CBAK's 18 GWh plan has no date on it, and its welder runs at half the winder's rate

Hold one winding machine at its rated speed for a full calendar year with no stops and it produces about 126 million cells ^[1]. Set that against CBAK's cumulative Model 32140 deliveries and the entire shipment history to date amounts to roughly nine months of one machine running flat out ^[7]. Divide the 18 GWh target by the same annual figure and each cell would have to hold about 143 Wh ^[3]. Winding speed was never going to be the wall.

The wall sits behind it. On-the-fly laser welding of current-collector disks is quoted at up to 120 cells a minute ^[6], half the winder's rate, so keeping the two in step means two welding stations per winder ^[4]. The other laser process, at an average 0.35 seconds a weld ^[7], comes to about 171 welds a minute ^[5]: adequate if a cell needs one of them, short if it needs two. How many stations sit at each step is not stated, and 3 GWh remains the only capacity figure with installed plant behind it ^[1].

The site plan is stacked rather than spread. Roughly 2.9 million square feet of planned gross floor area on about 58.8 acres ^[10] is a floor-area ratio near 1.13 ^[6], which puts material movement into lifts and onto the automated guided vehicles running between stages ^[11].

Two of the plant's numbers come back perfect, and both are perfect against a scope CBAK drew itself: 100 percent of the line's defined core manufacturing processes are automated [c11b], and the AI weld inspector shows a 100 percent detection rate for insufficient-fusion defects within its defined inspection scope, according to the company ^[14]. Neither states what share of total process steps or total defect modes those scopes cover. The bounded claims are the more useful ones, such as dual-roll continuous calendering said to improve anode thickness consistency by 30 percent ^[8] and a weld geometry that adds about a third to current-carrying area ^[7].

More than 100 critical parameters are inspected and tracked during production, and UV marking ties each finished cell's identification code to its jelly-roll code ^[12][13]. A failure traced to a jelly roll is containment. A failure traced to a month of output is a recall. That distinction is worth more to a 2027 design than the headline capacity, which CBAK describes as part of a long-term development plan and explicitly not the facility's current output, with no target date given ^[15]. Chenggang Yuan, deputy general manager of Nanjing CBAK New Energy Technology, framed the plant in the same register, as process discipline, automation and manufacturing traceability operating "as an integrated system" ^[17]. What a hardware team can actually design around is the format: one set of dimensions the whole six-fold build is tooled to produce ^[16][2]. The rest of the announcement is a plan.