Critical materials · Batteries

Battery technologies: the comparison table

In 2025 LFP took over half of EV batteries and more than 90% of grid storage, and averaged over 40% cheaper per kWh than NMC[2]. The table below puts every chemistry that matters on one grid: energy density, cycle life with the conditions attached, system efficiency and installed cost at the 100 MW reference point, the minerals each one draws, and a technology-readiness level gated on published evidence rather than press-release confidence.

How to read this page: measured sourced data · inferred analyst reading, basis linked · projected anchored to a real starting point · reported an outlet's claim, collected not verified. Bracketed citations link to the sources at the foot of the page.

Acronyms and units (39), open this first

Chemistry codes, engineering metrics and the JRC table's own column shorthand, all expanded. Nothing on this page assumes you carry these.

Cathode and cell chemistries

LFP · Lithium iron phosphate
The cheap, safe, long-lived workhorse. Lower energy density, no nickel or cobalt.
LMFP · Lithium manganese iron phosphate
LFP with manganese swapped in for some iron, which raises the voltage and the energy density.
NMC · Lithium nickel manganese cobalt oxide
The high-energy family. The digits give the metal ratio: NMC 811 is 8 parts nickel to 1 manganese to 1 cobalt.
NMCA · Lithium nickel manganese cobalt aluminium oxide
NMC with an aluminium dopant for stability.
NCA · Lithium nickel cobalt aluminium oxide
High nickel, no manganese. Energy-dense and cobalt-bearing.
LMR · Lithium manganese rich
Manganese-heavy cathode running at high voltage to reach energy density without much nickel or cobalt.
LCO · Lithium cobalt oxide
The original consumer-electronics cathode. Highest cobalt content of any commercial chemistry.
LTO · Lithium titanate
An anode, not a cathode. Replaces graphite to buy very long cycle life and fast charging, at roughly a third of the energy density.
Na-ion · Sodium-ion
Sodium replaces lithium as the shuttle ion. Three cathode families compete: layered oxides, Prussian white and blue, and polyanions.
Li-S · Lithium-sulfur
Very high energy per kilogram, short cycle life. Used where weight rules and replacement is acceptable.
NaS · Sodium-sulfur
A grid battery that runs molten, at 300 °C or so. Long duration, decades of field service.
ZEBRA · Zero Emission Battery Research Activity, the sodium nickel chloride cell
The other high-temperature sodium battery. Named for the research programme that produced it.
SSB · Solid-state battery
The liquid electrolyte is replaced by a solid one: sulfide, oxide-ceramic or polymer. Promises energy density and safety, not yet at volume.
VRFB / RFB · Vanadium redox flow battery / redox flow battery
Energy lives in liquid electrolyte tanks, power in a separate stack, so the two are sized independently.
Ni-MH · Nickel metal hydride
The pre-lithium rechargeable chemistry. Still shipping in hybrids and consumer cells.

Performance and cost metrics

TRL · Technology readiness level, 1 to 9
1 is a published idea, 9 is shipping at volume in a priced product. Every TRL on this page names the evidence that puts it there.
Wh/kg, Wh/L · Watt-hours per kilogram, per litre
Energy by weight and by volume. A car cares about both; a grid site cares about neither much.
RTE · Round-trip efficiency
The share of energy you get back out. Lithium-ion runs 83 to 90%, iron-air around 40 to 50%.
DoD · Depth of discharge
How far a cycle drains the cell. A cycle count means nothing without it: 4,000 cycles at 80% is a different claim from 4,000 at 20%.
SoH / SoC · State of health / state of charge
SoH is capacity remaining against new, so 'cycles to 80% SoH' is the honest way to state life. SoC is how full it is right now.
C-rate · Charge or discharge rate relative to capacity
1C empties the cell in an hour, 5C in twelve minutes. High C-rates age cells faster.
LCOS · Levelised cost of storage
Lifetime cost divided by lifetime energy delivered. The only fair way to compare a four-hour lithium system with a hundred-hour iron one.
N:P ratio · Negative to positive capacity ratio
The design margin, usually about 1.1, that keeps the anode slightly oversized so lithium plates into the graphite rather than onto it.
CAM · Cathode active material
The compound that actually stores charge, as distinct from the foil, binder and casing around it.
TR · Thermal runaway
The self-feeding heating reaction behind battery fires. JRC's safety column mostly describes how readily a chemistry enters it.

Applications, in the JRC table's shorthand

EV · Electric vehicle
LMT · Light means of transport
E-bikes, e-scooters, e-mopeds. Its own category in the EU Battery Regulation.
SLI · Starting, lighting, ignition
The 12 V starter battery in a conventional vehicle.
BESS · Battery energy storage system
Grid-connected stationary storage, the non-vehicle half of the market.

Materials, in the JRC table's shorthand

NG · Natural graphite
The anode material. Its processing is the most concentrated step in the whole battery chain.
REE · Rare earth elements
Used in Ni-MH, not in lithium-ion cells.
CRM · Critical raw material
An EU designation for materials that are both economically important and supply-risky.
Brackets, as in (Mn) or (Ir) · JRC's notation for a material used in some variants of the chemistry, not all
Element symbols · Li lithium, Na sodium, Ni nickel, Co cobalt, Mn manganese, Fe iron, P phosphorus, Al aluminium, Cu copper, Ti titanium, V vanadium, Si silicon, Ir iridium

Standards named in the safety column

UL 9540A · US test method for fire propagation in battery energy storage systems
Passing it is about whether one cell's failure spreads, not whether cells fail.
UL 1973 · US safety standard for stationary and motive battery packs
SAE J2464 · Automotive standard for abuse testing of EV batteries
IEC 62660-3 · International safety-requirements standard for lithium-ion cells in electric road vehicles
HF · Hydrogen fluoride
A corrosive gas some cell fires emit. Its absence is a real safety claim, not marketing.

LFP share of grid storage, 2025

>90%[2]

and over half of EV batteries

LFP installed, 100 MW / 4 h, 2023

$335–410/kWh[1]

full system, not the cell

Sodium-ion cell, CATL Naxtra

175 Wh/kg[3]

mass production from Dec 2025, company figure

The matrix

Every numeric cell is a range with the source that bounds it, and the tier marker says whose number it is: measured comes from the PNNL grid-storage database [1], reported is a manufacturer's own figure. Cycle counts carry their conditions, because a cycle count without depth-of-discharge and rate is marketing. Stationary technologies show no Wh/kg: density is not the buying criterion for a machine that never moves.

