling is the best-documented case of capacity built against feedstock that does not exist.

“Battery packs cost $108 per kilowatt-hour.” Established That figure is a volume-weighted survey average across all end uses — electric vehicles of several types, buses and stationary storage — drawn from 320 assessed data points, and it is not the price of anything in particular. The electric-vehicle pack price is US$99. Stationary storage is US$70. Two- and three-wheelers are US$133. LFP packs are US$81 and NMC US$128. China is US$84 and Europe US$131. Cells, not packs, are US$74. Using the headline as though it were an electric-vehicle pack price is the most common quantitative error in coverage of this subject, and it is also worth knowing that the survey does not disclose respondent counts — the only public methodology figures are “volume-weighted” and “320 data points assessed”.

“Toyota keeps slipping its solid-state dates.” Established This is largely folklore and repeating it makes the criticism weaker, not stronger. One clean slip is documentable: an executive's target of solid-state in hybrids by 2025, superseded around mid-2023 by a 2027-or-2028 target for electric vehicles. That 2027–2028 date has then held unchanged from September 2023 through October 2025 to January 2026 — three years of a stable public target, with a partner breaking ground on the enabling electrolyte plant on 30 January 2026 and Japanese ministry certification obtained in September 2024. There may be earlier dates this brief could not verify against a fetched source, and it does not assert them. The accurate criticism is different and sharper: the date held while the scope quietly shrank to limited batches, and the electrolyte plant completes at the end of 2027 — months before the deadline — at several hundred tonnes a year.

“Sodium-ion is where the real 2025–26 news is.” Established It is not, and this brief previously implied otherwise. Sodium-ion manufacturing capacity is just over 1% of lithium-ion capacity. The leading Western sodium-ion company shut down in August 2025 and cancelled a US$1.4 billion, 28 GWh plant. The first sodium-ion passenger car reached announcement only in February 2026, with more than 400 km of real range against 500 km advertised for the cell. Sodium-ion is real, shipping and correctly positioned for cold climates and stationary duty — and it is a rounding error. The real news of 2025 and 2026 is that grid storage added 108 GW in a single year, roughly 90% of it LFP, and that stationary pack prices fell below electric-vehicle pack prices for the first time.

“Nio ships a 150 kWh semi-solid pack.” Established It does not; the product is dead. Production was halted in November 2025 after only a few hundred packs, because the pack cost about as much as an entire ET5 and 97% of customers preferred the 75 kWh option given battery swapping. This is not a footnote. It is the single best-documented case anywhere of a solid-state-adjacent product completing the full journey from announcement in January 2021 through series production in April 2024 to withdrawal in November 2025, and it deserves to be in the body of any argument about whether customers will pay for energy density.

“The market accepted worse batteries to save money.” Established Only on one axis. LFP's cycle life is better than NMC's, not worse; its tolerance of charging to 100% state of charge without the degradation penalty that confines NMC to roughly 80% is better; its cost per kilowatt-hour is more than 40% better. The only metric surrendered was energy density — the headline one — at roughly 20% gravimetric and 33% volumetric at cell level, and even that was substantially engineered back at pack level through cell-to-pack architecture. The industry gave up one metric and improved on every other axis it cared about.

“There is a cost per kilowatt-hour for installed battery storage.” Frontier There is not, and the disagreement is large enough to be a finding rather than noise. One analyst survey puts global turnkey system cost at US$117 per kWh in 2025, down 31%, with China at US$73, Europe at US$177 and the United States at US$219. A second puts all-in capital cost at US$125 per kWh for four-hour-plus projects excluding China and the United States, split roughly US$75 of Chinese core equipment and US$50 of installation and connection. A United States national laboratory puts a four-hour utility-scale system at US$334 per kWh for 2024 in 2024 dollars on a bottom-up model including balance of system, labour, permitting, tax, contingency, developer overhead and profit. Do not average them; the scopes differ. Note also the embarrassment for forecasting: that laboratory's 2035 mid case of US$243 per kWh is already above the survey's 2025 actual United States figure. These are system-level numbers and they are the sibling brief's subject; they appear here only because the incomparability is the point.

