Repurposing EV batteries for grid storage

Kann frames the decision as a fork, and it is the standard framing: an end-of-life pack goes either to materials recycling or to stationary storage, and which one wins is an economic calculation over the value of recovered cathode and anode material against the value of grid services, net of processing or refurbishment cost in each case, with new lithium iron phosphate storage as the competing option. Kann qualifies that competing price: outside the United States it has been a falling knife, while inside it tariffs and similar factors make it more complicated. Historically the fork resolved to recycling. The packs themselves are in better shape than that framing implies. Campbell says they arrive physically almost new, because a vehicle battery is an extremely durable engineered object, and electrically they arrive when the owner gets frustrated rather than when anything fails, typically around 20% less capacity and 20% more impedance. They are predominantly nickel-based NMC cells, because what comes back today is roughly what was built a decade ago.

Campbell’s first move is to reject the fork. Redwood does not choose: it sends the pack to a grid site to earn grid services and then recovers the metals afterwards, which he describes as a detour rather than a destination. That reframing is what makes the economics work, because it means the grid stint does not have to beat recycling, only add to it. The second move is the surprising one. Because the cost of getting a pack onto a grid site is now very low, the amount of useful life a pack needs to have left is correspondingly small, on the order of one to two years and low hundreds of cycles. Kann works out the implication out loud: for that to be true against a new purpose-built system with a ten-year warranty, the fully installed cost of the second-life asset has to be far below the new one. Campbell agrees, and volunteers that he was sceptical of second-life storage until roughly the previous year, on the reasonable grounds that a repurposed object rarely beats something optimized for the job. What changed his mind was the return volume arriving at the same time as a cheap way to install the packs.

The integration cost is the part that fell, and the description is almost anticlimactic. A five-minute electrical inspection checks cell balance and impedance and whether the pack’s internal electronics still report diagnostics; if those come back clean, the pack is wheeled into the field and plugged in. The packs are not opened and the modules are not removed, so the site amounts to a parking lot of electric vehicles with the wheels taken off. Kann presses on the obvious objection, which is that this sounds like something anyone could do. Campbell’s answer is three things rather than one: power electronics designed to talk to an extremely wide variety of pack types and dispatch each sensibly as part of one coherent asset, site mechanical design cheap enough not to eat the margin, and access to the feedstock in the first place, since Redwood already collects north of 80% of end-of-life electric vehicle packs in the United States. The third of those is positional rather than technical, and Campbell says so himself: collecting packs like this is not simple, and what is hard to copy is aggregating a heterogeneous stream arriving from a million places.

That heterogeneity also shapes where these assets fit on the grid. Campbell says second-life packs can compete head to head in the two-hour and four-hour markets, but they shine at longer durations, four to eight hours and possibly as much as twenty. The technical reason is that a pack’s degradation shows up as cell imbalance, and because cells sit in series the whole pack is limited by its weakest member during hard discharge; discharge slowly and that weakness matters much less. Kann correctly points out that this is true of any battery, not just a used one, and that the reason nobody builds twenty-hour lithium-ion projects is economic, since cost scales roughly linearly with duration. Campbell concedes the point directly: it is not a different equation, the energy is simply cheap enough that stacking more of it stays affordable. On scale, the supply curve is unusually predictable because it was set a decade ago by vehicle production. Roughly five gigawatt-hours a year is coming off US roads now against roughly fifty gigawatt-hours of US storage deployed in the prior year, so the retirement stream is about a tenth of deployment; Redwood expects to deploy low single-digit gigawatt-hours of second-life storage this year and next.