Figure AI Ramps Up Humanoid Robot Production to One Unit Per Hour

Figure AI has increased production of its Figure 03 from one robot per day to one per hour in less than four months. The company said more than 350 robots have now been produced through its BotQ manufacturing facility.

Figure described BotQ as the foundation for scaling commercial deployment and AI development for its humanoid robotics platform. The company said the faster production ramp was achieved through dedicated assembly lines, custom manufacturing software, and stricter supplier quality controls.

“By overcoming the massive technical hurdles of high-volume manufacturing, Figure has unlocked a major catalyst for humanoid development: physical scale,” the company stated. “Every robot that rolls off the BotQ line at our new hourly cadence is more than just a unit of hardware.”

The company said its manufacturing software now operates across more than 150 workstations. Figure also reported end-of-line first-pass yields above 80%, while its battery production line achieved a 99.3% first-pass yield with more than 500 battery packs shipped.

Figure revealed it has produced over 9,000 actuators across more than 10 product variants. Each humanoid robot undergoes more than 80 verification tests, including full-body stress movements such as squatting, jogging, and shoulder presses before deployment.

The growing robot fleet is helping improve Figure’s Helix system by feeding operational data back into the platform. The company said this supports better diagnostics, fault recovery, and long-duration performance in real-world environments.

Alongside the production update, Figure introduced a new perception-conditioned whole-body control capability for its System 0 model. The feature allows Figure 03 robots to navigate stairs, ramps, and uneven terrain using onboard stereo camera perception without task-specific programming.

Figure said the model was trained through reinforcement learning in simulation before being transferred directly to physical robots without additional tuning. The company described the update as part of its broader push toward end-to-end learned control systems for humanoid robots.