DoD depth of discharge, how far each cycle drains the cell · SoH state of health, capacity left against new · RTE round-trip efficiency, the share of energy you get back · LCOS levelised cost of storage, lifetime cost over lifetime energy delivered · TRL technology readiness level, 1 a published idea to 9 shipping at volume. All terms in acronyms and units.

ChemistryCell Wh/kgCycle lifeSystem RTEInstalled $/kWh (2023, 4 h)LCOS $/kWhMinerals drawnTRL (ours)TRL (JRC)Sold into
Lithium-ion
LFPLithium iron phosphate140–210 reported[5]2400–4550 measured[1]grid system at 80% DoD (PNNL 2023 estimate band)83% [1]335–410 [1]0.16–0.20 [1]lithiumironphosphorusgraphitecopper9 inferredOver half of EV batteries and over 90% of grid storage shipped globally in 2025 (2026)[2]9agrees, as "LFP"EVs, grid storage (the default chemistry of both)
NMC 811Lithium nickel manganese cobalt oxide (8:1:1)150–300 measured[13]1520–4805 measured[1]grid system at 80% DoD (PNNL 2023 estimate band)83% [1]372–455 [1]0.18–0.26 [1]lithiumnickelmanganesecobaltgraphitecopper9 inferredMulti-hundred-GWh/yr shipped by multiple manufacturers into priced vehicles (2026)[2]9agrees, as "NMC"Long-range EVs, premium segments
NCALithium nickel cobalt aluminium oxide322 measured[13]lithiumnickelcobaltaluminiumgraphitecopper9 inferredShipped at scale in production EVs for over a decade (2026)[2]9agrees, as "NCA"Long-range EVs (one automaker's ecosystem, mostly)
LTOLithium titanate anode84–96 reported[6]8000–20000 reported[6]5C/5C, 10-90% SoC, 25 °C (Toshiba figures for the 20 Ah-HP cell; upper end for the standard cell family)lithiumtitaniumnickelcobaltcopper9 inferredSCiB in commercial buses, trains and industrial fleets since 2008 (2026)[6]not mappedBuses, rail, port equipment, anywhere cycle count beats weight
Sodium
Sodium-ionSodium-ion (layered oxide)160–175 reported[3]10000 reported[3]company claim at launch, conditions not publishedsodiumironmanganesehard carbonaluminium8 inferredNaxtra mass production from Dec 2025; first mass-production sodium-ion passenger vehicle announced with Changan (2025-12)[3,4]9 / 8diverges, JRC row "Layered oxides"Entry EVs, grid storage, cold-climate and heavy-duty starter packs
Flow & long duration
Vanadium flowVanadium redox flown/a, stationary5250 measured[1]grid system at 80% DoD (PNNL 2023 point estimate)65% [1]585–715 [1]0.24–0.28 [1]vanadium9 inferredMulti-hundred-MWh commercial systems operating; PNNL models it as a standard 100 MW procurement (2026)[1]9 / 7agrees, as "VRFB"6-12 h grid storage; energy and power sized independently
ZincZinc-based (alkaline and zinc-bromine families)n/a, stationary2000–4563 measured[1]grid system at 80% DoD (PNNL 2023 estimate band, 10 MW reference)76–62% [1]369–621 [1]0.19–0.45 [1]zincbromine8 inferredCommercial deliveries at MW scale; PNNL's largest modeled system is 10 MW, an order below the lithium reference (2026)[1]not mappedBehind-the-meter and distribution-scale storage
Iron-airIron-air (reversible rusting)n/a, stationaryiron7 inferredForm Energy validated >100 h continuous discharge at nameplate power in a fielded system; first commercial sites commissioning (2025)[10]8 / 4diverges, JRC row "Fe-air"Multi-day (100 h) storage
High temperature
NaSSodium-sulfur (high temperature)n/a, stationary4500 reported[9]full-discharge cycles (NGK figure); 15-year calendar lifesodiumsulfur9 inferredNGK NAS systems in commercial grid service since 2002, GWh-scale fleet (2026)[9]9 / 8agrees, as "Na-S, Operating at 120-360 °C"6 h+ grid storage, operating at ~300 °C
Legacy
Lead-acidLead-acidn/a, stationary2368 measured[1]grid system at 68% DoD (PNNL 2023 point estimate)77% [1]399–463 [1]0.27–0.30 [1]lead9 inferred160 years in production; the recycling loop the EU Battery Regulation treats as the benchmark (2026)[1]not mappedSLI, UPS, telecom backup
Pre-commercial
Solid-state (Li-metal)Solid-state lithium-metal (ceramic separator)301 reported[7]lithiumnickelmanganesecobaltzirconium6 inferredB1 samples shipped to automotive customers Oct 2025; Eagle pilot line inaugurated 4 Feb 2026. TRL 7 requires a C-sample in a real vehicle (2026-02)[8,14]8 / 6diverges, JRC row "Li-ceramic-NMC"None yet sold; automotive target

System RTE, installed cost and LCOS are PNNL's 100 MW reference systems (zinc: 10 MW, its largest modeled size), 2023 estimates, low to high band [1]. Cell-level and system-level numbers answer different questions; this table keeps them in separate columns rather than blending them. Per-chemistry cell prices exist as a commercial series but its licence does not allow republication [12].

What a kWh is made of

Every chemistry above draws a different basket of minerals, and the basket is computable rather than quotable. The bars below are derived from each cathode's formula, its practical specific capacity and the cell voltage, with the anode sized to match the charge the cathode delivers inferred. The arithmetic is in scripts/build_mineral_intensity.js and every assumption it makes is listed under the chart.

Active materials only. Current collectors (the copper and aluminium foil every cell needs), electrolyte, separator, binder and housing are excluded, because none of them follow from stoichiometry and assuming them would put an unsourced number inside a checkable one. Bars are the midpoint of each derived range; hover for the range. Red and amber mark the materials the EU Battery Regulation names for due diligence.

LFP carries 0.95–1.04 kg of graphite per kWh against 0.09 kg of lithium, and no nickel, cobalt or manganese at all. Its iron and phosphorus appear on no concentration list. The chemistry that took over 90% of grid storage keeps almost all of its critical-materials exposure in one place, the anode.

