Closed-Loop Recycling - Battery Design

In September 2026 GM announced that cells containing cathode active material (CAM) made with 100% recycled nickel, cobalt and manganese are now in customer vehicles [1][2]. A week earlier Porsche reported test cells with a fully recycled cathode, lithium included [3]. Both are being called “closed-loop recycling”. This page sets out what that term should mean to a cell or pack engineer, runs a sanity check on the GM numbers, and looks at what CATL, Porsche, Redwood, BMW and the University of Birmingham are doing in the same space. It sits alongside Recycling Process Engineering, which compares the recovery routes, and Direct Cathode Recycling, which covers relithiation in detail.

What “closed loop” actually means

In an open loop, a recycler sells recovered metal into the commodity market and it ends up wherever the price takes it: alloy, chemicals, or somebody else’s battery. In a closed loop, material from a spent product goes back into the same product. For batteries that can be read at three levels, and announcements rarely say which one they mean:

Element level: recycled nickel, cobalt or lithium salts go back into cathode precursor. This is what most hydrometallurgical flowsheets deliver, and it is what the EU recycled content targets count.
Material level: the cathode particle itself is repaired and reused, the direct recycling route.
Company level: the metal comes out of the OEM’s own retired packs and back into its own new cells. This can be physical (the same atoms) or a material account, where recycled tonnage is booked against an OEM’s production in the way green electricity is booked against a factory.

The engineering questions to ask of any “closed-loop” claim are therefore: which elements, at what percentage, from what feedstock, by which route, and is it physical or mass-balance accounting?

Case study: GM and Cirba Solutions

The published facts are straightforward. GM recovered 80 end-of-life high-voltage packs, which Cirba Solutions disassembled and processed into black mass at its Ohio facility [1][4]. That black mass was refined into new CAM through a further partnership that has not been named publicly [5]. Ultium Cells, the GM and LG Energy Solution joint venture, made the cells, and the modules and packs were built at Factory Zero and Spring Hill [1][6]. GM states the pilot yielded more than 12 tonnes of CAM and that the vehicles include variants of the Cadillac Lyriq, Vistiq and Escalade IQ, Silverado EV and Sierra EV [1].

Three things are worth reading carefully:

“100%” applies to nickel, cobalt and manganese only. Lithium is not part of the claim. GM’s own figure for current processes is recovery of up to 95% of nickel, cobalt and manganese and up to 80% of lithium, depending on chemistry and route [2]. Ultium cells are NCMA [7], so our reading is that the aluminium dopant and the lithium are most likely virgin material.
Cirba made black mass; someone else made the CAM. The hydrometallurgical refining, precursor co-precipitation and calcination steps happened at a partner GM has not named. Some secondary write-ups attribute the whole flowsheet to Cirba, which the primary reporting does not support.
No electrochemical data has been published. GM says the cells performed comparably to cells made from virgin material in its testing [2]. That is a company claim; no capacity, cycle life or impedance comparison is in the public domain, and we could not find a peer-reviewed paper on this material.

Sanity check: does 80 packs make 12 tonnes?

We can check this from the headline numbers. Our assumptions: the new CAM is a high-nickel NCMA of roughly LiNi_0.89Co_0.05Mn_0.05Al_0.01O₂, a high-nickel layered oxide uses 1.2 to 1.6 kg of CAM per kWh of cell, and the Ultium cell is about 103 Ah at 3.7 V, as on our GM Ultium page. Nickel’s mass fraction in the CAM is its share of the formula weight:

w_{Ni} = \frac{0.89 \times 58.69}{97.1} \approx 0.54

where 58.69 g/mol is the atomic mass of nickel and 97.1 g/mol is the formula weight of the NCMA. Applied to 12 t of CAM, this gives the following.

Quantity	Value	Working
Nickel in 12 t CAM	6.5 t	12 t x 0.54
Cobalt / manganese	0.36 t / 0.34 t	Same method
Cell energy enabled	7.5 to 10 MWh	12,000 kg / 1.2 to 1.6 kg/kWh
Ultium cells	~20,000 to 26,000	Energy / 381 Wh per cell
Equivalent new packs	~40 to 100	200 kWh truck pack to 100 kWh Lyriq-class pack
Nickel needed per returned pack	~85 kg	6.5 t / 95% recovery / 80 packs
Nickel in a 100 kWh NCMA pack	~73 kg	100 kWh x 1.35 kg/kWh x 0.54
Nickel in a 66 kWh NMC622 pack	~41 kg	66 kWh x 1.7 kg/kWh x 0.36

Our arithmetic from GM’s headline figures; composition and kg/kWh values are our assumptions, not GM data.

