Figure AI hits 350 robots at scale

Humanoid robots are usually announced like concept cars — slick videos, careful demos, vague timelines. But manufacturing is the real test. On April 29, Figure AI said it had crossed 350 Figure 03 robots built at its BotQ factory and ramped output from one robot per day to one per hour in less than 120 days. (figure.ai) That is the news here. Not a new demo — a claim that the company can actually make these things repeatedly. ### What changed this week? Figure’s update was very specific: BotQ has now delivered more than 350 third-generation humanoids, and line throughput improved 24x in under four months. (figure.ai) The company framed that as a manufacturing milestone, not just a robotics one — basically saying the bottleneck is shifting from “can we build a humanoid?” to “how fast can we scale the line?” ### What is BotQ, exactly? BotQ is Figure’s in-house production facility for humanoid robots. When Figure introduced it in March 2025, the company said the first-generation line would be capable of making up to 12,000 humanoids per year. (figure.ai) So the latest announcement is not a surprise in direction, but it is the first concrete proof point that the factory is moving beyond blueprint-stage ambition. ### Why does “one robot per hour” matter? Because robot startups often talk in annual capacity numbers that are easy to float and hard to verify. “One robot per hour” is much more tangible. (figure.ai) It means a line that is behaving like a real production system with repeatable steps, stable parts flow, and fewer manual heroics. The jump from one per day to one per hour is also a 24x throughput gain — fast enough to suggest the line itself is still being actively debugged and improved. ### Is 350 robots actually a lot? For cars, no. For humanoids, yes. (figure.ai) This category is still early, and most companies are closer to prototypes, pilot fleets, or tightly staged customer trials than anything you would call volume manufacturing. A few hundred units is not mass adoption, but it is enough to start building service procedures, fleet software, spare-parts logistics, and the kind of reliability data you only get once many machines are out in the world. That is the unglamorous part — and it is the part that usually kills robotics programs. (figure.ai) ### Where are these robots going? Figure has already shown industrial deployment work with BMW. In a November 2025 update, the company said Figure 02 robots at BMW had run 10-hour weekday shifts, loaded more than 90,000 parts, and contributed to production tied to over 30,000 X3 vehicles. That does not mean the new 350-unit Figure 03 fleet is all deployed at BMW, but it does show Figure has at least one real factory use case to absorb more capable hardware. ### Is this just about hardware? Not really. Figure paired the production news with software progress around whole-body control. (figure.ai) The company said its System 0 policy now uses onboard stereo vision to handle stairs, ramps, and uneven terrain without task-specific programming. That matters because a humanoid is only commercially useful if the hardware line and the autonomy stack improve together. A fast factory that ships brittle robots does not help much. ### What’s the catch? The catch is that almost all of these numbers come from Figure itself. (figure.ai) There is still a big gap between “we can build 350 robots” and “customers can profitably run thousands of them.” Reliability, maintenance cost, task flexibility, and safety in mixed human environments still decide whether this becomes a real market or another robotics hype cycle. But the manufacturing ramp is a serious signal. ### Bottom line? Figure is trying to prove that humanoids are becoming a product category, not a lab project. Hitting 350 robots and one-per-hour output does not settle that argument. (figure.ai) But it moves the conversation from spectacle to operations — and that is where real industries start.