NMC 811 needs 0.62–0.69 kg of nickel per kWh and 0.08–0.09 kg of cobalt, and still needs its graphite. Going high-nickel cuts cobalt mass without removing a single mineral from the list. Sodium-ion is the row that removes minerals: no lithium, no cobalt, and hard carbon in place of graphite, which takes the chain's most concentrated processing step off the bill entirely.

Where each mineral's story lives

Supply-concentration figures below render from the same shared module that /data/critical-materials/ renders in full, reference year 2023-2024. This page states the intensity; that page owns the geography.

GraphiteBattery Regulation due diligence

The anode in every commercial lithium-ion cell. It is the single largest mineral mass in the cell.

  • Processing: China 95%

The most concentrated processing step in the chain, and the one an LFP cell cannot avoid.

In: LFP, LMFP, NMC 622, NMC 811, NCA

Hard carbon

The sodium-ion anode. Made from pitch, biomass or polymer precursors rather than mined.

Not on the concentration list

Not graphite, and not on any critical raw materials list. Sodium-ion changes the anode question as much as the cathode one.

In: Sodium-ion (layered oxide)

Titanium

The LTO anode. Buys cycle life, costs energy density, and adds mass.

Not on the concentration list

In: LTO anode (NMC 622 cathode)

NickelBattery Regulation due diligence

The energy carrier in high-nickel cathodes. More nickel, more range, more exposure.

  • Refining: China 70%

Class-1 nickel is the battery-grade cut, a smaller market than the nickel headline price suggests.

In: NMC 622, NMC 811, NCA, Sodium-ion (layered oxide), LTO anode (NMC 622 cathode)

CobaltBattery Regulation due diligence

Structural stabiliser in NMC and NCA. The material every chemistry roadmap is trying to design out.

  • Refining: China 75%
  • Mining: DRC 70%

In: NMC 622, NMC 811, NCA, LTO anode (NMC 622 cathode)

Manganese

Cheap structural element in NMC, LMFP and sodium layered oxides.

  • Processing: China 95%

Abundant as an ore and highly concentrated as a battery-grade product. The two facts are often confused.

In: LMFP, NMC 622, NMC 811, Sodium-ion (layered oxide), LTO anode (NMC 622 cathode)

Iron

The cathode metal in LFP and LMFP, and in sodium layered oxides.

Not on the concentration list

Not a critical raw material by any list. This is what LFP bought when it gave up energy density.

In: LFP, LMFP, Sodium-ion (layered oxide)

Phosphorus

The phosphate in LFP and LMFP, supplied as purified phosphoric acid.

Not on the concentration list

Phosphate rock is not scarce; battery-grade purification capacity is a narrower question than the ore.

In: LFP, LMFP

LithiumBattery Regulation due diligence

Shuttle ion. Every Li-ion cathode carries it, and the anode holds it when charged.

  • Processing: China 65%

Finland's Keliber project is the EU's own lithium hydroxide bet.

In: LFP, LMFP, NMC 622, NMC 811, NCA, LTO anode (NMC 622 cathode)

Sodium

Shuttle ion in sodium-ion cells. Produced from soda ash and salt, neither scarce nor concentrated.

Not on the concentration list

Not a critical raw material, and that is the entire commercial argument for the chemistry.

In: Sodium-ion (layered oxide)

Aluminium

Dopant in NCA. Also the cathode current collector in every cell, which this intensity does not count.

Not on the concentration list

In: NCA

Assumptions behind the derivation, and how it checks out

Practical specific capacity is this file's central assumption, given as a range per chemistry rather than a point value. It is not a citation. Voltages come from JRC Table 2 (JRC145289, CC BY 4.0) where it publishes a nominal voltage for the same chemistry. The check below divides active mass per kWh by the cell mass implied by JRC's published Wh/kg for the same chemistry. A real cell is roughly half active material by mass, so a share in that region means the inputs are sane, and a share above 100% would mean they are not.

ChemistryCathodeCapacity (mAh/g)Voltage (V)AnodeActive kg/kWhActive share of cell mass
LFPLiFePO41451603.2Natural or synthetic graphite2.9083.1973577%
LMFPLiMn0.6Fe0.4PO41401553.5Natural or synthetic graphite2.7162.9935772%
NMC 622LiNi0.6Mn0.2Co0.2O21651803.7Natural or synthetic graphite2.3272.5393576%
NMC 811LiNi0.8Mn0.1Co0.1O21902103.7Natural or synthetic graphite2.1132.3233270%
NCALiNi0.8Co0.15Al0.05O21902053.7Natural or synthetic graphite2.1442.3236975%
Sodium-ion (layered oxide)NaNi0.33Fe0.33Mn0.33O21201403.2Hard carbon3.3063.8326069%
LTO anode (NMC 622 cathode)LiNi0.6Mn0.2Co0.2O2 with Li4Ti5O12 anode1651802.3Lithium titanate5.0245.488no JRC density to check against

Excluded: Current collectors (copper, aluminium foil), electrolyte salt and solvent, separator, binder, tabs, cell can, module and pack hardware. These do not follow from stoichiometry and are design choices, so they are left out rather than assumed.

What the materials cost, and what the cell costs

Multiply the kilograms above by what those materials sell for and you get a floor: the least a kWh of that chemistry can cost while metals trade where they do inferred. Set it against what JRC says a cell of that chemistry actually cost in 2025 [13] and the gap is everything else: the foils, electrolyte and separator this floor does not count, the conversion of ore-grade metal into battery-grade compound, manufacturing, yield loss, capex and margin.

Mineral prices are USGS Mineral Commodity Summaries 2026 (Public domain (US Government work)), 2025 figures, fetched 2026-08-05. Cell costs are JRC's 2025 EUR/kWh converted at the ECB reference rate of 1.1554 USD per EUR on 2026-08-05. Bars show midpoints; hover for the ranges, the largest cost leg and the share of active mass actually priced.

LFP's priced active materials come to $5.175.69 per kWh, about 8% of what JRC says an LFP cell cost in 2025. NMC 811 sits at 21% and NMC 622 at 26%. A nickel cell is two and a half to three times as exposed to metal prices as an LFP one, and an LFP cell's price is mostly a manufacturing number rather than a mining number.

That asymmetry decides who a price spike hurts. The 2022 lithium spike raised every chemistry's floor, but the chemistries whose floor is a quarter of their price felt it as a cost crisis, and the ones at a twelfth felt it as an irritation. It also says where each chemistry's remaining cost reductions have to come from: LFP's from scale and yield, high-nickel's from the metals or from designing them out.

