Date: 2026-03-06 hits: 102
Since the beginning of the year, humanoid robots have been rapidly entering battery factories. BMW is using them on high-voltage battery assembly lines, CATL is transforming its battery production lines, and EVE Energy is using large-scale models to centrally manage their switching between workstations.
Meanwhile, battery companies are treating robots as their next most important customer – LG Energy Solution has launched a cylindrical cell product line specifically designed for humanoid robots, and XPeng has made all-solid-state batteries a core selling point for its humanoid robots.
These are not two coincidental parallel news stories. Robots need better batteries to operate continuously in factories, and battery factories need more flexible robots to keep up with the pace of product iteration.
These two demands began to converge at a visible pace in early 2026.
In late February 2026, BMW announced the deployment of the humanoid robot AEON at its Leipzig plant in Germany, specifying its application not as welding or stamping, but as high-voltage battery assembly and energy module manufacturing. Additional testing was conducted in April, with a pilot phase scheduled for the summer – a very solid timeline.
BMW's choice of this process has its engineering logic: battery assembly demands extremely high operational consistency and involves repetitive, cumbersome, and delicate operations under high pressure—a combination of processes with the heaviest ergonomic burden and the highest risk of error.
Meanwhile, Mercedes-Benz took a different path. On February 11, Apptronik completed a $520 million funding round, with Mercedes-Benz and Google among the participants. The funds will be used to expand the production capacity and commercial deployment of Apollo humanoid robots.
Mercedes-Benz had previously tested Apollo at its Berlin digital factory and its Keczemét plant in Hungary, performing handling and quality inspection tasks.
The two major European automakers, within the same timeframe, bet on the same direction in different ways—linking humanoid robots with battery production processes.
BMW chose direct import, while Mercedes-Benz opted for capital investment, but both focused on battery production lines. This is not a coincidence, but rather a similar judgment formed by the manufacturing industry under the pressure of electrification transformation.
Battery companies are simultaneously expanding at both ends.
If BMW and Mercedes-Benz are seen as "introducers," CATL's logic is more proactive and aggressive—it's simultaneously investing in both factories and capital, attempting to integrate the humanoid robot ecosystem into its supply chain.
On the factory side, CATL has deployed humanoid robots on its battery PACK production lines. According to the company, this is the "world's first new energy battery production line to achieve large-scale deployment of humanoid intelligent robots," and it has provided data on improvements in efficiency and defect rates.
On the capital side, CATL's strategy is more systematic.
Its subsidiary, Morningway Capital, led a nearly 1 billion yuan Series B financing round for Songyan Power; CATL was the lead investor in Galaxy General's previous financing round; its subsidiary, Puquan Capital, followed suit in Zhongqing Robotics; and CATL, along with industry players such as JD.com, Huawei, and Xiaomi, participated in Qianxun Intelligent's nearly 2 billion yuan financing round.
Furthermore, CATL has partnered with Zhiyuan Robotics on a factory application project.
CATL's approach goes beyond the scope of "strategic investment." It's using capital to absorb robotics companies into its ecosystem, while simultaneously accumulating data through actual deployment in factories.
Today, robots are its production line "labor force"; tomorrow, they might be "workers" driven by CATL batteries, ultimately manufacturing batteries for CATL. The completeness of this closed loop will be key to judging the value of this strategy.
Compared to CATL's capital-driven approach, EVE Energy is taking a different path: first selling batteries to robots, then using robots to transform its own factories.
On the product side, EVE Energy has already connected with leading humanoid robot and robot dog customers, and has partially completed sample delivery and assembly.
The product matrix is designed for three operating conditions: for extended battery life scenarios, the 21700-58E and 26105-G26E are used, with energy density exceeding 300Wh/kg and a 35% increase in range; for high-load heat dissipation, the all-tab high-power 21700-50PL is used, simultaneously reducing internal resistance and heat generation by 74% and doubling power.
The 46137LMX series cylindrical batteries, used in extreme safety scenarios, passed the needle penetration non-flammation test. This product matrix isn't just a show presentation; it's an engineering breakdown for three specific operating conditions.
