Figure robots run five months on BMW line

Humanoid robots are finally getting judged like factory equipment instead of internet clips. That is the real story here. The viral version says Figure proved a robot could work for five months on a BMW line. The fuller version is stronger — Figure and BMW say the deployment at BMW’s Spartanburg plant stretched to 11 months, with two Figure 02 robots working on a live assembly line and contributing to real vehicle production. ### What actually happened at BMW? Figure’s November 19, 2025 update said its Figure 02 robots were deployed at BMW Group Plant Spartanburg after months of setup, then ran on an active assembly line every working day. BMW had first disclosed the partnership in August 2024 as a trial in Spartanburg, where the robot handled sheet-metal parts used in chassis assembly. So the “five months” line was basically an earlier milestone inside a longer rollout. (figure.ai) ### What were the robots doing? This was not general-purpose magic. It was a classic industrial pick-and-place task — grabbing sheet-metal parts from racks or bins and loading them into welding fixtures. That sounds simple, but it is exactly the kind of repetitive, awkward, ergonomically annoying work factories want to automate if they can. BMW framed the whole exercise around taking tiring tasks off human workers while testing where humanoids actually fit in production. (figure.ai) ### Why is the 11-month number more important? Because endurance is the whole question. A humanoid that walks around a demo floor is one thing. A humanoid that shows up every workday, inside a plant, around real cycle times and safety constraints, is another. Figure says the BMW deployment logged 1,250 operating hours, handled more than 90,000 parts, and contributed to production of more than 30,000 BMW X3 vehicles. Those are narrow metrics, but they are factory metrics — and that is the point. (assemblymag.com) ### Does this mean humanoids are ready for every factory job? No — not close. The BMW task was tightly scoped, repetitive, and engineered around one workflow. That is still a win. Factories do not need a robot that can do everything on day one. They need one that can do one annoying job reliably enough to justify the trouble. Turns out that is how most automation spreads — first a bounded task, then a neighboring one, then a bigger footprint if uptime and economics hold. The available BMW and Figure disclosures support that narrower interpretation, not a full humanoid takeover story. (figure.ai) ### Why are automakers early adopters? Car plants already live inside structured environments with repetitive material handling, strict process control, and expensive labor bottlenecks. That makes them a natural test bed. BMW has now expanded humanoid testing beyond Spartanburg — it introduced a pilot at Plant Leipzig in 2026, its first use of humanoids in a German facility. That suggests the company sees enough promise to keep experimenting, even if the technology is still early. (figure.ai) ### What is happening around the rest of the market? The broader humanoid race is now about manufacturing capacity as much as software. Figure has been talking openly about scaling production, and 1X said on April 30, 2026 that its Hayward, California NEO Factory had begun full-scale production, with stated capacity of 10,000 units per year. Separate market forecasts from TrendForce project global humanoid shipments could exceed 50,000 units in 2026. Forecasts are not deployments, but they show how fast expectations are rising. (bmwgroup.com) ### So what is the bottom line? The interesting part is not that a humanoid lasted five months. It is that Figure and BMW turned a flashy claim into something more grounded — an 11-month, measured factory deployment with real throughput numbers. That still does not prove humanoids are broadly economical. But it does move the category out of pure concept mode and into the much harsher world of uptime, task fit, and production math. (figure.ai 1) (figure.ai 2)