Bottom Line: Southeast Europe (SEE) enters the humanoid robotics era with a new project – the Minth Group/AgiBot facility in Šabac, Serbia – but a wider, mostly unconnected set of capabilities sits around it: Croatian vision-AI and autonomous-vehicle engineering, Slovenian motor manufacturing already serving mobile robots, and Bulgarian semiconductor-automation expertise.
None of this adds up to a hub yet. What exists today is a Chinese-designed product being assembled, tested, and (the Serbian government hopes) re-exported as “Made in Serbia” – not a European-owned technology stack.
Whether SEE becomes a genuine embodied-AI manufacturing base for Europe, rather than a lower-cost extension of Chinese supply chains, depends on three things converging over the next three to five years: local actuator and reducer manufacturing, capital that links the region’s scattered capabilities into an actual supply chain, and a geopolitical environment that still allows Chinese-origin hardware built in Serbia to reach EU and Western markets.
What Actually Exists Today
The only dedicated humanoid mass-assembly line in the SEE region is the Šabac, Serbia facility, opened August 29, 2026 as a joint venture between Chinese automotive supplier Minth Group and Chinese robotics firm AgiBot. Phase one is a €20 million, 3,000 m² operation targeting more than 5,000 units a year of humanoids and quadruped “robot dogs,” doing final assembly, motion/vision calibration, and load testing. A planned €200 million expansion – the Inđija Robotics Industrial Park – would push capacity toward 20,000 units a year.
This is real and operating, but it is worth being precise about what it is: contract assembly of a Chinese-engineered product. The core IP – actuators, vision models, control software – comes from AgiBot in Shanghai. Local based contribution is structural integration, wiring, calibration, and testing, using an automotive supplier base (with strong background in China) that already does die-casting, harnessing, and battery enclosures for the auto industry. That’s a legitimate and valuable role, but it is not, on its own, an embodied-AI industry.
The Wider Regional Base: More Than One Project
Zoom out from Šabac and a different picture emerges – not a single assembly line surrounded by empty space, but a potential cross-border pool of talent and adjacent capability, scattered across five or six technology layers a humanoid robot actually needs. None of it was built with humanoids in mind, and none of it is currently connected to the Šabac line commercially. But taken together, one can see potential for Southeast Europe to integrate such know-how into a regional powerhouse.
Vision, perception, and autonomous software. Croatia perhaps is the strongest node here: Gideon Brothers builds AI-driven, 3D-vision autonomous mobile robots and is an NVIDIA Isaac Perceptor development partner – genuine perception-AI IP, not integration work. Rimac’s spinoff Verne runs Europe’s first commercial robotaxi service (launched in Zagreb, April 2026), though it’s worth being precise that the self-driving stack itself is licensed from China’s Pony.ai – Verne’s own expertise is vehicle design, fleet operations, and EU manufacturing, not perception software. In Serbia, RT-RK (Novi Sad) is a genuine embedded-AI and automotive-software house – part of the TTTech Group, which builds safety-critical autonomous-driving compute – giving the region real depth in the software layer that sits between sensors and motion control. Several European powerhouses in this segment are active for years already in the region, such as Aumovio in Serbia and Romania – proving that the talent pool exists and is continuously being developed.
Precision motion and motors. Slovenia carries this layer: Yaskawa Europe’s Kočevje plant is one of Yaskawa’s largest production sites outside Japan for articulated industrial robots, and Domel (Železniki) already manufactures BLDC gear motors specifically for AGVs, AMRs, and unmanned ground vehicles – closer to humanoid-actuator engineering than generic automotive motor production.
Power electronics and batteries. This is a layer that matters for humanoid robots as much as EVs. Serbia could soon host planned ElevenEs’s LFP battery gigafactory in Subotica, though the project’s timeline faces uncertainty following the tragic death of its founder. Other pojects such as ZF Friedrichshafen’s €160 million EV powertrain-component plant in Pančevo. Romania’s Bosch and Continental facilities (Timișoara/Cluj) add ECU and drive-control manufacturing know-how to the pool. None of this is robot-specific, but it’s exactly the power-management and battery-density expertise a humanoid platform needs, sitting a few hundred kilometers from Šabac.
