Humanoid Robot Battery & Power Supply BOM: CATL, Unitree vs Tesla Optimus

The runtime problem, in numbers

A humanoid spends almost no BOM dollars on energy — the pack is roughly 1–3% of total cost — but it dominates the product spec. Two hours of walking, lifting and manipulation is just barely enough for a work shift; every extra minute demands more weight, which demands more motors, which demands more battery — the classic robot power spiral. The published numbers make the constraint obvious:

Note the pattern: even Tesla's larger 2.3 kWh pack only buys ~2 hours of active work, because a walking humanoid draws high, spiky currents. That is why the design focus is not bigger batteries but hot-swap packs and lower actuator energy consumption (see our BOM cost breakdown).

Chemistry: why high-nickel, and when LFP loses

• High-nickel NMC/NCA (250–300 Wh/kg) — dominant. It is the only chemistry that fits enough energy in a weight- and space-constrained torso while delivering the high-rate discharge walking demands. This is the chemistry in most humanoid packs today.
• LFP — niche. Safer and cheaper, but its lower density powers a humanoid for only ~1 hour. It survives only in indoor service robots where safety and cost beat endurance.
• Solid / semi-solid-state — fastest growing. The next lever for both density and safety; semi-solid packs are already appearing in robot demos (e.g. Joyson Electronics' solid-liquid hybrid cell) and are expected to extend runtime before 2030.

The Chinese EV battery chain is the hidden supplier

Here is the under-appreciated point: humanoid batteries are just EV cells repackaged. The world's leading battery makers are Chinese, and they are already on the robot's board. CATL demonstrated the first heavy-duty humanoid in regular factory operation (the 50 kg-payload Galbot S1) running on CATL batteries, and its portfolio already includes 280 Wh/kg Qilin cells and 350 Wh/kg condensed (aviation-grade) cells — the exact density range humanoids will demand. Other suppliers draw on the same cylindrical and pouch-cell ecosystem that serves China's EV industry, which is why a humanoid pack can be costed and scaled almost on EV terms.

What to watch

• Semi-solid-state migration: which humanoid first ships a mass-production pack above ~300 Wh/kg, and the runtime jump it unlocks.
• Hot-swap standardization: whether quick-swap packs become a fleet model (swap-and-charge off-peak) rather than single packs.
• Power vs payload trade-off: heavy-duty humanoids (50 kg payload, like Galbot S1) need even larger packs — watch for dedicated heavy-humanoid battery modules from CATL/EVE/others.

Frequently asked questions

How big is a humanoid robot's battery?

Most humanoids run 0.5–2 kWh. Unitree H1 uses a 0.864 kWh pack for under 4 hours; H2 PLUS uses 0.972 kWh for about 3 hours; Tesla Optimus Gen 2 uses a 2.3 kWh high-nickel pack for about 2 hours of dynamic operation.

What chemistry do humanoid robot batteries use?

High-nickel NMC/NCA dominates (250–300 Wh/kg) because weight-constrained humanoids need maximum density and high-rate discharge. LFP is safer but too low-density beyond ~1 hour; solid and semi-solid-state are the fastest-growing segment.

Who supplies humanoid robot batteries?

The Chinese EV battery chain is the de facto supplier. CATL powers the Galbot S1 heavy humanoid and supplies Qilin-class 280 Wh/kg and condensed 350 Wh/kg cells; others draw on the same cylindrical/pouch supply that serves EVs.

How much does a humanoid battery cost?

The pack is roughly 1–3% of total BOM cost, a small share, but it dominates runtime and weight — the bottleneck for autonomy rather than the biggest line item.