1,000 to 2,000 a year: the economics Mali is selling
EAt the end of May 2026, Finance Minister Siniša Mali gave Serbia's robot plan the numbers a market audience actually cares about: how many, how fast, and to do what. His figure — 1,000 to 2,000 humanoid robots a year — is the one that has stuck, and it matches the range used across subsequent coverage.
The output case
- Volume: 1,000–2,000 humanoid robots per year.
- Build time: the first factory to be completed within about six weeks of the announcement.
- Use cases: making coffee and drinks, hotel reception, and defence-industry applications.
- Training: a robot-training centre expected to open by the end of 2026.
A robot does not need sleep or lunch breaks. It works 24 hours a day, seven days a week.
What holds up, what to test
The 24/7 argument is real but partial: uptime is a genuine advantage, yet it says nothing about unit cost, reliability, or whether there is demand for 1,000–2,000 units a year at a price that clears. The use-case list — barista, receptionist, defence — spans wildly different engineering and regulatory bars, and treating them as one market overstates how ready any single one is. The output range is a target, not a shipped result.
The market read
For buyers, the number that matters is not the annual ceiling but the first credible unit shipped and supported. A plant sized for up to 2,000 a year is meaningful capacity for the region if it fills; the open question is demand, not aspiration. That is the metric Rojium will track as the line comes online — actual availability, price and support, against the 1,000–2,000 headline.
Sources
Turn the use case into a machine list.
Describe the site, payload and timeline. Rojium will map suitable hardware and service options.
Robot showcase
Compare the first catalogue machines.
Open the robot showcase to review pilots by vertical, use case and request-only commercial terms.
Related articles
Europe builds its own humanoids: who, for whom and with whose money
NEURA, Humanoid, Hexagon, Wandercraft and others build humanoids in Europe, usually around an industrial customer that is also a parts supplier and an investor. Who is who, and which numbers are still plans.
Boston Dynamics isn't rushing to IPO - what that says about real humanoid timelines
A Hyundai executive told Reuters a 2027 listing for Boston Dynamics "won't be easy" - Atlas isn't deployed at scale, the company is still unprofitable. Unitree and AGIBOT reached public markets within a year of trying. What the gap says about real deployment timelines.
Unitree's stock cools after the IPO rally: what it means for buyers
Unitree's stock hit a fresh post-IPO low on 16 September, then rebounded more than 6% the next day. None of that changes what a European buyer is actually choosing between — Unitree's price list or AGIBOT's.