While Tesla audits Chinese suppliers for Optimus, Figure AI stands out in September 2026 as the humanoid with the most visible factory ramp: its BotQ plant has sustained rates on the order of one Figure 03 robot per hour after surpassing 350 units in April, with trade-press reports placing the 1,000th-robot milestone over the course of the year. That is throughput, not a keynote.
In parallel, BMW Group advanced Physical AI in production with a Figure 03 project in Spartanburg, USA. The stated focus is logistics: sequencing unsorted parts and just-in-sequence flows. BMW has not published deployment quantities or plant-performance metrics—a typical caution while humanoids still share floors with human operators and AGVs.
The combination of owned factory plus automotive customer anchors Figure in the early industrial segment. Unlike home demos, logistics sequencing has measurable ROI: less walking time, fewer picking errors, more resilience to absenteeism. If Figure 03 sustains useful cycles without constant intervention, the buy argument becomes operational, not aspirational.
Figure’s stack pairs hardware redesigned for manufacturability with policy models trained for manipulation. The bottleneck is no longer only walking without falling; it is robust grasp, replanning when parts are out of pose, and safe handoff with humans. Spartanburg is the test bed Wall Street understands.
Versus Optimus, Figure wins today on production evidence and an OEM customer logo. Tesla wins on potential automotive-chain scale and brand capital. The 2027 race will be decided by cost per useful work-hour, not by followers on X.
Risks for Figure include actuator yield and cost, dependence on a still-limited number of pilot sites, and the familiar demo-to-deploy gap if Spartanburg remains an endless trial. Competition from other Chinese and U.S. humanoids promising lower prices with less software sophistication also looms.
For unions at BMW plants and rivals, F.03’s arrival raises bargaining over assisted versus replaced roles. BMW’s messaging emphasizes logistics assistance, not replacement of skilled welders or assemblers—a political line as important as the assembly line.
Neural Pulse sees BotQ as a sector narrative shift: from viral videos to units per hour. That industrial KPI is what was missing for manufacturing CFOs to open serious 2027 budget lines.
The implicit pitch to investors is that Figure can learn from real fleets faster than labs that barely build dozens of units. Each robot at Spartanburg is a generator of physical-contact data no simulator fully replaces.
Even so, without public metrics on uptime, interventions per hour and task defects, the market should keep disciplined skepticism. A humanoid “on the plant floor” can mean a test cage or flow integration—those are different worlds.
In short, through September 2026 Figure AI combines sustained Figure 03 production at BotQ with logistics trials at BMW Spartanburg, positioning itself as the early industrial reference against Tesla Optimus’s still-preparing volume push.
We will track verifiable production figures, any expansion beyond Spartanburg and productivity data BMW or Figure choose to share. Until then, Figure’s lead is evidentiary, not a coronation.
The lesson for the rest of the physical-AI ecosystem is clear: the model matters, but the factory sets the pace of real-world learning.
BotQ at one robot per hour is still modest versus classic automotive—and revolutionary versus where the humanoid sector stood eighteen months ago.
