Humanoid robots are real but limited: 90-minute battery life means factories need 2 robots per task

Summary

Humanoid robots are moving out of the prototype phase into limited real-world deployment, but a fundamental battery constraint is forcing factories to operate multiple robots per task—a friction that undercuts the economic case for widespread adoption.

The constraint is severe and measurable. Agility Robotics' Digit, the first humanoid robot operating under a paid factory contract, runs for roughly 90 minutes to 2 hours on a full charge when carrying loads. Recharging takes 50 minutes to an hour. That means a factory deploying Digit for a single task needs at least two robots on hand, swapping them out during downtime. Agility's chief technology officer, Pras Velagputi, confirmed the batteries are among the densest lithium-ion chemistries available—nickel, manganese, cobalt with graphite anodes—packed tightly into the robot's torso. Even that optimization yields only incremental gains. Some power is recovered through regenerative energy during walking and arm swinging, similar to braking in electric vehicles, but not enough to close the gap.

The battery problem is structural, not a near-term fix. Battery technology has not followed Moore's law curves. While the industry has long assumed power density would improve on its own timeline, progress has been slow relative to the urgency of the robotics use case. Fusion-based approaches like Avalanche's micro-reactors remain speculative and face the same thermodynamic hurdles as larger fusion efforts—reaching net-positive energy output remains unsolved, and scaling it to small form factors introduces additional engineering constraints.

The gap between hype and deployment is becoming visible. Tesla's Optimus demonstrations at recent events relied on remote teleoperation by staff, a detail that shifted market perception once disclosed. Meanwhile, a crowded competitive field has emerged—One X with its Neo robot, Figure (valued in the billions), Unitree (the Chinese competitor now selling units via Apple Pay for around $20,000), and others—all converging on the same form factor while facing the same battery bottleneck. The number of industrial robots globally hit 4 million last year, but almost none are humanoids; the existing installed base is rigid-arm systems better suited to fixed tasks.

The unresolved question is whether alternative power delivery—electrified floors, overhead charging racks, or tethered systems—becomes the pragmatic middle ground, or whether robotics companies must solve battery density through materials science or alternative energy sources before humanoids become economically viable at scale. Until then, the robots are real but remain constrained to narrow use cases where the cost of dual-robot deployment makes sense.