Stacking vs. Winding: The 6.3% LFP Cell Energy-Density Gap Revealed by the Xiaomi SU7 Teardown

What’s the story: the same 186.7 Ah, different weights

A teardown comparison of the two types of LFP (lithium iron phosphate) cells used in the Xiaomi SU7 reported a gravimetric energy density of 186.7 Wh/kg for the stacking-type BYD FinDreams Blade cell and 175.7 Wh/kg for the winding-type CATL cell. Both cells have the same rated capacity of 186.7 Ah. The gist of this article is that the gap comes almost entirely not from a difference in the energy stored but from a difference of roughly 200 g in weight, and that where that weight comes from is the real issue in comparing the two processes.

Looking first at the industry situation, the relative merits of stacking and winding have long been discussed in theory, but it is rare to see cells from the same vehicle, with the same chemistry and the same capacity, torn down side by side and compared in numbers. This comparison was part of the “100 Cell Teardown Project” run by China’s Aichi New Energy, was published on Bilibili on August 7, 2026, and was reported by CarNewsChina on August 11, 2026. The CATL cell was introduced as a Shenxing LFP cell.

The tension lies in the interpretation. The report stated that the CATL cell is 6.3% lower in gravimetric energy density, and looking at that number alone it is easy to jump to the conclusion that winding is structurally disadvantaged. Yet the CATL cell also showed signs of choosing safety margin, such as a thick aluminum current collector and a coated separator. This article answers three questions. What exactly does 6.3% mean, where do stacking and winding differ in principle, and how should the scope of rights in the patents surrounding this difference be read?

Basic principle: stacking books versus rolling up a scroll

Stacking cuts the cathode, separator and anode into individual sheets and piles them layer by layer like a stack of books. Because it can fill all the way to the corners, it makes it easier to reduce dead space (empty space without active material) inside a prismatic cell. Winding rolls long electrode strips up like a scroll to make a jelly roll. Winding is fast, which gives it an advantage in tact time (the time needed to process one unit in a step), but slack space remains in the bent sections at both ends of the roll.

By analogy, when you put paper into a square box, stacking individual sheets lets you fill the box right to its corners, but putting in a scroll leaves gaps beside its rounded edges. In exchange, a scroll can be made quickly in one go. The internal makeup of the CATL cell confirmed in this teardown is as follows.

• Winding cores: each cell contains four jelly rolls
• Current collectors: aluminum foil thicker than 20 μm on the cathode side and 11 μm copper foil on the anode side
• Separator: about 12 μm thick with a ceramic/polymer coating
• Coating width: cathode material covers about 78 mm of a collector about 81 mm wide
• Nail penetration test: the voltage during the test was reported as 3.392 V

The point to note here is that a thick aluminum foil and a coated separator are “inactive weight” that does not contribute to energy storage. They help safety and mechanical strength, but in the Wh/kg calculation they only enlarge the denominator. By contrast, the current collector and separator specifications of the BYD cell were not disclosed in this report, so the two cannot be compared on the same yardstick.

Comparison: what the gap really is, from teardown data and back-calculation

The table below summarizes the reported teardown figures. The CATL cell’s dimensions and weight, which the original post listed as “not disclosed”, are stated in the original report and have been filled in.

Item	BYD FinDreams Blade cell	CATL Shenxing LFP cell
Process	Stacking	Winding, four jelly rolls
Dimensions (mm)	620 × 24 × 103	213 × 80 × 95
Weight (kg)	3.2	3.4
Rated capacity (Ah)	186.7	186.7
Gravimetric energy density (Wh/kg)	186.7	175.7
Volumetric energy density (Wh/L)	389.8	369.0
Current collector / separator	Not disclosed in the report	Al over 20 μm, Cu 11 μm, separator about 12 μm (ceramic/polymer coated)

What this table says: at the same 186.7 Ah, the weights differ by 200 g (3.2 kg vs 3.4 kg), gravimetric energy density diverges to 186.7 vs 175.7 Wh/kg, and volumetric energy density to 389.8 vs 369.0 Wh/L.

Calculating these figures backwards reveals what the gap really is. Multiplying the BYD cell’s 186.7 Wh/kg by 3.2 kg gives about 597 Wh, and dividing that by 186.7 Ah gives about 3.2 V. Dividing the same 597 Wh by the CATL cell’s weight of 3.4 kg gives about 175.6 Wh/kg, effectively matching the reported 175.7 Wh/kg. In other words, the energy in both cells was calculated on the same basis, and the gap in gravimetric energy density comes entirely from the 200 g weight difference.

Back-calculated item	BYD cell	CATL cell	Gap
Cell energy (Wh, based on 186.7 Wh/kg × 3.2 kg)	about 597	about 597 (same assumption)	None
Cell volume (L, product of dimensions)	about 1.53	about 1.62	CATL about 6% larger
Gravimetric energy density ratio	Baseline	about 5.9% lower than BYD	BYD about 6.3% higher than CATL
Volumetric energy density ratio	Baseline	about 5.3% lower than BYD	BYD about 5.6% higher than CATL

What this table says: “6.3%” holds when read as how much higher the BYD cell is than the CATL cell; read as how much lower CATL is than BYD, it is about 5.9%. Either way, the source of the gap is weight and volume, not the amount of energy.

