Minth and AgiBot Open Europe's First Humanoid Robot Plant, Targeting 20,000 a Year

In this storyAgiBotMMinth Group

Serbia opened what its government calls Europe's first mass-production base for humanoid robots on Saturday, in the western city of Šabac. The plant is operated by Chinese auto-parts manufacturer Minth Group in partnership with Shanghai robotics company AgiBot.

President Aleksandar Vučić and Minth founder Qin Ronghua attended the opening. Production runs out of Minth's existing Metal Parts Majur site in Šabac, and the first robot assembled there was given a serial number during the ceremony.

The first phase carries an investment of €20 million. A separate Robotics Industrial Park planned for Inđija in northern Serbia takes total planned investment to €200 million, according to a government statement. Once the full project is built out, stated annual capacity is up to 20,000 humanoid robots and robot dogs aimed at European and global markets.

Those are completion figures, not current output. Vučić said he expected more than 5,000 robots a year to be assembled at the facility initially, with around 200 people working directly in the new division at the start.

The plant is the manufacturing end of a partnership that has been building for over a year. Minth and AgiBot signed a three-year strategic cooperation agreement in March 2025, and in March 2026 the two launched AgiBot's full robot portfolio in Germany, showing the compact X2 humanoid to roughly 400 engineers and executives in Munich. Minth acts as AgiBot's European sales agent and localisation partner.

Minth brings 15 years in the European market, more than 40% share of regional aluminium trim and battery-housing structural components, and production sites in Germany, France, the UK, Czechia, Poland and Serbia. AgiBot, founded in 2023, says it had delivered more than 5,000 humanoid robots as of early 2026.

The pairing is unusual in a useful way. AgiBot supplies the embodied-AI stack and the hardware design; Minth supplies something a robotics startup cannot buy quickly — 78 factories and offices, a European industrial footprint built over 15 years, and customers who already sign purchase orders with it for structural parts. Localised assembly also softens the tariff and procurement-policy exposure that Chinese hardware faces in Europe.

The second half of the deal matters more than the factory. Minth's own plants serve as training grounds and data-collection sites for AgiBot's robots. Part grasping, high-precision assembly, logistics scheduling and equipment monitoring on live lines produce the messy, variable data that lab environments cannot fake — changing light, inconsistent material surfaces, people moving through the workspace. Whoever accumulates the most of that data compounds fastest.

Where it is thinner: 20,000 units a year is a completion-stage figure attached to a €200 million industrial park that has not been built. The number in circulation describes intent, not installed capacity. Phase one is €20 million and roughly 200 direct jobs, and even the 5,000-unit initial figure was framed as a hope rather than a committed run rate.

Demand is the unresolved variable. Automakers and suppliers are still running pilots rather than placing volume orders, and a humanoid has to beat purpose-built automation on cost per task before a plant manager signs. Capacity is the easy half of this problem.

The strategic read is that Chinese robotics has stopped treating Europe as an export market and started treating it as a production base. That is a different competitive posture. Exporting invites tariffs, content rules and political friction; assembling locally with an incumbent European supplier as the front end largely sidesteps them. Serbia, outside the EU but inside its industrial supply chains and offering low-cost skilled manufacturing labour, is a sensible first node for that strategy.

The moat here is distribution and data, not the robot. Humanoid hardware is converging fast and the specification sheets are starting to look alike. What is hard to replicate is a partner with existing lines, existing quality systems and existing customer relationships across six European countries — plus the right to run robots inside those lines every day. Western humanoid developers largely lack that; they have to sell into factories rather than already live in them.

For Indian readers the relevant contrast is structural. India has a growing robotics cohort and a manufacturing base actively courting global capital, but nothing resembling this pairing — a domestic component major that is simultaneously a robotics distributor, a training environment and a contract assembler. The Minth model suggests the leverage point for Indian robotics is not another lab-stage humanoid, but a tie-up with an established auto-component or electronics manufacturer that owns the shop floor and the customer.

The risks are conventional and real. Announced capacity in robotics has consistently outrun delivered volume across the industry. The €200 million park is a plan. Geopolitical exposure works both ways, and European scrutiny of Chinese-controlled industrial assets could tighten regardless of where final assembly sits. Morgan Stanley's forecast of roughly 1.5 million units of annual humanoid demand by 2035 is the number underwriting projects like this — and it is a forecast, not an order book. Watch shipped units and named industrial customers, not factory capacity.

For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.