Figure published a production update this week disclosing that BotQ, the dedicated humanoid manufacturing facility it announced last year, has reached a one-Figure-03-per-hour cycle time on its first production line. According to the company, BotQ has gone from one robot per day to one robot per hour in roughly 120 days, which is a 24x throughput improvement, and has cumulatively delivered over 350 Figure 03 units against pilot deployments at customer sites. The first-line nameplate capacity is reportedly 12,000 robots per year, scaling toward a four-year total of 100,000 across additional lines.
The first frame on these numbers is the comparison to where humanoids actually are. A modern automotive assembly plant builds something like 30 to 60 finished vehicles per hour per line. One humanoid robot per hour is, in absolute terms, still slow, and the cycle time reflects the fact that Figure 03 has a meaningfully higher parts count than the prior generation, including a redesigned actuator stack and the new tactile-sensing hand assembly that drives most of the manipulation performance. Figure’s framing of the milestone as “the throughput needed for mass production” is accurate in a specific sense: it is the throughput needed for a single line to clear five-figure annual volume, which is the threshold the company has to hit to justify the next round of capex on lines two through n. Beyond that threshold the curve gets steeper, because most of the cycle-time gains so far have come from process engineering on a fixed BOM, and the next 10x has to come from either parallelization or design-for-manufacturing changes that ripple back into the robot itself.
The second frame is competitive. Tesla is converting the Fremont Model S line to Optimus Gen 3 with a stated target of a million units per year out of a future Texas plant, on a timeline that even Tesla has had to walk back at the last two shareholder meetings. Boston Dynamics is shipping the first 2026 electric Atlas units to Hyundai and DeepMind on a smaller production base. Agility Digit is in Toyota Canada at single-digit unit counts. Unitree is reportedly past 5,500 cumulative units shipped, mostly into research and smaller industrial pilots. Figure is alone in the cohort that has both a paid commercial deployment at scale, currently with BMW, and a stated, dated, and physically observable production ramp. Of the public humanoid roadmaps you can actually evaluate against, Figure’s is the one closest to behaving like a normal manufacturing program.
The honest question is whether the demand curve catches the supply curve. Figure has been signaling residential deployments for Figure 03 as part of the long-term roadmap, which is the kind of statement that has to clear a regulatory and liability bar that none of the current OEMs have actually cleared. The near-term volume case is still industrial: automotive, logistics, fulfillment, a small amount of cleanroom semiconductor work. If 100,000 humanoids over four years lands in that mix, it is a meaningful labor-market story for two or three large industrial customers and a footnote for everyone else. If it lands in homes, the story is entirely different and so is the regulatory bill. We will know which one Figure is actually planning for when the next round of customer announcements posts, presumably with the next Q3 production update.