· Industry· 日本語で読む
Figure AI livestreams humanoid robots working 200 hours at logistics warehouse; conducts 10-hour comparison with humans
TMTPost reported that three Figure AI humanoid robots performed work in a logistics warehouse for 200 hours during a livestream. The task involved turning packages over so their shipping labels faced down. According to the report, humans narrowly outperformed the robots in a 10-hour sorting comparison.
What happened
In an article published May 25, 2026, TMTPost reported that three Figure AI humanoid robots livestreamed 200 hours of work in a logistics warehouse, turning packages over so their shipping labels faced down. The latter part of the livestream included a 10-hour sorting comparison with human trainees, in which humans narrowly outperformed the robots, according to the report. The article also said the robots resumed work after a brief system interruption.
Why it matters
The Figure AI case showed package-turning work in a warehouse through a 200-hour livestream and a 10-hour comparison with humans. However, the reported result was limited to a narrow human lead. Because the article did not provide figures such as the number of items processed, the result should be distinguished from data that would allow a quantitative assessment of on-site productivity.
Analysis
The 200-hour livestream provided a view of work over a longer period than a one-off demonstration, but that alone is not enough to determine the stability of continuous operation or practical performance. The 10-hour comparison with humans reported in the article also concerned a specific warehouse task and does not establish a general difference in sorting capabilities. The reported work was limited to turning packages over so their shipping labels faced down. Without figures such as work speed, the number of items processed, total downtime and success rate, it is difficult to interpret the result that humans narrowly outperformed the robots as a quantitative productivity difference.
The article also cited automotive and electronics factories as sites where Chinese companies are deploying robots, contrasting them with warehouse trials. However, package handling in a warehouse and work on a factory floor involve different objects and process conditions. Rather than comparing which type of site robots have entered, it is necessary to check processing capacity and operational records, including downtime, under the conditions of each task. We have also reported on [AgiBot showcasing operational cases at the China International Import Expo; reported cumulative effective intelligent operating hours exceed 62,000](/en/news/agibot-showcases-operational-cases-at-import-expo-reported-effective-i-bfc739), which covered cumulative operating hours while noting the need to distinguish an individual unit’s operating time from the scope and calculation conditions of a cumulative figure. The livestream duration in this case and the cumulative hours in that article are different kinds of measures and cannot be directly compared.
The report supports the conclusion that Figure AI publicly demonstrated a specific warehouse task over an extended period and conducted a comparison with humans. Linking factory deployment records to warehouse task results requires common evaluation conditions, including the volume processed in each task and operational status that accounts for downtime. Points to verify include the number of items processed in the reported 10-hour comparison, how work continued across the full 200-hour livestream, and how demonstration conditions differ from those in actual operations. Without this information, the length of the livestream cannot be treated as evidence of operating capability in a mass-production setting.
Sources
Translated from the Japanese original: /news/karada-beichuu-robotto-wo-hikaku-suru-ronkou-wo-keisai-c01b4a. The Japanese version is authoritative if they differ.