Two conclusions come out of this. First, the output is plausible in energy terms: 12 t of CAM is roughly as much cathode as 80 new packs would need, and an independent Spanish estimate arrived at a similar figure of about 8 MWh [8]. Second, the nickel balance is tight. Unless most of the 80 returned packs were large, high-nickel Ultium packs, they would not have contained enough nickel to make 12 t of high-nickel CAM on their own. A returned fleet of older, smaller NMC packs would cover less than half of it. GM’s wording, that the material includes metal recovered from its batteries, leaves room for other recycled feed such as cell manufacturing scrap [8]. That would still be recycled nickel, but it would not be a pack-to-pack loop in the strict sense.

Which packs were returned therefore matters, and GM has not said. None of the reporting we found states whether the 80 packs came from customer vehicles, warranty returns or GM’s own fleet [1][23].

Nigel: My guess is that GM is recycling its own packs: ones built for engineering and demo vehicles, or packs that failed end-of-line testing. For a first loop you would want feedstock you know everything about, and your own development fleet is the obvious source.

That reading also closes the nickel balance. Development and validation fleets for the Hummer EV, Silverado EV and Sierra EV carry Ultium packs of around 200 kWh, which on our assumptions hold roughly 145 kg of nickel each (200 kWh x 1.35 kg/kWh x 0.54). A mix of these and Lyriq-class packs would cover the 6.5 t comfortably without any scrap feed. Either way, a loop fed by known in-house packs is a qualification loop; it is not yet evidence that customer returns are flowing back.

The scale also needs context. A cell plant of 40 GWh a year produces about 4.6 MWh an hour, so the whole pilot output is under two hours of production at that rate. This is a qualification exercise, not a supply chain.

What others are doing

CATL and Brunp

CATL’s recycling subsidiary Brunp is the volume reference point. CATL says Brunp processed more than 120,000 t of spent batteries in 2024, producing 17,100 t of recycled lithium salts, and claims recovery rates of 99.6% for nickel, cobalt and manganese and 96.5% for lithium [9]. In July 2026 the European Patent Office gave two Brunp inventors the European Inventor Award for the underlying “directional recycling” process; the EPO describes it as regenerating cathode material directly from spent cells and manufacturing scrap while keeping its functional structure, and reports a 61% lower carbon footprint than conventional cathode production [10].

Two cautions. CATL’s own description of the process lists intelligent disassembly, hydrometallurgy and material restoration [9], so it reads as a hybrid route rather than pure direct recycling. And the recovery and carbon figures are company and award-body statements without a published test method. What is not in doubt is the scale: Brunp handles in a few hours more material than the whole GM pilot, and as a captive recycler inside the world’s largest cell maker it is a closed loop at company level by default.

Porsche and cylib

Porsche and the Aachen recycler cylib have made battery cells whose cathode active material is 100% recycled, including lithium, recovered from Porsche’s own high-voltage packs using a water-based process [3]. The cells are going into test vehicles rather than customer cars, and the recovered material is held in a material account, to be available to Porsche and selected partners for cell production from 2028 [3]. So Porsche is ahead of GM on element coverage and behind it on production deployment.

Redwood Materials

Redwood takes the blended approach. When its Panasonic supply deal was announced, the CAM was to contain 100% recycled cobalt and around 30% recycled nickel and lithium [11]; the later Toyota agreement set minimums of 20% recycled nickel and lithium and 50% recycled cobalt [12]. This is the more realistic model for volume: recycled content as a percentage of a mixed feed, rising as end-of-life packs arrive. We have not verified current delivery status against those announcements.

BMW: recycled content as a purchasing requirement

BMW approaches it from the buyer’s side rather than the recycler’s. When it announced the Gen6 cylindrical cell for the Neue Klasse in 2022, BMW said it had agreed with its cell partners that they would use a specified proportion of secondary lithium, cobalt and nickel, and would make the cells using only renewable electricity [20]. In other words, recycled content was written into the cell supply contract. That is the clearest public evidence we found that recycled content is now a line in OEM cell specifications and RFQs, not just a sustainability report aspiration.

The published figures are worth reading closely. For the iX3, BMW states that the Gen6 cells “are made of 50% secondary cobalt, lithium, and nickel materials”, with CO₂e per watt hour 42% lower than the Gen5 cell [21]. For the i3 seven months later, the wording is that the cells “incorporate a share of secondary cobalt, lithium and nickel”, with a 33% reduction per watt hour [22]. Neither release says whether 50% means half of each element or half of the combined mass, whether it is physical content or a mass-balance allocation, or how much comes from end-of-life packs rather than production scrap. Our reading is that the later, vaguer wording is the more cautious one, and that the iX3 figure should not be quoted as a per-element value until BMW defines it.