ChemistryActive-material floor (USD/kWh)JRC 2025 cell cost (EUR/kWh)Floor as share of cellLargest cost legsPriced share of active mass
LFP5.175.69608%Lithium $4.12–4.54 · Graphite $0.95–1.04 · Iron $0.1–0.1182%unpriced: Phosphorus
LMFP4.985.590–955%Lithium $3.9–4.32 · Graphite $0.87–0.95 · Manganese $0.17–0.1982%unpriced: Phosphorus
NMC 62222.6924.7575–85JRC prices NMC as one family, not by grade, so 622 and 811 share this cost band.26%Cobalt $8.45–9.22 · Nickel $8.18–8.93 · Lithium $5.15–5.62100%
NMC 81118.1920.0975–85JRC prices NMC as one family, not by grade, so 622 and 811 share this cost band.21%Nickel $9.32–10.3 · Lithium $4.4–4.86 · Cobalt $3.61–3.99100%
NCA20.7422.39100–12017%Nickel $9.66–10.43 · Cobalt $5.62–6.06 · Lithium $4.56–4.92100%
Sodium-ion (layered oxide)6.27.23no matching JRC costNickel $5.82–6.79 · Manganese $0.17–0.2 · Sodium $0.16–0.1960%unpriced: Hard carbon
LTO anode (NMC 622 cathode)42.7246.63no matching JRC costLithium $15.84–17.31 · Cobalt $13.6–14.83 · Nickel $13.16–14.3641%unpriced: Titanium
Why this floor is a floor under a floor
  • Active materials only. Copper and aluminium foil, electrolyte, separator, binder and housing are not counted, because they do not follow from stoichiometry.
  • Graphite: Flake, not the spherical or synthetic anode grade a cell actually uses. Anode-grade costs several times this, so the graphite leg of the floor is a lower bound on a lower bound.
  • Nickel: LME cash is the class-2-inclusive benchmark. Battery cells need class-1 nickel sulfate, which trades at a premium, so this understates the nickel leg.
  • Phosphorus is not priced at all. A cell needs purified phosphoric acid, not phosphate rock. The rock price would understate it by an unknown multiple, so it is left out rather than guessed.
  • Ti is not priced at all. LTO needs titanium dioxide or tetrachloride. MCS quotes titanium sponge metal, a different product at a different price.
  • Hard carbon is not priced at all. A manufactured material from pitch, biomass or polymer precursors. There is no commodity quote for it.
  • Sodium-ion carries a floor and no comparison: the floor is computed for a layered oxide, and the only sodium cell cost JRC publishes sits on its Prussian white row, a different cathode. Borrowing it would compare a floor against the price of something else.

Cost per kWh cycled, as duration grows

Installed cost per kWh flatters whatever is cheapest to build. Levelised cost of storage divides lifetime cost by lifetime energy delivered, which is the only fair way to set a four-hour lithium system against a day-long one [1].

PNNL point estimates at 100 MW (zinc 10 MW), 2023 vintage where modelled and 2021 for gravity and hydrogen [1]. PNNL does not model Sodium-ion, Sodium-sulfur (NaS), Iron-air, Zinc-bromine flow, so they are absent here rather than estimated into the chart. 10 technologies shown.

Lithium-ion LFP costs $0.18 per kWh cycled at four hours, $0.17 at ten and $0.31 at twenty-four [1]. Pumped hydro runs the other way, $0.24 at four hours and $0.12 at twenty-four, because its extra hours are a larger reservoir rather than more cells. The two cross between four and ten hours, well short of the multi-day durations the argument usually invokes. Compressed air sits below lithium at every duration PNNL models, where the geology allows it at all.

What a battery does for a power system

Fingrid buys four frequency products and the market buys a fifth thing, energy. Four of the five ask a battery to hold output for seconds or minutes, which is why lithium won them so easily. The fifth is the one that runs into the wall measured further down.

ProductWhat it buysActivationHow long output must holdFit for a battery
FFR, fast frequency reserve[16]Sub-second injection when the grid has too little inertia to slow a frequency fall on its ownFull activation in at most 1.3 s at 49.7 Hz, 1.0 s at 49.6 Hz, 0.7 s at 49.5 HzMinimum 5 s of activation, or 30 s if power is withdrawn faster than 20% of capacity per second; must be able to reactivate within 15 minutesexcellentSeconds of energy and instant response. This is the product batteries were made for, and it exists precisely because inertia is falling.
FCR-D, disturbance reserve[17,19]Arresting a frequency excursion outside the normal band, holding 49.5 to 50.5 Hz50% of capacity within 5 s and 100% within 30 s on a stepwise fall from 49.9 to 49.5 HzContinuous full activation for 20 minutes, cut from 30 minutes in August 2022excellentFingrid shortened the endurance requirement from 30 to 20 minutes in a change aimed explicitly at energy storage.
FCR-N, normal operation reserve[17]Continuous two-directional regulation to hold frequency in the 49.9 to 50.1 Hz bandContinuous response within the normal bandContinuous full activation capability for 30 minutesgoodTwo-directional, so it charges and discharges all day. Cycling is the cost, and cycle life is the spec that decides the business case.
aFRR, automatic restoration reserve[18]Returning frequency to 50 Hz after containment has arrested the fall, on a signal from the operatorFull activation within 5 minutes, on activation signals sent every 4 seconds; minimum bid 1 MWMinutes, until mFRR or the market takes overgood
Energy arbitrageBuying cheap hours and selling expensive ones, which is the only service on this list that is about energy rather than frequencyMarket schedule, typically hourly or sub-hourlyThe full rated duration, most often 2 to 4 hoursconditionalPays only when the daily price spread beats the round-trip loss and the degradation. A flat price day earns nothing.

The one thing it cannot sell

Fingrid defines inertia as "the ability of the kinetic energy stored in the rotating masses in the electricity system to resist changes in frequency" [16]. A battery has no rotating mass and stores no kinetic energy, so it supplies none of it inferred. What it supplies instead is a response fast enough to substitute for part of what inertia buys, which is why FFR exists as its own product and why its requirement is written in fractions of a second: full activation in 1.3 seconds at 49.7 Hz and 0.7 seconds at 49.5 Hz [16].