The insight that follows is this. The question in comparing processes should shift from “does stacking hold more energy?” to “on which components did the wound cell spend an extra 200 g to hold the same energy?” Candidates include the tabs and connecting parts that tie together four jelly rolls, the thick current collector, the coated separator and the case thickness. How much each contributes was not broken down in this report, so the inability to separate the limits of the process itself from the results of design choices is also a limitation of this comparison.

Patent perspective: the Blade’s rights lie in “shape” rather than “process”

The patent related to the BYD Blade cell whose bibliographic data was verified for this article is U.S. granted patent US12113227B2 (title “Cell, power battery pack, and electric vehicle”, applicant BYD Co Ltd). Its priority date is January 9, 2019, filing date June 21, 2019 and grant date October 8, 2024; according to Google Patents its legal status is Active and its anticipated expiration date is January 21, 2040. The application publication is US20220102787A1.

Restating the gist of the claims, it covers a cell in which the length (L) of the cell body is greater than its width (H), the width is greater than its thickness (D), the length exceeds 600 mm, and the length-to-width ratio (L/H) falls within the range of 4 to 21. The specification presents the effect that such long, flat cells span the entire width or length of the pack, reducing internal partition beams and raising space utilization. Interestingly, the specification does not explicitly address whether the electrodes are stacked or wound. In other words, the center of gravity of this right lies not in the stacking process but in the cell’s external proportions.

Fitting the reported dimensions of the BYD cell in the SU7 (620 × 24 × 103 mm) into the frame of this claim, they can be read as L = 620 mm, H = 103 mm and D = 24 mm, with an L/H of about 6.0. Numerically it appears to fall within the scope of the claim, but this only shows the possibility that BYD’s own product is an embodiment of its own rights. The full set of elements of claim 1 and the scope of rights of the Chinese family patents could not be checked against the full text this time and remain as verification items.

An example of rights on the stacking/stack-folding side is LG Energy Solution’s (original assignee LG Chem) U.S. granted patent US9484593B2 (“Stack-folding type electrode assembly”). Its priority date is October 31, 2013, filing date October 31, 2014, grant date November 1, 2016, legal status Active, and anticipated expiration date October 31, 2034. The original post introduced this patent as claiming “a structure that folds and stacks bi-cells and full-cells with a continuous separator sheet”, but the core of claim 1 as confirmed on Google Patents is a configuration in which a binder coating is placed on both the first separator inside the unit cells and the second separator surrounding the unit cells, with the second binder content higher than the first. The stack-folding structure itself is a premise, and the distinguishing point of the right lies in the relationship between binder contents.

Dividing the design-around scenarios into two directions gives the following.

• Designing around the shape patent: designing a cell length of 600 mm or less, or an L/H outside 4 to 21, could take it outside the literal scope of rights. The counterargument is this. It amounts to giving up the effect of eliminating pack partitions with long cells, so the price of the design-around may come back as a loss of space efficiency at the pack level.
• Designing around the stack-folding patent: making the binder contents of the inner and outer separators equal, or reversing them, could lower the risk of literal infringement of claim 1 of US9484593B2. The counterargument is that the adhesion of the outer separator weakens, which could revive the loosening and swelling problems the specification points out.

Summarized from an FTO (freedom-to-operate) perspective, the possibility of a rights conflict is determined less by the process choice of “stacking or winding” itself than by specific parameters such as the cell’s external proportions and separator adhesion design. Conclusions may vary with claim-by-claim comparison, and they are organized here only at the level of possibility.

Limitations and outlook: cell numbers and pack numbers are different

Stacking has a large effect in reducing dead space in large prismatic cells, but the precision required of the stacking and notching (cutting electrode tab shapes) equipment that aligns individual sheets is high, increasing the burden of equipment investment. Winding still holds a tact-time advantage, so it may continue to be adopted in cost-sensitive segments.

There is also a counter-scenario to the optimistic outlook. This comparison is only of cell-level figures, and the two cells are completely different in shape. The 620 mm long Blade cell and the 213 mm CATL cell differ in arrangement, cooling structure and part count within the pack, so the cell-level gap of 5 to 6% could shrink or even reverse at the pack level. Pack-level energy density was not presented in this report.

As a strategic alternative, stack-folding (Z-folding), a compromise between the two processes, continues to be considered. However, as seen above, rights in this area also hinge on detailed parameters such as separator adhesion design, so a process switch must proceed together with both equipment review and FTO review.

Practical action: planning and IP staff should build a table that breaks down, item by item, the “inactive weight needed to hold the same Wh” (current collectors, separator, tabs, case) for each of their own cell formats, and compare against the same items whenever a teardown report on a competing cell comes out.

This article is a commentary on technology and patent trends, not legal advice, and judgments on the scope of rights of individual patents require confirmation by a patent attorney or lawyer.

Items requiring verification

• Thickness and makeup of the current collectors and separator of the BYD Blade cell (not disclosed in this report)
• Component-by-component breakdown of the 200 g weight difference between the CATL and BYD cells
• How the original teardown data labels the basis on which the reported “6.3%” was calculated
• Full text of claim 1 of US12113227B2 and the scope of rights of its Chinese family patents
• Pack-level energy density of the SU7 battery packs fitted with the two cells

Sources

1. CarNewsChina, “BYD vs CATL battery cells dissected: Xiaomi SU7 teardown reveals key differences”, 2026-08-11
2. Google Patents, US12113227B2
3. Google Patents, US9484593B2