“Northvolt was a recycling failure, and it went bankrupt in 2024.” Established Northvolt was a cell manufacturer; Northvolt Revolt was a recycling subsidiary. Its collapse is a manufacturing-cost and production-ramp story. The chronology, which most secondary sources get wrong, is two-stage: United States Chapter 11 in November 2024, then Swedish bankruptcy on 12 March 2025, filed by five named Swedish entities, with the German and North American subsidiaries not filing in their own jurisdictions. Sources saying it filed for bankruptcy in 2024, and sources saying it filed Chapter 11 in November 2024 after raising around fifteen billion dollars, are each half right. There is also a third act: Lyten completed the acquisition of Northvolt Ett, Ett Expansion and Northvolt Labs on 27 February 2026 for nearly five billion dollars, four months later than targeted, and will restart Skellefteå in the second half of 2026 making NMC lithium-ion — not its own lithium–sulfur. A lithium–sulfur company taking over a bankrupt NMC gigafactory in order to make NMC is itself a datum about which technology is bankable.

“Cuberg was Northvolt's solid-state division.” Established It was not, and the error appears in otherwise careful reporting. Cuberg's cells were lithium-metal with a liquid electrolyte. The distinction matters here more than it would elsewhere, because the whole argument of this brief turns on which announcements are describing an all-solid-state system and which are describing something with a liquid or polymer phase still in it.

“Solid-state batteries have reached 600 Wh/kg.” Handwave A claimed 600 Wh/kg cell with 1,500 km of CLTC range, going into rental and ride-hailing fleets in 2026, describes an in-situ polymerised — that is, semi-solid — system. It would be roughly 70% above the most aggressive announced laboratory figure from the world's largest cell maker and roughly double any shipping cell. No mechanism is published and no third-party measurement exists.

“Semi-solid, quasi-solid and condensed-matter batteries are kinds of solid-state battery.” Established They are marketing categories with specific and different technical contents, and keeping them apart is most of what it takes to read this field accurately:

“Mercedes has driven 1,200 km on a solid-state battery.” Established A lightly modified EQS covered 1,205 km from Stuttgart to Malmö in September 2025 on a Factorial FEST cell, which is a quasi-solid polymer electrolyte with an ultrathin lithium-metal anode — explicitly not all-solid-state. The pack carried roughly 25% more energy than the standard pack at similar mass and volume, which is the real and creditable result. It also requires pneumatic actuators to accommodate cell volume change during cycling, and the average speed of the run was not disclosed. Production is stated as by the end of the decade.

“Dendrites have been solved.” Frontier They have been explained, which is a different achievement and a real one: penetration of a ceramic electrolyte is stress-driven fracture at the dendrite tip, deflecting along grain boundaries in mixed-mode fracture. Laboratory cells now cycle at practically relevant areal capacity under practically relevant pressure. What has not happened is transfer to multi-Ah, multi-layer, low-pressure cells at automotive lifetimes — and the metric by which suppression is usually claimed has no agreed measurement protocol, so a share of the published progress may be measuring its own methodology.

“Recycling recovers X% of battery metals.” Established The most recent authoritative figure is from 2023, not from 2025 or 2026: recovered volumes reached over 40% of available feedstock for nickel and cobalt and 20% for lithium, and the market value of recycled battery metals grew nearly elevenfold between 2015 and 2023. No credible 2025 or 2026 global recovery-rate figure could be found for this brief and none should be accepted without a source. The genuine story of 2024 to 2026 is not the rate at all — it is that capacity is being built against feedstock that does not exist, and that Eramet, BASF at Tarragona, Umicore, Northvolt Revolt and Li-Cycle have all closed, paused or gone insolvent in consequence.

“QuantumScape has produced a breakthrough energy density.” Frontier The QSE-5 is a genuine engineering achievement and its volumetric figure of 844 Wh/L at cell level is its strongest number. Its gravimetric figure of 301 Wh/kg is roughly 20% above the top of the commercial liquid-electrolyte distribution measured for 2018 and below the 350 Wh/kg cell the incumbent announced in April 2026. Fifteen years and more than three billion dollars produced, on the metric the company exists to improve, about four years of incremental lithium-ion's historical trend rate.