Related: Figure 03 humanoid robot starts handling logistics at BMW’s Spartanburg plant.
Figure AI says it has cracked one of humanoid robotics’ hardest problems, not building a capable robot but building enough of them fast enough to matter. In an update posted to its newsroom, Figure said its BotQ manufacturing facility increased production of the Figure 03 humanoid from one unit per day to one unit per hour, a 24 times increase in throughput achieved in under 120 days. The company has now delivered more than 350 of the third-generation robots, a scale that founder and CEO Brett Adcock has framed as the real test of whether humanoid robotics can become an industry rather than a demo.
The jump from prototype-level output to an hourly cadence is the kind of milestone that tends to separate companies that can show off a robot from companies that can actually deploy one. Figure said the transition ran through dedicated production lines for every critical module of the system, coordinated by custom-built manufacturing execution software running across more than 150 networked workstations. Getting there required tightening quality at the supplier level first, since a single bad part can stall a line, so Figure said it qualified suppliers against strict incoming inspection criteria and layered in more than 50 in-process inspection points across the build. That effort shows up in the numbers the company is now reporting, an end-of-line first-pass yield above 80 percent and climbing weekly, a battery line running at 99.3 percent first-pass yield with more than 500 packs shipped, and over 9,000 actuators produced across more than 10 distinct part variants.
Getting a robot off the line is not the same as trusting it to work, so Figure said every unit goes through more than 80 functional verification tests before sign-off, including multi-limb stress testing and burn-in sessions where the robot performs squats, shoulder presses, and jogging at cycle counts in the thousands to flush out early failures before they reach the field. It is the kind of grinding, unglamorous quality process that consumer electronics and automotive manufacturers have run for decades, now being applied to a machine that walks.
A bigger fleet also creates a bigger support problem, and Figure says it built new infrastructure to handle that rather than treating it as an afterthought. The company described an internal Field Service Management system built to service Figure 03 units wherever they end up, whether that is Figure’s own headquarters, a customer’s factory floor, or eventually a home. Paired with a custom Fleet Management System that tracks the health, location, and operational status of robots in real time, plus over the air software updates, Figure said it can now push new behaviors and fixes to its entire fleet at once and run coordinated recall campaigns when a hardware issue needs a fix across every unit already in the field.
Scale is also feeding Figure’s software roadmap directly. The company said the larger fleet is generating the data needed to train Helix, its humanoid AI model, and pointed to a new capability it calls perception-conditioned whole-body control. Previously, Figure’s S0 walking controller reasoned only about the robot’s own body, meaning it could walk confidently on flat ground but needed manual intervention on stairs or uneven terrain. The updated system now feeds camera-based 3D perception into the same controller, trained through reinforcement learning across randomized terrain in simulation, and Figure says the resulting behavior transfers to real hardware with no additional fine-tuning, letting robots climb stairs they were never explicitly programmed to handle.
Entrelligence has previously covered Figure 03 robots going to work on BMW’s factory floor in Spartanburg, and this production update is the other half of that story, the manufacturing capacity that has to exist before robots can show up at customer sites in meaningful numbers. Figure has not confirmed a specific annual production target publicly on its own site, though the company has signaled that velocity is still increasing rather than plateauing. If BotQ keeps climbing from here, the bottleneck in humanoid robotics may shift away from the factory floor and toward the harder, messier problem of teaching robots to be useful once they arrive.
Source: Figure AI newsroom, “Ramping Figure 03 Production” (April 2026). Production figures are Figure’s own, not independently audited. Credit: Figure AI.