The distinction gets blurred constantly, usually as "synthetic inertia". It matters because the two are bought for different reasons: inertia sets how fast frequency falls in the first seconds, and fast reserve sets how quickly something arrests the fall. Fingrid's own FFR page says the need for FFR is set by inertia levels alongside activation speed and disturbance size, so the products are complements, not substitutes.

The duration wall, measured

"Batteries cannot cover a windless week" is said constantly and almost never with a number. Fingrid publishes hourly wind output under CC BY 4.0 [20], so here is the number, read straight out of the measured record measured.

The five longest runs of consecutive hours with Finnish wind below 747.8 MW, which is 10% of the highest wind hour in the window (7477.8 MW). Window 2026-01-01 to 2026-08-05, 5,208 hours. The window spans 8 months (2026-01 to 2026-08), not a full year. A window missing a winter understates the wall.

Finnish wind ran below 747.8 MW for 154 consecutive hours from 2026-01-18, which is 6.4 days [20]. Holding wind's contribution flat across that lull would have taken 313 GWh of stored energy, about 17% of everything Finland consumed in the same period. Lulls of six hours or more accounted for 1,210 of the window's 5,208 hours.

At the cheapest installed lithium cost on this page, PNNL's 2023 four-hour LFP system, 313 GWh of energy capacity runs to $105.1128.4 billion [1] projected. The number is deliberately the wrong product: nobody builds a four-hour system to hold six days, and the LCOS curve above shows what happens to lithium's economics as duration stretches. That is the argument for iron-air and pumped hydro stated in Finnish hours rather than in adjectives, and it is why the hundred-hour design duration on the iron-air row is aimed at exactly this gap.

Two honest limits on the figure. The window covers 8 months rather than a full year, so a longer record would probably find a worse lull, not a better one. And Finland does not actually ride out lulls on storage: it imports, and it runs hydro, nuclear and combined heat and power, all of which sit in the same Fingrid file. This measures the storage-only counterfactual because that is the claim the battery debate makes, not because it is how the system runs.

What goes wrong, and what the rules reach

Five problems, each stated as a fact before it is stated as a worry. Where the fact already sits somewhere on this page, it is referenced rather than repeated.

Thermal runaway is a chemistry property, not a battery property

measured

JRC's safety column in the table above separates the families rather than treating fire as one risk: LFP has a relatively mild thermal runaway, LMFP the lowest risk of any lithium chemistry, and the sulfide and ceramic solid-state routes are described as thermally stable. The standards named there, UL 9540A and SAE J2464, test whether one cell's failure propagates, not whether cells fail.[13]

For a grid site the design question is containment and spacing, not whether the chemistry can burn. For a chemistry choice it is one more reason LFP took the stationary market.

A cycle count is a claim about conditions

measured

Every cycle-life figure in the comparison table carries its depth of discharge and, where the manufacturer gave one, its rate. PNNL's grid systems are quoted at 80% DoD and Toshiba's LTO cell at 5C into 10-90% state of charge. Calendar life runs alongside it: PNNL puts a lithium system at 13 to 16 years whatever its cycling.[1,6]

Two chemistries with the same headline cycle count can differ by a factor on the same duty. The number without its conditions is marketing, which is why this page refuses to print one.

The recycled-content rules name four metals, and graphite is not one

measured

From 18 August 2031 a new EV battery must contain at least 16% recycled cobalt, 85% recycled lead, 6% recycled lithium and 6% recycled nickel, rising to 26, 85, 12 and 15% in 2036. The due-diligence articles of the same regulation name cobalt, lithium, natural graphite and nickel, so graphite is inside the law but outside the recycled-content list.[21]

Graphite is the largest single mineral mass in a lithium cell and the most concentrated processing step in the chain. The recycled-content instrument does not reach it.

Recycled content needs dead batteries, and the fleet is young

inferred

Recovery targets bind earlier than content targets: 50% of lithium recovered from waste batteries by the end of 2027 and 80% by 2031, and 90% rising to 95% for cobalt, copper, lead and nickel. Collection targets run alongside, 63% of portable batteries by 2027 and 73% by 2030.[21]

A recovery rate is a share of what arrives, not a quantity. Batteries sold into a growing fleet do not return for a decade or more, so a high recovery rate on a small return stream still yields little secondary metal against fast-growing demand. That is the gap between recycling working and recycled content being available.

The anode is the exposed step, and it has already been used

measured

China required export permits for graphite items from 1 December 2023, covering high-purity synthetic graphite and natural flake including the spherical and expanded grades that go into anodes. CSIS puts China's share of global graphite refining above 90%, and the concentration panel above reads 95% for anode-grade processing.[22,23]

Every lithium chemistry on this page draws roughly a kilogram of graphite per kWh and cannot avoid it. Sodium-ion is the only row that does, because hard carbon is manufactured rather than mined.

Recycled-content minimums, and the material they leave out

MaterialFrom 18 Aug 2031From 18 Aug 2036Intensity in a cell, this page's derivation
Cobalt16%26%NMC 811: 0.08–0.09 kg/kWh
Lead85%85%not in a lithium cell
Lithium6%12%NMC 811: 0.09–0.10 kg/kWh · LFP: 0.09 kg/kWh
Nickel6%15%NMC 811: 0.62–0.69 kg/kWh
Graphitenamed for due diligence, absent from recycled contentno targetno targetLFP: 0.95–1.04 kg/kWh, the largest mineral mass in the cell

Percentages are Article 8 minimums for EV batteries [21]. Intensities are this page's derivation, active materials only. Setting the two columns side by side is the point: the instrument reaches the metals with the smallest mass in the cell and not the one with the largest.

The regulation is well aimed at cobalt, where 26% recycled content by 2036 meets a metal that a modern cell already uses sparingly, 0.08–0.09 kg/kWh in NMC 811 against 0.18–0.20 in NMC 622. It does not reach graphite at all, and graphite is where the concentration sits: 95% of anode-grade processing in one country, export permits since December 2023, and about a kilogram in every kWh of every lithium chemistry on this page [22] inferred.

Where JRC reads it differently

JRC puts three of the eight mapped chemistries higher than this page does: sodium-ion at 9 against 8, iron-air at 8 against 7, ceramic solid-state at 8 against 6 [13]. Its band records where the furthest-ahead developer has reached, and its own footnote counts a technology as commercialised once it sells in some market segment, earbuds and small electronics included. The rubric here asks for multi-GWh shipping into priced products before it writes a 9. Both columns stay in the table: where they diverge, a leader has arrived and volume has not.

