Eight hours. That’s a normal shift for a factory worker, and it’s the number Xiaomi wanted everyone to see this week. On November 3, the company posted a video of a Xiaomi humanoid robot working a full eight-hour shift in its car factory in China. Just car parts, storage racks and a clock ticking in the corner.
It’s a neat clip. But is it a turning point, or just very good marketing? Let’s dig in.
So what did Xiaomi actually show?
Xiaomi Tech shared the video on X while the company picked up awards at the World Internet Conference in Wuzhen. The setting is the Center Console Side-Cover Sorting Workstation at the Xiaomi Auto Factory. In plain English, the robots pull side covers for a car’s center console out of bins, carry them to storage racks and slot them into place.
The footage runs at up to 100 times normal speed, with a clock counting the working time. According to Electrek, one segment shows two legged robots plus a wheeled humanoid upper body, while small autonomous mobile robots shuttle the bins around. Interesting Engineering describes two humanoids doing the work. Either way, it’s a team effort.
Xiaomi’s big claim is about autonomy. The company says the robots “can independently detect and correct anomalies while coordinating with the factory’s production line system.” It also says the robot covered the station’s full day of production needs, and it called the test “another step toward real-world industrial deployment.”
Why an eight-hour shift is a big deal
To see why, rewind to March. Back then, Xiaomi put a humanoid to work as an “intern” at its Beijing EV factory, installing self-tapping nuts. It hit a 90.2% success rate over three hours. That was fast enough for the station’s 76-second cycle time, which is the time each unit gets at that spot on the line.
By July, Electrek reports, the success rate had climbed to 98%. Xiaomi also added two harder stations, including this side-cover sorting job. So in roughly eight months, Xiaomi’s humanoid work went from a three-hour trial to a full shift.
That matters because factories think in shifts, not demos. A robot that works for twenty minutes is a science project. One that lasts a whole shift starts to look like a coworker. And Xiaomi isn’t thinking small. Last December, CEO Lei Jun said the company plans to “deploy humanoid robots in large numbers across its factories within five years.”
What Xiaomi didn’t tell us
Here’s where we put our skeptic hat on. Xiaomi hasn’t said how many parts the robots sorted. It hasn’t shared a success rate for the eight hours, either. We also don’t know how often a human had to step in, or how the robots stayed powered. Battery swaps? A cable? Charging breaks? No idea.
As Interesting Engineering put it, “working duration alone does not establish production readiness.” That’s a fair point, since in theory a robot could run for eight hours and still fumble plenty of parts.
There’s history here, too. At Tesla’s 2024 “We, Robot” event, humans remotely controlled the Optimus robots, and Tesla didn’t disclose that at the time, Electrek notes. So a healthy dose of doubt is normal now. Still, Xiaomi deserves some credit. It has published success rates and cycle times before, which Electrek says Tesla hasn’t done for Optimus’s useful work. We looked at the open questions around Tesla Optimus production recently, and they sound familiar.
Wheels, legs and the brain behind the Xiaomi humanoid robot
One detail we love is that wheeled robot. Electrek’s Fred Lambert argues wheels beat legs on a flat factory floor. That’s because wheeled bases are more stable, can carry bigger batteries and need fewer costly actuators. Actuators, by the way, are the motors that make a robot’s joints move. Here’s our explainer on robot actuators if you’re curious.
Legs still make sense in homes with stairs. But factories grew up around carts and forklifts, not staircases. So a robot on wheels with good hands and arms might be the practical sweet spot. In fact, Lambert says the real value sits in the hands, arms and AI.
Then there’s the software side. Xiaomi has an open-source AI model called Xiaomi-Robotics-1, pre-trained on 100,000 hours of real-world robot data. Xiaomi has also started trial runs of the model in staff apartments. In one 30-minute run, it completed more than 50 task chains across three rooms.
Okay, but should you care about a Xiaomi humanoid robot?
Honestly, yes, even if you’ll never set foot in a car plant. Sorting parts is dull, repetitive work, and that’s the kind of job most people are happy to hand over. We’ve seen a similar move with UBTECH’s humanoids in Volkswagen logistics. The Xiaomi humanoid robot fits right into that pattern.
For countries like Armenia, there’s another angle. For example, lower labour costs often help emerging markets attract factories. If humanoids get good enough to handle whole shifts, that edge could shrink over time. On the flip side, smaller manufacturers might one day rent robots to fill hard-to-staff roles.
But let’s not get ahead of ourselves. Xiaomi hasn’t shared the numbers that would prove this is ready for prime time. Our take: it’s one of the more convincing humanoid demos so far, with a big asterisk. So here’s the real question. When Xiaomi finally publishes a success rate for a full shift, will it look like 98%, or something a lot less flattering?