Humanoid robots are edging into factories and homes, but their batteries run dry too quickly to be truly useful. Most full-size humanoids operate for just 2 to 4 hours on a charge, with battery packs typically under 2 kWh. Even premium units priced near one million yuan (about $140,800) often need to head back to a charging station before they can finish a shift.
That endurance gap is now pulling the automotive supply chain into the robotics business. In a sign of the shift, a Chinese auto parts maker has announced that its semi-solid-state battery has reached mass-production readiness, according to D1EV, an industry news outlet.
Why robot batteries aren't just scaled-down car batteries
Robot batteries cannot simply be shrunk versions of EV packs, analysts note. Market researcher TrendForce points out that most humanoid robots carry less than 2 kWh of battery capacity because of limited torso space and weight constraints. Unitree's H1 robot, for example, has a 0.864 kWh pack that lasts under 4 hours at rest, while Tesla's Optimus Gen2 uses a 2.3 kWh high-nickel ternary battery that lasts about 2 hours of dynamic operation. Keeping a robot upright without mechanical locks consumes 200 to 300 watts just for balance.
Enlarging the battery is not a fix: the torso is already packed with joints, motors and controllers, and a heavier pack raises the center of gravity, making balance harder. The solution has to come from chemistry.
In May 2026, Joyson Electronics and Beijing Enpower Energy jointly set up Junen New Energy to develop solid-state batteries. Within months, the venture unveiled a semi-solid-state cell using a sulfide electrolyte. The design reduces liquid electrolyte content and optimizes the cell chemistry and pouch structure, achieving energy densities of 300 to 600 Wh/kg, with a typical cell density of 380 Wh/kg. Junen says this boosts range by about 60% and supports over 2,000 charge cycles, with 80% charge in 30 minutes via wired fast charging or wireless charging.
High power output is another hurdle. Robots need about 3C continuous discharge for lifting and fast movement, with peak bursts of 5C to 15C, sometimes hitting 20C. That demands far more from the battery than consumer electronics do.
Junen's integrated package—cells, modules, battery management system (BMS), data services and charging solutions—targets humanoid robots, wheeled robots, robot dogs and industrial AGVs. The company says its BMS, adapted from Joyson's automotive-grade technology, estimates state of charge with 1% to 3% accuracy, two to three times better than conventional systems, and can warn of thermal runaway 30 minutes to 2 hours in advance.
Tier 1 suppliers bet on robots as 'second growth curve'
Joyson is not alone. Since 2025, Bosch, Schaeffler, Valeo and Aptiv have all moved into embodied AI. The logic is technology overlap: perception sensors, domain controllers, actuators and thermal management are similar to what these firms already build for cars.
Valeo is adapting ultrasonic sensors, cameras, radar and lidar, plus actuators and thermal and power management, for robots. Aptiv has launched its PULSE fusion perception system for collaborative and autonomous mobile robots. Schaeffler, drawing on decades of bearing and precision component know-how, set up a dedicated robotics company in China in February and plans to deploy thousands of humanoid robots in its global factories by 2035. Bosch is investing in and partnering on robot components, and has signed a deal for 2,000 embodied robots in China.
The deeper play is data. Goldman Sachs research cited in the report found that AI models that achieve 80% to 90% accuracy in simulation often drop below 50% in real-world conditions. Tier 1 suppliers are turning their own factories into robot training grounds, using real production-line variations—lighting changes, vibration, material tolerances—as scarce training data. That helps solve the cold-start problem for robot deployment and builds a data feedback loop.
"Whoever wins, they can't avoid reliable core component supply," the report quotes industry sentiment. Being the "shovel seller" is the common choice. Market forecasts cited in the article project demand for solid-state batteries in humanoid robots could exceed 74 GWh by 2035, a thousand-fold increase from 2026.
Still, questions remain over data ownership, standardization, and the yield and cost of scaling from samples to mass production. The semi-solid-state battery's journey from sample to production is just one test of the Tier 1 suppliers' technology-scenario-data path, the report concludes.