JRC also prints a second number this page has no equivalent for, the followers level. Iron-air sits at 8 for the leader and 4 for everyone else, and zinc-air at 9 against 5, the two widest gaps in the table [13]. One company knowing how to build something is a different fact from an industry knowing how, and procurement lives on the second one.

What the table says

On PNNL's 2023 numbers a lead-acid grid system costs more per installed kWh than LFP [1]. The chemistry that held the grid market for a century now loses on price, the one dimension it was supposed to keep.

Sodium-ion entered mass production in December 2025 at 175 Wh/kg, inside first-generation LFP's density band, with no lithium, no cobalt and aluminium current collectors on both electrodes [3,4]. Its cost case rests on materials that do not spike; whether that discount materialises at GWh volumes is the open question the weekly watcher on this page will track.

Iron-air runs at a reported 40 to 50% round-trip efficiency against 85 to 90% for lithium-ion [11], and validated more than 100 hours of continuous discharge in a fielded system [10]. That pairing is not a defect sheet, it is a different product: a battery priced to sit full for weeks and cover the gap lithium cannot reach. The first project outside the US makes the point in its own rating: 10 MW against 1,000 MWh in northwest Ireland, due 2029, is the 100-hour duration written as a nameplate [15].

The full landscape: JRC Table 2

The Commission's Joint Research Centre compares 29 battery technologies in one grid in its 2025 battery status report, and publishes it under CC BY 4.0 [13]. It is reproduced below in full, verbatim, including the blanks. The TRL column is JRC's own two-speed reading: a level for technology leaders and, where it differs, a level for followers.

JRC states the limits of its own table, and they travel with the data: it compiles commercial figures, manufacturer announcements and research results into one grid, so cells are not uniformly comparable. Two rows show it without needing an argument. VRFB's energy density is quoted in Wh/L where every other row uses Wh/kg, and Fe-Fe carries two costs an order of magnitude apart. The bracketed numbers are JRC's own reference markers into its bibliography, kept so a reader can follow its trail rather than ours.

Reading its shorthand: EV electric vehicle · LMT light means of transport, e-bikes and e-scooters · SLI the 12 V starting, lighting and ignition battery · BESS grid-connected stationary storage · NG natural graphite · REE rare earth elements · TR thermal runaway · a material in brackets, as in (Mn), is used in some variants and not all. Full list in acronyms and units.

TechnologyTRL leaders / followersNominal voltage [V]Energy density [Wh/kg]Durability [cycles]Cost 2025 [EUR/kWh]SafetyCritical raw materialsApplicationRemarks
Li-ion
NMC93.6-3.7150-300 [11]1000-1500 [11]75 [12] 85 [13]Li, NG, Si, Ni, Mn, Co, Cu, AlEV, LMT
NMCA83.6-3.7200-330 [14]3000 [13]69 [15]Coatings may improve electrodes stability [16]Li, NG, Si, Ni, Mn, Co, Cu, AlEV, LMTEnergy density and cycle life improved by 20% [17], cost reduced by 8% relative to NMC [15]
LFP93.2120-240 [18]2000 [11], [19] 8000 [20]60 [21]Relatively mild thermal runaway [22]Li, NG, P, Cu, AlEV, LMT, BESSRD&I on 1 MW charging for EVs [23] energy density [24], cost reduction and safety [22]
LMFP83.5210-240 [25], [26], [27]3000 [25] 4000 [26]90-95 [28]Lowest risk of TR among Li-ion batteries [29]Li, NG, P, Mn, Cu, AlEV, LMT, BESSIntegral Power and Euro Manganese join forces for LMFP production in Czech Republic [30]
NCA93.6-3.7322 [31]3000 [31]100-120 [32]Li, NG, Ni, Co, Al, Cu, AlEV, LMT, aviationNCA chemistry investments will double in the aviation industry to 250 million EUR in 2034 [33]
LMR74.1-4.6+33% vs. LFPMore stable than NMCLi, NG, Mn, Ni, Cu, AlEVGM+LG plan to start production in 2028 [34]
NMC+LFP Mixed chemistry (dual layer cell)– / 63.73260 [35]3000 [35]SAE J2464 and IEC 62660-3 compliant cellsLi, NG, Ni, Mn, Co, P, Cu, AlEV, BESSOEMs also consider mixed chemistry battery packs (combining cells of different chemistries) for EVs [36], [37]
SSB
Si-sulfide-NMC– / 63.6283 [38], 390 [39]1000 [39]Thermally stable and robust [40]Li, NG, Si Ni, Mn, Co, Cu, AlHigh-end EVs, medicalA review on sulfide based solid electrolytes' properties and challenges [41]
Li-sulfide-NMC7 / 63.6440 [39]1000 [39]95 [42]Thermally stable and robust [40]Li, Ni, Mn, Co, Cu, AlHigh-end EVs, medicalHybrid electrolyte systems can improve compatibility at the sulfide electrolyte-lithium interphase. [43]
Li-ceramic-NMC8 / 63.6220 [44]6000-14000 [44]Non-flammable at 300 °C [45]Li, Ni, Mn, Co, Cu, AlHigh-end EVs, medicalNew FeCl3 cathode and dry 3D printing process might cut cost to 40-50% of current NMC [46]
Li-polymer-NMC9 / 63.7180-275 [47]300-500 [47]Cells pass all safety test [48]Li, Ni, Mn, Co, Cu, AlHigh-end EVs, hobby, medicalPassive fillers (e.g. LiAlO2) can improve ionic conductivity of polymer electrolyte [49], [50]
Quasi-solid, hybrid9 / 73.6350 [51], 400 [52]1200 [51]No smoke, fire, or explosion, 100% SOC nail penetr. tests [52]Li, NG, Si Ni, Mn, Co, Cu, AlHigh-end EVs, hobby, medical
Na-ion
Prussian white9 / 72.2-3.2180 [52]500085 [53]Solid byproducts of battery fire, less prone to emit HF [54](Mn), AlMid-range EVs, LMT, SLI
Prussian blue9 / 71.3-1.922 [55]50000 [55]AlBESSNatron Energy ceased operations [55]
Polyanions8 / 62.1-3.2175 [56]10000 [56]Intrinsically safe at material level [56]Ni, V, AlMid-range EVs, LMT, SLI90% usable power at -40°C [56]
Layered oxides9 / 82.1-3.2180 [52]3000 [52]Ni, Mn, Ti, AlMid-range EVs, SLI
NiMH
Ni-MH91.2140 [57]400-1000 [58] 1500 [57]150-200 [59]Aqueous solution of K4P2O7 reduces volatile compounds in case of fire [60]Ni, REEEVs, BESS, consumer electronicsExpected market growth of 40% by 2030, mostly for EVs and consumer electronics [61], competing with 1.5 V Li-ion batteries [62]
RFB
VRFB9 / 71.26 [63]20-40 Wh/L [63]quoted in Wh/L, not Wh/kg10000 - 20000 [64]260 (10 h duration) [65]Mn-V/V couple can reduce V content, improve safety [66]V, Ti, (Mn)BESS2 MWh system installed in the US with a warranty life of 30 years [67]
Zn-Br81.760-856000 20 y [68]380Ti, NGBESSCommercial modules of 25 kW available in US [68]
Fe-Fe– / 51.21 [69]40-60 [69]1831 [70]30 [70], 230-380 [71]NGBESSUnlimited cycles, 25 y warranty [72]
Me-air
Li-air9 / 73.2-41200 [73] 5000 [74]100-200 [74], 1000 [73]Li, (NG), (Ir)Medical, EVs, consumer electronicsButton cells commercially available [75]
Zn-air9 / 51.4-1.51200 [76]7200 [77]Targeted 20 [78], 30 [79](NG), (Ir)Medical, EVs, consumer electronicsREE-free air catalyst (FeCo-N-C dual atom) can improve performance and cycle life. [77] Button cells commercially available.
Fe-air8 / 40.6-1.4 [80]20-250 [81]1300 [82], Unlimited [83]Targeted 20 [84]Passed UL9540A safety testing; no flame or thermal event propagation [83](NG), (Ir)BESSLow RTE, slow charge / discharge. First battery connected to the grid in NL for testing [85]. A 500 MW / 50 GWh plant is under construction in the US [86]. A low TRL EU project aim for 250 Wh/kg and 60% RTE [82].
ZNB
Zn-Br (non-RFB)81.850-70 [87]6000 or 20 y [88]50UL 1973, UL9540A [88]NG, TiBESSA 15 MWh battery to be installed for a microgrid in California [89]
Zn-salts61.5-1.930-40 [90]4000-10000 [91]10-100Considered safeNGBESSThin films of 2D porous fluorinated framework improve cycle life. [92] A 5 kWh system installed in NL [93]
Li-S
Li-S8 / 62.5 -3.0400-500 [94]300-500, 800 [94]Projected 60-100 [94]Li, NG, Cu, Ti, AlDrones, military, hobby, (EV, BESS)Working range: -40 °C to 50 °C. [95] Effect of environment on Li-S batteries yet done at cell level only. [96] Cycle life of 25000 reported for Li-S SSB. [97]
High-temperature batteries
Li-S, Operating at 150 °C82.5 -3.0450Li, NG, Cu, Ti, AlSpecial industrialSpace, boreholes drilling
Na-S, Operating at 120-360 °C9 / 81.6-2.2150-2404500-7000 [98]150-200Considered safe, often UL 1973 or UL9540A certifiedAlBESS
Na-NiCl (ZEBRA), Operating at 270-350 °C9 / 82.6120-140 [99]450085 [99]Considered safe, potential risks from molten sodium and sulfur if damagedNi, AlBESS

