Tuesday 10 November 2026 949 stories on file Full archive
Daily Edition
newscms

Volume III Edition Daily

Humanoid Robots Are Doing Real Work in 2026, but Only at a Handful of Plants

Humanoid robots moved past the demo stage in 2026, but the gap between stage and shop floor is still wide. Figure AI and Agility Robotics hold the clearest operating records among Western programs, while Tesla's Optimus…

Robotics & Drones 1,039 words 5 min read

Humanoid Robots Are Doing Real Work in 2026, but Only at a Handful of Plants — Robotics & Drones No Image Robotics & Drones
Lead image · Filed 7 November 2026, 04:01

Humanoid robots moved past the demo stage in 2026, but the gap between stage and shop floor is still wide. Figure AI and Agility Robotics hold the clearest operating records among Western programs, while Tesla's Optimus remains the least verified large program, according to a Technology.org review published in July 2026. Real industrial humanoid use exists, yet it is concentrated at a small number of named sites.

The review separates company-reported operating data from widely circulated unit claims. Tesla has never published an Optimus production count, and claims that it has passed 50,000 cumulative units, that Figure has 10,000 deployments, or that more than a thousand Optimus robots work on Tesla production lines do not come from the companies involved.

That distinction matters now because investors and factory operators are deciding whether to budget for humanoids in 2027. A robot can stand on a factory floor, collect data, and still not replace labor in a measurable way. Elon Musk said as much on Tesla's Q4 2025 earnings call on January 28, 2026, when he described Optimus as "not in usage in our factories in a material way," with units used primarily for learning rather than productive tasks.

Figure's BMW Pilot Produced Concrete Numbers

Figure's strongest evidence comes from BMW Group Plant Spartanburg. According to Figure's November 2025 report cited by Technology.org, the Figure 02 robot completed an eleven-month pilot, logged more than 1,250 operating hours, loaded more than 90,000 sheet metal parts, and took roughly 1.2 million steps while supporting production of more than 30,000 BMW X3 vehicles.

The robot worked ten-hour shifts, Monday through Friday, placing metal parts into welding fixtures to five-millimetre precision. Figure reported above 99 percent placement accuracy per shift and an 84-second cycle time. That is still not full-scale production, but it is among the more concrete public records of a humanoid doing repetitive work on an active automotive line.

Figure has since retired the 02 units and moved Figure 03 into a logistics sequencing task at the same BMW plant. BMW has also established a Center of Competence for Physical AI in Production and plans to extend humanoid deployment to Plant Leipzig from summer 2026.

Figure 02 humanoid robot at BMW

Agility Has the Broader Footprint

Agility Robotics has a wider reported footprint than Figure. The company says Digit has accumulated more than 65,000 operating hours across nine customer facilities, with GXO, Schaeffler, Toyota Motor Manufacturing Canada, and Mercado Libre named as commercial customers. Agility's RoboFab plant in Salem, Oregon was designed for capacity of up to 10,000 Digit units annually at full output.

The numbers are self-reported, and the review does not independently verify them. Still, naming specific customers and facilities puts Agility ahead of programs that cite only internal pilots.

Tesla Is the Least Verified Large Program

Tesla had targeted 5,000 Optimus units for internal factory use in 2025, according to Technology.org, while reporting by The Information in mid-2025 put actual output at a few hundred, below ten percent of the target. China's April 2025 export restrictions on rare earth magnets and longer-than-planned Gen 3 design finalization, particularly around the dexterous hand, contributed.

On the Q1 2026 earnings call on April 22, Musk said Optimus production at Fremont would begin in late July or August. He warned that initial output would be slow and said the production rate for the year was impossible to predict, citing 10,000 unique parts across a new line. As of mid-July 2026, Fremont Optimus production had not started, and Tesla's Q2 delivery report, published July 2, contained no Optimus numbers. Musk indicated at Davos in January 2026 that public sales were likely by the end of 2027.

The next checkpoint is Tesla's Q2 earnings call on July 22, 2026, when investors should get a clearer read on whether the Fremont Optimus line is beginning production or still preparing for it.

Tesla Optimus robot

Unitree Wins on Volume, Not Margin

Unitree Robotics is the clearest contrast to Western industrial programs. The Hangzhou company shipped about 5,500 humanoid units across its product line in 2025 and is targeting 10,000 to 20,000 units in 2026. Its G1 humanoid starts around $16,000, roughly a tenth of the price level associated with full-size Western platforms.

The company reached a $1.3 billion valuation in June 2025 in a round led by ByteDance, Alibaba, and Tencent, and has won approval for a STAR Market listing, China's technology board IPO route. Volume has not insulated it from margin pressure: Unitree's Q1 2026 net profit fell 52 percent compared with a year earlier, even as humanoid stocks rallied on Optimus procurement orders and Unitree's IPO approval.

What This Means for Buyers

For a plant manager weighing a 2027 budget, the honest answer is that humanoids are a pilot item, not a line item. Figure's BMW numbers are real but narrow, Agility's customer list is the best proxy for scale, and Tesla's program has not yet produced verifiable factory output. Anyone promising a humanoid that pays for itself next quarter is selling a roadmap, not a machine.

There is also the question of what deployment even means. A robot that runs for a few hours under supervision and then needs a human to reset it counts as a pilot, not a worker. Figure's 1,250 hours over eleven months is roughly 3.8 hours per day, which is honest but modest. Agility's 65,000 hours across nine facilities averages out to about 7,200 hours each, but the spread across sites and tasks is unknown.

Buyers should ask for three things before signing: the operating hours per unit per month, the number of interventions per shift, and whether the vendor will guarantee a cycle time in writing. None of the programs covered here publish all three. Until they do, the safest planning assumption is that a humanoid replaces a fraction of a task, not a person.

The Technology.org review is available at Technology.org, and the underlying figures come from company reports and earnings calls rather than independent audits. That caveat should travel with every headline about humanoids for the rest of 2026.

For further context on the robotics market, see our robotics and drones coverage.