Warehouse AMRs can already run a full shift with predictable docking and energy budgets. Many bipedal humanoids, by contrast, still advertise on the order of a few hours per charge. That gap is not a marketing accident. It is what you should expect when you replace a wheeled base on a flat floor with a walking machine that must continuously fight gravity, balance disturbances, and peak torque events.This is an energy-system problem involving locomotion physics, thermal limits, and operations—not a single vendor runtime claim. Public manufacturer materials commonly cite roughly two to five hours depending on duty cycle. Treat those numbers as scenario-dependent, not as a promise that every tote loop or machine-tending cycle will match a brochure.The physics tax of walkingWheeled robots mostly pay for rolling resistance, acceleration, and accessory loads. Bipeds pay for continuous joint torque to keep the center of mass above a moving support polygon. Every step involves swing-leg acceleration, stance-leg support, and ankle and hip corrections when the floor or payload surprises the controller. Degrees of freedom that look impressive in a specification table are also actuators that draw current even when the robot is just standing.Peak events matter more than average watts. Catching a fall, lifting a tote from a low rack, or recovering from a slip can spike current far above steady walking. Packs and motor drives must handle those spikes without tripping protection or overheating. Thermal limits on actuators and batteries then constrain how long a robot can sustain high-load work even if the pack still has energy on paper. A robot that walks gently in a lab video may need derating on a real line where stops, starts, and manipulation dominate.The form factor is part of the tax. Human-height mass, torso packs, and cable routing trade away space that wheeled platforms use for larger packs and simpler cooling. Until energy density and actuator efficiency improve together, bipeds will keep paying more joules per meter than comparable wheeled bases on the same floor.Design responses already shippingHot-swap and self-swap systems attack the problem as operations as much as chemistry. A body keeps working while a depleted pack charges offline. Autonomous or assisted exchange replaces the fantasy of one eight-hour pack with a logistics loop of spare batteries, docks, and timing. Continuous duty becomes a property of the fleet and energy station, not one pack’s watt-hours.Wireless and opportunity charging let a robot top up between short tasks without a human plugging a cable. Inductive feet and charging mats work when a duty cycle brings the robot back often enough that average power stays inside the battery envelope. Opportunity charging rewards predictable routes and short gaps more than heroic one-shot endurance.Duty-cycle engineering matters too. A credible four-hour robot plus a twenty-minute charge strategy can beat an optimistic eight-hour pack that never ships. Buyers should ask whether runtime was measured while standing, walking slowly, or running continuous tote loops. Wheeled-legged hybrids and wheeled upper bodies appear where stairs are rare because they retain human-height manipulation while cutting the balance tax.What buyers should askProcurement teams should ask for sustained and peak current at the intended payload, swap time and human involvement, spare-pack assumptions, charge infrastructure cost, and the exact measurement conditions. Ask what happens to payload or speed after thirty minutes of continuous high load. If a vendor cannot separate peak from average or describe the energy station, treat the brochure runtime as a ceiling, not a plan.TakeawayBattery life for humanoids is a system problem: locomotion efficiency, actuator heat, pack chemistry, and recharge logistics all matter. Until energy density and walking efficiency improve together, uptime winners will design swap and charge operations as carefully as they design gaits. Compare commercial platforms in the HumanoidRobotList directory, and read runtime and swap notes on Digit, Walker S2, Atlas, and NEO before modeling a full shift.