Clean Energy Technology Observatory: Battery Technology in the European Union, 2025 Status Report on Technology Development, Trends, Value Chains and Markets, JRC145289, Table 2, printed pages 11-13. Licence CC BY 4.0 (Commission Decision 2011/833/EU). Critical raw materials are given at cell level, active and passive materials. Where a solid-state or metal-air row reaches TRL 9, in some market segments (e.g. earbuds, small electronics) the technology is commercialized. Changes made in reproduction: none to values; four cells had subscripts restored (K4P2O7, FeCl3, LiAlO2) or text completed across a page break, each flagged in the data module.

Technology watch

Two layers with a hard gate between them. The register is curated: every level is set by hand against the nine gates below and cites the dated evidence that clears it. The feed underneath is collected: a weekly watcher reads battery news, ranks it by how much maturity it evidences, and proposes moves. It never edits the register.

The nine gates, and what has to be evidenced to clear each

No standards body publishes TRL by battery chemistry, so every level here is this page's inference inferred. An inference nobody can argue with is worthless, so the gates state what must be publicly evidenced. Disagree with a specific gate rather than with a feeling.

TRLWhat must be publicly evidenced
1A chemistry proposed and published. Results are calculations or coin cells.
2Concept formulated with an application named. Still coin-cell scale.
3Experimental proof of the critical function, published, at coin-cell or small-pouch scale.
4Multilayer pouch or small cylindrical cell built in a lab. Specs are self-reported.
5A-sample cells delivered to a named external party. Pilot equipment ordered.
6Pilot line producing Ah-scale cells. B-sample out. Third-party test data exists.
7C-sample qualified on production-representative equipment, or a field trial in a real vehicle or a real grid site.
8Start of production on a commercial line. First commercial delivery at a stated price.
9Multi-GWh a year shipped to multiple customers inside priced products.

Three rules keep the register honest. TRL attaches to the technology and never to a company, so a manufacturer failing is a dated note and not a level change. Every level carries dated evidence and a link, or it is not published. A manufacturer's own spec is labelled as the manufacturer's claim.