University of Birmingham and the Faraday Institution ReLiB project

ReLiB, led by Birmingham under Faraday Institution funding, is working on the problem that the industrial routes skip: what to do with mixed or blended cathodes without dissolving everything. Driscoll et al. took blended cathode from a first-generation Nissan Leaf cell retired at about 40,000 miles, used ascorbic acid to leach the LiMn₂O₄ spinel selectively within minutes, and regenerated the remaining nickel-rich layered oxide with a hydrothermal treatment that also removed the PVDF binder [13]. The lithium and manganese in the leachate can be upcycled into NMC rather than downcycled [13]. The current phase of ReLiB targets separating 10 kg batches of commercial black mass into purified, regenerated CAM by April 2027 [14], and an industry sprint with Recyclus is adapting the same physical separation methods to LFP black mass [15].

Our reading is that this is the UK’s most relevant contribution: it attacks feedstock variability, which our Direct Cathode Recycling post identifies as the main barrier to taking direct routes from lab to line. It is still at kilogram scale.

Does recycled CAM perform?

The best public evidence predates all of the above. Ma et al. at Worcester Polytechnic Institute made NMC111 from a mixed spent-battery feed by a hydrometallurgical route and built 1 Ah multilayer cells. Cycled at 45°C, the recycled cells reached 80% capacity after 4,200 cycles against 3,150 for a commercial control [16]. That is one chemistry, one feed and one cell format, and NMC111 is far more forgiving of impurities than the nickel-rich CAM GM is making. It shows parity is achievable, not that it is automatic.

Side by side

Programme	Route	Recycled elements in CAM	Scale	In customer vehicles	Performance data public
GM / Cirba	Black mass, then hydromet and resynthesis by unnamed partner	Ni, Co, Mn 100%	12 t CAM pilot	Yes	No
Porsche / cylib	Water-based process	Li, Ni, Co, Mn 100%	Pilot	Test vehicles	No
CATL / Brunp	Disassembly, hydromet and material restoration	Not stated per cell	>120,000 t/yr feed	Not stated	No
Redwood	Hydromet, blended feed	Co 50 to 100%, Ni and Li 20 to 30%	Commercial contracts	Not verified	Not found
BMW Gen6	Contractual requirement on cell suppliers	Li, Co, Ni “50%” (iX3), undefined	Series production	Yes	No
Birmingham ReLiB	Selective leaching plus direct regeneration	Phase by phase	Lab, 10 kg target	No	Yes, peer reviewed

Closed-loop cathode programmes, October 2026. From the references below; company claims as stated.

Where the regulation sits

Under the EU Battery Regulation, EV batteries must contain at least 16% recycled cobalt, 6% lithium and 6% nickel from 18 August 2031, rising to 26%, 12% and 15% in 2036 [17]. Recyclers must recover 90% of cobalt, nickel and copper and 50% of lithium by the end of 2027, rising to 95% and 80% by the end of 2031 [17]. One detail matters for anyone tempted to hit the targets with their own scrap: battery manufacturing waste counts towards recycled content, but by-products reused directly in the production process do not [18]. The recycled content declaration also has to be carried on the battery passport.

Set against those minimums, GM and Porsche are far ahead on percentage but tiny on volume. The regulation does not need 100% recycled cells. It needs a few percent in every cell, which is the Redwood model.

What to take from this

The chemistry question is largely answered. Recycled nickel, cobalt and manganese can be refined to a purity that a high-nickel cell line will accept, and GM has now shipped it. The open questions are volume and accounting. There are not yet enough retired packs to feed a gigafactory, so for the rest of this decade most recycled feed will be manufacturing scrap [19], blended with virgin material. “100% recycled” headlines tell us a pilot passed qualification; they do not tell us what a production cell contains.

BMW has shown that recycled content can be a contractual cell requirement; its own releases also show how easily the number loses its meaning. For engineers specifying cells, the practical step is to ask for recycled content element by element (Li, Ni, Co, Mn, Al), the feedstock split between end-of-life and scrap, and whether the figure is physical or a material account. For pack designers, the lesson from ReLiB and Brunp is that the value recovered at end of life depends on how cleanly the cathode can be separated, which brings us back to design for disassembly and chemistry labelling. As retired Ultium packs start to arrive in volume early in the 2030s, we expect the GM loop to become a genuine pack-to-pack one, and that is the point at which it will be worth revisiting these numbers.