At the factory level, EVE Energy proposed a three-layer architecture: "AI digital employees + physical robot employees + Co-TEE scheduling model." Digital employees replace planning and decision-making processes that previously relied on human experience; physical robots cover handling, loading, inspection, and visual quality control workstations; and a self-developed Co-TEE model optimizes scheduling for multiple robots, workstations, and tasks operating concurrently, solving the questions of "who to assign, when, at which workstation, and how to complete what task." On-line testing is planned to begin this March.
This architecture addresses the core challenge in battery manufacturing: while main processes like coating, rolling, formation, and sorting are relatively fixed, daily parameters, including specifications, customer demands, cycle time fluctuations, and equipment status, are constantly changing. This is precisely the gap that allows AI and robots to intervene.
EVE Energy's approach resembles a "two-sided experiment": simultaneously expanding its customer base for robotics products and validating its own factory's flexible manufacturing capabilities through robotics.
This complements CATL's "capital + factory" strategy, outlining two distinct offensive postures among domestic battery companies in this market.
Furthermore, on March 2nd, LGES announced the official launch of its next-generation battery product line for robots and drones at InterBattery 2026.
The 2170 cylindrical cell product matrix features clear performance gradations for humanoid robot applications: H51 emphasizes a balance between energy density, power, and weight; H52A supports up to 8C ultra-high power output, achieving fast charging in approximately 15 minutes; and M58 prioritizes high energy density for long-duration battery life.
The rationale for choosing cylindrical batteries is stated in the product description: a rigid casing provides greater safety redundancy, standardized dimensions support supply chain stability, and high energy and power density are achieved within a limited space.
Furthermore, according to South Korean media citing industry sources, LGES is the battery supplier for Boston Dynamics' Atlas robot. If true, this would be the most direct testament to how "robot batteries" have gone from being a concept at trade shows to becoming a real supply chain for leading customers.
Meanwhile, XPeng's path offers another perspective.
In November 2025, XPeng released its humanoid robot, IRON, highlighting its all-solid-state battery as a key selling point: a 30% reduction in weight and a 30% increase in battery capacity, with mass production expected by the end of 2026.
LG is pursuing a mass-production approach with mature cylindrical batteries, using standardization to ensure stable supply; XPeng is betting on the performance premium of solid-state batteries, using lightweight design and high density to differentiate itself.
These two paths are not mutually exclusive, but their respective timelines and cost curves will be revealed in the next two to three years.
The emergence of LGES and XPeng signifies that "robot batteries" have moved from concept to real product competition. Whoever develops and stably supplies high-rate, long-lasting, fast-charging battery cells first will control the lifeline of this new hardware category.
Both are heading towards the same goal.
Putting these four stages together, the logical structure of this "two-way race" becomes quite clear.
Robots entering battery factories solve the flexibility problem on the manufacturing side. Battery product iteration is accelerating, but line replacement costs are extremely high. Replacing fixed production lines with reprogrammable robots is essentially trading capital expenditure for operational flexibility.
BMW choosing battery assembly as the first large-scale process for robots, and EVE using a large model to coordinate its robot fleet, are both implementations of the same logic in different scenarios.
Batteries entering the robot body solve the robot's usability problem. Currently, most mainstream humanoid robots have a battery life of only a few hours, and there is a significant gap between the actual demand for continuous shifts in factories and the current energy density of batteries.
This gap is the reason for LG's H52A and XPeng's gamble on solid-state batteries.
Both curves begin to accelerate significantly in early 2026—not because of a sudden breakthrough invention, but because the demands on both sides have simultaneously reached a critical point: the reliability of the robot body has just crossed the threshold for factory trials, and the product iteration speed of battery companies has just begun to catch up with the power and energy density requirements of the robots.
The truly interesting aspect of this two-way race isn't who reaches the finish line first, but rather that when the robots that manufacture batteries and the people who manufacture batteries for those robots finally merge into one entity, the entire manufacturing industry's cost structure and competitive barriers will have to be rearranged.
CATL (Contemporary Amperex Technology Co., Limited) was the first to recognize this, hence its simultaneous bets on both ends. Whether other players see this, and whether they can enter the game in time, will be the most noteworthy industry question to track over the next two to three years.