The register, edited by hand

TechnologyTRLEvidence that clears the gateIn the comparison table
Lead-acid9160 years in production; the recycling loop the EU Battery Regulation treats as the benchmark (2026)[1]yes
LFP9Over half of EV batteries and over 90% of grid storage shipped globally in 2025 (2026)[2]yes
LTO9SCiB in commercial buses, trains and industrial fleets since 2008 (2026)[6]yes
NaS9NGK NAS systems in commercial grid service since 2002, GWh-scale fleet (2026)[9]yes
NCA9Shipped at scale in production EVs for over a decade (2026)[2]yes
NMC 8119Multi-hundred-GWh/yr shipped by multiple manufacturers into priced vehicles (2026)[2]yes
Vanadium flow9Multi-hundred-MWh commercial systems operating; PNNL models it as a standard 100 MW procurement (2026)[1]yes
Sodium-ion8Naxtra mass production from Dec 2025; first mass-production sodium-ion passenger vehicle announced with Changan (2025-12)[3,4]yes
Zinc8Commercial deliveries at MW scale; PNNL's largest modeled system is 10 MW, an order below the lithium reference (2026)[1]yes
Iron-air7Form Energy validated >100 h continuous discharge at nameplate power in a fielded system; first commercial sites commissioning (2025)[10]yes
Solid-state (Li-metal)6B1 samples shipped to automotive customers Oct 2025; Eagle pilot line inaugurated 4 Feb 2026. TRL 7 requires a C-sample in a real vehicle (2026-02)[8,14]yes
Sulfide all-solid-state6Idemitsu broke ground on a lithium sulfide electrolyte plant in Jan 2026 against a Toyota mass-production target around 2027-28. A plant under construction evidences a pilot-to-production programme, not a product; TRL 7 needs a C-sample in a real vehicle. (2026-01)[25,24]not yet, no sourced spec
Sodium-ion (Prussian blue)4JRC's own Table 2 row records that Natron Energy, the chemistry's most visible commercial developer, ceased operations. The chemistry's published energy density on that row is 22 Wh/kg, an order below the layered oxides. (2026)[26,13]Natron ceasing operations is recorded as a company event. It does not by itself lower the chemistry's level; the low level here rests on the published cell performance.not yet, no sourced spec

This week in battery technology

Collected weekly and ranked by the maturity class of the event rather than by how loud the headline is reported. Every claim below is the outlet's, not this site's, and none has been verified here.

Week to 2026-08-31. The watcher also logged 13 headlines that cleared a maturity class but matched nothing on the watchlist, which is how a chemistry the register does not yet name gets found, and 0 company events. A company failing never lowers a chemistry's level.

Named, watched, not yet tabulated

A comparison row costs at least one sourced spec and a dated TRL evidence line. These chemistries have headlines and promises so far; the weekly technology watcher reads their news, and each one gets its row when the evidence clears the bar.

LMFP: manganese-boosted LFP; several makers announce, few ship
LNMO: cobalt-free high-voltage spinel; pilot lines
Silicon-anode Li-ion: drop-in energy gain, swelling still the fight
Lithium-sulfur: aviation-weight promises, cycle life the wall
Sodium-ion (Prussian white): the second sodium cathode family
Semi-solid / polymer electrolyte: the halfway house already in some packs
Sulfide all-solid-state: the other solid-state route: Toyota-Idemitsu electrolyte plant building, Samsung SDI targets 2027 mass production
Sources and method (26)
  1. [1] PNNL, Energy Storage Cost and Performance Database v2024 (ESGC workbook; 100 MW reference systems, zinc 10 MW)
  2. [2] IEA, Global EV Outlook 2026, electric vehicle batteries chapter (CC BY 4.0)
  3. [3] CATL, Naxtra sodium-ion launch, Super Tech Day, 21 Apr 2025 (company announcement)
  4. [4] CATL, first mass-production sodium-ion passenger vehicle with Changan (company announcement)
  5. [5] electrive, first details of BYD's second-generation Blade battery, 9 Dec 2024 (reporting BYD figures)
  6. [6] electrive, Toshiba 20 Ah-HP SCiB (LTO) cell announcement, 12 Jan 2022 (reporting Toshiba figures)
  7. [7] QuantumScape, 'A first look at the QSE-5 B sample' (company blog; self-reported cell data)
  8. [8] electrive, QuantumScape delivers B1 samples of its solid-state cell, 24 Oct 2025
  9. [9] NGK Insulators, Sodium-Sulfur (NAS) Battery, presentation to Sandia LDES workshop, 2021
  10. [10] Form Energy, battery technology page (company; 100-hour iron-air)
  11. [11] Energy-Storage.news on Form Energy's iron-air efficiency trade-off (reported round-trip efficiency)
  12. [12] Our World in Data chart of Benchmark Mineral Intelligence cell prices by chemistry (cite-only; Benchmark's licence does not allow redistribution)
  13. [13] European Commission JRC, Clean Energy Technology Observatory: Battery Technology in the European Union, 2025 Status Report (JRC145289), Table 2, printed pp. 11-13. CC BY 4.0
  14. [14] QuantumScape, Eagle Line inauguration for solid-state pilot production, 4 Feb 2026 (company press release)
  15. [15] Form Energy and FuturEnergy Ireland, agreement for a 10 MW / 1,000 MWh iron-air project in northwest Ireland, operational target 2029 (company announcement)
  16. [16] Fingrid, FFR (Fast Frequency Reserve): product page, technical requirements and activation thresholds
  17. [17] Fingrid, FCR (Frequency Containment Reserves) product page: FCR-N and FCR-D frequency bands and activation
  18. [18] Fingrid, aFRR (automatic Frequency Restoration Reserve) product page: bid size and full activation time
  19. [19] Fingrid, Changes in the requirements of Frequency Containment Reserve provision from energy storage (2022): FCR-D continuous full activation cut from 30 to 20 minutes
  20. [20] Fingrid Open Data, hourly wind generation and load (CC BY 4.0), snapshot in static/data/finland-power/hourly/fingrid.json
  21. [21] European Commission / EUR-Lex, Sustainability rules for batteries and waste batteries: recycled-content minimums, collection targets, material recovery rates (Regulation (EU) 2023/1542)
  22. [22] IEA policy database, China: Announcement on the optimisation and adjustment of temporary export control measures for graphite items (export permits from 1 December 2023)
  23. [23] CSIS, China's New Graphite Restrictions: scope of the controls and China's share of global graphite refining
  24. [24] Toyota and Idemitsu, beginning of cooperation toward mass production of all-solid-state batteries for BEVs (company announcement)
  25. [25] electrive, Toyota partner Idemitsu breaks ground on its lithium sulfide plant for all-solid-state batteries, 30 Jan 2026
  26. [26] JRC CETO 2025 Status Report (Table 2, Prussian blue row): records that Natron Energy ceased operations

Grid-system figures are the PNNL ESGC v2024 workbook slice fetched by scripts/fetch_pnnl_storage.js (100 MW reference, zinc 10 MW); the snapshot records the fetch date. Manufacturer cell figures are the companies' own claims at the linked announcements, shown under the reported tier, never blended with measured data. TRL levels are this page's inference under the nine-gate rubric documented in the plan; each level cites the dated evidence that clears its gate. The Benchmark Mineral Intelligence per-chemistry price series is cited for existence only; its licence does not allow republication.