Precision electronics and sensors. Bulgaria’s Plovdiv/Trakia cluster is the deepest node: Milara International is a genuine global leader in semiconductor wafer-handling robotics – cleanroom-grade precision assembly, not generic PCB work – and Sensata Technologies produces the pressure, thermal, and electromechanical sensors that would underpin tactile or force sensing.
Adjacent autonomous hardware. The region’s drone and UAV sector is a useful proof point that SEE can already manufacture complex autonomous hardware at commercial scale: Bulgaria’s Dronamics has built a cargo-drone logistics model now cited among the region’s most-watched startups, and Serbia recently opened a UAV manufacturing joint venture with Israel’s Elbit Systems near Belgrade – mostly defense-oriented, but evidence that autonomous aerial systems, not just ground vehicles, are being built locally.
The honest read: this is a real, broad base – vision AI, motors, batteries, precision electronics, and autonomous hardware all genuinely present in the region – but it is a set of separate national success stories, not a supply chain. Nothing here currently ships a component into Šabac or to neighboring hubs. Instead, it offers a glimpse of what Southeast Europe could achieve with a unified talent pool.
What Has to Happen for This to Become a Hub
Three gaps separate “assembly enclave” from “integrated hub”:
Actuators and reducers. Harmonic strain-wave reducers and frameless BLDC motors are the highest-margin components in a humanoid robot, and none are currently made at humanoid-grade precision in SEE. Domel is the closest existing candidate – it already builds BLDC motors for mobile robots – but scaling to joint-actuator tolerances requires new gear-cutting capability nobody in the region has yet committed capital to probably.
Coordination, not just capability. Vision AI in Croatia, motors in Slovenia, batteries and power electronics in Serbia and Romania, precision sensors in Bulgaria – none of it is currently linked in any way, shape or form.
Talent. The region has strong traditional mechanical and automotive engineering, but a real shortage of systems-mechatronics specialists trained in dynamic bipedal balance and real-time kinematic control – a narrower and newer skill set than industrial robotics or automotive electronics.
The Geopolitical Wildcard
Two developments cut against the “SEE as Europe’s nearshore hub” thesis:
On July 29, 2026, the United States announced a ban on imports of advanced foreign-made robots, a category that explicitly covers the kind of Chinese-designed humanoids and quadrupeds being assembled in Šabac. That complicates one major export market for exactly the product this facility makes.
The Šabac project sits inside a broader, deepening Serbia-China relationship – reporting has linked the robotics investment to wider defense cooperation between Belgrade and Beijing, which Serbian EU-accession watchers flag as a complicating factor for the country’s EU path. A facility whose output is meant to be relabeled “Made in Serbia” for EU export is politically sensitive in a way a purely European-owned project would not be.
Both mean SEE’s near-term humanoid story is entangled with US-China trade policies and geopolitical positioning.
Strategic Outlook
Near-term (1–3 years): SEE functions as a contract assembly and testing node, with some companies picking up isolated component or subsystem contract within new EU development or globally on their own commercial merits.
Medium-term (3–5 years): The realistic path to “hub” status could run through Slovenia towards Bulgaria – provided an OEM is willing to source locally. Absent that demand signal, these remain adjacent industries rather than a humanoid supply chain.
For Policy Makers and Investors: Maximizing regional competitiveness requires:
1. Capital allocation specifically targeted at precision capabilities and humanoid industry in particular, not general industrial incentives.
2. Deliberate supply-chain matchmaking between the region’s existing motor, sensor, and vision-AI firms and humanoid OEMs – European ones, or supplementing Chinese hardware’s route into the EU.
3. Not mistaking contract assembly of foreign IP for the development of native manufacturing capacity.
