The forecast doing the persuading here belongs to Fortune Business Insights, which puts the global humanoid market at 6.6 trillion won in 2025 and 222.4 trillion won in 2034 [12]. That is 33.7 times in nine years, about 48 percent compounded annually with no down year written into it [1], and it prices the narrative rather than measuring the value behind it. Samsung's answer is scaled to that uncertainty: the account carries no capex number, no headcount for the new robotics office and no ship date [18], so what is on the table today is a reporting line, and reporting lines are cheap.
Units make the gap concrete. If the second half matches the first, 2025 ships about 38,200 humanoids, and 6.6 trillion won spread across them prices the average machine near 173 million won [5]; hold that price and the 2034 number needs roughly 1.29 million units a year [6], and since prices fall as volume arrives, the real requirement is higher. Set that against who is shipping now: the first-half total implies about 5,100 units in the same period a year earlier [2], Chinese vendors took more than 97 percent of it [14], which leaves under 575 units in the half for every non-Chinese maker combined [3], and the five named Chinese firms account for 87 percent, roughly 16,600 machines, with the rest of the Chinese share going to companies the survey does not name [4].
The sequencing is what Samsung has actually changed. One CTO holding both the drivetrain and the perception teams signals software taking the lead over hardware [4], and the working purpose is a body drawn around the algorithms from the first design pass rather than a frame with software layered on later, which is how Samsung expects to arrive at commercial readiness in one go [5]. The IP is stacked to match: the SAMSUNG BOT trademark filed in 2022 [6], hip joint geometry for bipedal walking filed in 2024 [7], and language-model control that turns spoken instruction into physical motion [9]. Internalizing everything from design to the brain [2] is also a decision not to license anyone else's stack, and that costs time.
The claimed moat is data. Samsung's own semiconductor, phone and appliance lines are the collection base, with a data factory under construction at Gumi validating what the Umyeon-dong campus develops [10][11]. Against big tech firms working from laboratory simulation, which is the comparison the report itself draws [11], the argument holds. Against thousands of machines already earning hours in other people's facilities it is weaker, because a deployed fleet collects its data on the customer's clock.
Branches worth pricing. The AI-first design compresses the cycle, Samsung lands a factory-integrated humanoid, and its own plants become a captive first customer most rivals lack. Or hands and legs commoditize on the Chinese cost curve, in which case internalizing drivetrains is capital in the low-margin half and the better business was selling brains and silicon into other people's bodies. Or humanoids sit in pilots into the 2030s, and the org chart turns out to be the cheapest option Samsung ever wrote. My weight sits on the second, with the caveat that captive factory demand is a hedge the pure-play robot makers cannot buy. The read breaks the day an outside AI partner appears inside that robotics office, or a Samsung Bot demo runs on someone else's model.
