Report 089 · Energy Storage
The stock answer is "all of them, eventually." A 2026 study in Applied Energy built the cost model instead of assuming it, and the result runs backwards from intuition: the chemistry with the cheapest metals is the one worth reusing, and the chemistry stuffed with nickel and cobalt is the one worth shredding. Recyclers have to be paid to take the good ones.
Every few months a story runs about retired electric-vehicle batteries getting a second life on the grid. The framing is always the same and it is always appealing: the pack is only "worn out" by car standards, a house or a substation asks far less of it, so plug it into a rack and let it work another decade.
The appeal is real. The framing hides the actual question, which is not whether a used pack still holds charge. It is whether anyone can afford to test it.
A paper published in Applied Energy in April 2026, by a Carnegie Mellon group with co-authors at the National Laboratory of the Rockies and Argonne National Laboratory, is the first I have read that answers that properly. It builds an open-source, process-based cost model of a repurposing facility that actually complies with UL 1974, the safety standard for evaluating used batteries for second life, and then asks the only question that decides anything: what is the most a repurposer could pay for a used pack and still beat a new battery.
The authors call it the breakeven acquisition price, and their definition is worth reading slowly because the sign matters:
"A positive value indicates that a repurposer is able to pay up to the breakeven price for used EV battery packs, whereas a negative value indicates that a repurposer would have to be paid at least the breakeven price to take used EV packs."
So a positive number means the pack is an asset with a price. A negative number means the pack is garbage with a disposal fee attached. One model, one axis, and every retirement decision in the industry sits somewhere on it.
First the processing. For an 85 kWh Tesla Model S pack at a facility running 500 MWh per year, the study estimates repurposing costs of roughly $12/kWh at the pack level and $41/kWh at the module level, measured per kWh of original EV nameplate capacity, before any of the components and labor needed to assemble a finished storage system.
The gap between those two numbers is not disassembly labor, which is what I would have guessed. It is testing time, and the mechanism is almost funny once you see it. The cycling tests that dominate UL 1974 take about 20 hours per unit, and a unit is a unit: a pack takes about as long to test as a module does. A Model S pack contains 16 modules. So routing the same kilowatt-hours through the module line means running the twenty-hour test sixteen times instead of once, which means more cyclers in parallel and more labor to set them up. Pack-level repurposing lands at less than a third of module-level cost for that reason alone.
That is a very specific, very actionable finding, and it is sensitive to exactly what you would expect: pack size and modules per pack. The authors say so directly rather than letting the $12 travel unqualified.
Now the part that reorders the picture. The study compares repurposing against recycling for three cathode chemistries, using Argonne's EverBatt model for the recycling side, and the answers point in opposite directions.
For LFP, lithium iron phosphate, breakeven acquisition prices for repurposing are positive across every second-life application and every sensitivity case tested. Even in the harshest one, EV charge support, the most cycling-intensive job they modeled, a repurposer could pay $46 to $56 per kWh for the used pack. In most other applications LFP second life did not end because the battery wore out. It ended because the authors capped the simulation at 30 years across both lives, and the battery was still going.
Meanwhile, recycling LFP produces negative breakeven prices no matter which process you use, because the recovered minerals are worth less than the cost of recovering them. Recyclers must be paid to take LFP packs.
For NCA, the nickel-cobalt-aluminum chemistry, it flips. No second-life application was consistently competitive with recycling. NCA degrades faster and its metals are worth more, so the shredder wins. And the authors note that if new cell prices keep falling, repurposing NCA will likely stop being competitive in any application at all.
For NMC622, it depends, which here is a real answer rather than a hedge. Only peak shaving beat recycling across all first-life scenarios. Everything else turned on how the pack was used in the car and how hard the second life would work it. That is precisely the profile of a battery you should test individually rather than route by policy, which is why their recommendation for NMC is sorting.
The authors compress the whole thing into one sentence that I think is the most portable idea in the paper:
"for a given battery chemistry, the value of recycling strongly depends on the value of materials that can be recovered and the value of repurposing strongly depends on the durability of the battery."
Two different properties, pulling in opposite directions, and no chemistry is good at both. Cheap durable cells are worth reusing and worthless to shred. Expensive fragile cells are worth shredding and a bad bet to reuse. The intuition that a more valuable battery is more worth saving has the logic exactly inverted.
Because LFP already won the grid. I wrote about why that happened: cheap, durable, thermally better behaved, and content to be big and heavy in a place where weight does not matter. This paper says the same properties that made LFP the default for new stationary storage also make it the only chemistry whose used packs are unambiguously worth buying.
That is a convergence with a real consequence. The second-life supply that is actually economic is LFP, and the market it would feed is a market already dominated by new LFP. So second-life storage is not competing with the batteries it replaces. It is competing with new cells of the same chemistry, at $96.3/kWh in the study's baseline, on a life-adjusted basis. That is a much harder fight than the usual framing admits, and it is the fight the breakeven price is measuring.
The authors also flag the saturation risk plainly: if repurposed systems flood the market, prices for both new and used batteries fall, which could push nickel-based chemistries to negative breakeven prices across most applications. The economics of second life are not a fixed property of the technology. They move with the price of the thing they compete against.
The limitations section is unusually candid and I am going to reproduce its shape rather than skip it, because the headline numbers are only as good as their fences.
Note that this is a peer-reviewed paper, open access under CC BY, not a preprint or a vendor study. The recycling side comes from Argonne's EverBatt model rather than from the authors' own measurements, and I did not open EverBatt myself, so treat those figures as the paper reports them.
This is the layer I work in. I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator; I do not own the company and earn nothing from this link, and I flag it because cycling and testing are what I actually build rather than only write about. Full policy here.
The finding that changes my thinking is the twenty-hour test. Testing cost per kilowatt-hour is set by how many units you touch, not how much energy you move. That makes throughput a diagnostics problem, not a labor problem. Anything that shortens the qualification cycle, better onboard history, better state-of-health estimation from partial cycles, better screening before a pack ever reaches a cycler, attacks the single largest line item in the whole model. The paper even points at the gap: the older studies it corrects were built on testing protocols that predate UL 1974 entirely, with test times ranging from 8 to 40 hours and no standard behind them.
The second thing I would take to a procurement meeting: ask which chemistry the pack is before asking anything else about its second life. If the answer is NCA, the honest response is that the shredder probably outbids you. If it is LFP, the pack has a positive price and the conversation is about who captures it. If it is NMC, you are in the sorting business, which means your testing cost is your business model.
"EV batteries get a second life" is not wrong, it is just underspecified to the point of being useless for a decision. The 2026 model puts a price on the specifics: about $12/kWh to qualify a pack under UL 1974, about $41/kWh if you break it into modules first, and a breakeven acquisition price that is comfortably positive for LFP, mostly negative for NCA, and genuinely case-by-case for NMC.
The clean version is the authors' own: reuse LFP, recycle NCA, sort NMC. And the reason underneath it is the part worth carrying, because it will outlive these particular numbers. Recycling pays for materials. Repurposing pays for durability. The battery that is cheapest to build out of is the one nobody wants to tear apart.
Scope note: all cost figures are modeled outputs, not observed prices from operating facilities, and they describe one modeled pack type at one modeled facility scale. Breakeven acquisition price is a ceiling on what a repurposer could pay while remaining life-adjusted competitive with a new system, not a market price and not a forecast. Chemistry-level conclusions apply to the specific cells and applications simulated and, as the authors state, may not represent all LFP, NCA or NMC622 batteries. This is general engineering and economic analysis, not investment advice or a recommendation about any specific product, facility or company.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.