Competitiveness

'Humanoid robots, the credibility game in Europe'

William Shi, European CEO of Agibot, the Chinese robot manufacturer, speaks: in the Old Continent 'we are thinking about adaptive models'

by Marco Gervasi

9 April 2026

3' min read

Translated by AI

Versione italiana

Key points

Ten Thousand Robots
European benchmark value
Data Governance
Shi: "In Europe we think about adaptive models

3' min read

Translated by AI

Versione italiana

China has not won the race for humanoid robots: it has rewritten it. With thousands of units employed and an unrivalled manufacturing capacity, Beijing now controls the physical component of a sector undergoing great transformation. Explaining what this means for European industry is William Shi, European CEO of Agibot.

Ten Thousand Robots

William leads Agibot's European expansion from its Milan office, chosen as a strategic base for entry into the continental market. Agibot recently announced its 10,000th humanoid robot, becoming one of the first companies in the world to reach this threshold on an industrial scale. It is a trajectory that has convinced some of the most important Chinese and global investors - including Tencent Holdings, HongShan Capital Group (formerly Sequoia China), C-Capital and Lg Electronics - to bet on embodied robotics. The company is targeting a valuation of between USD 5.1 billion and USD 6.4 billion with a listing on the Hong Kong Stock Exchange planned for this year.

The value of the European benchmark

William's perspective is that of someone who knows both sides: the technology from the inside and the expectations - and resistance - of the European customer. "Europe is a demanding market, but it is also the place where companies that know how to conform build global credibility," he notes. "Those who exceed the European benchmark have gained great credibility."

The revolution brought by humanoid robots is not just about their capabilities, but about their brains. In Shanghai, at Agibot's headquarters, robots are 'educated' before being delivered to customers. The space dedicated to their training - already three thousand square metres inadequate - is a controlled environment in which the robots learn specific tasks. Every gesture is recorded, labelled, transformed into input for the Go One foundational model. "It is not a research lab," William points out. "It is already aimed at production."

Data Governance

Agibot's architecture reveals much about how China thinks about robotics. The robot is structured in three layers: the Cerebrum, the cloud layer for interactive intelligence; the Cerebellum, on-device processing for locomotion and manipulation control; and the Ontology, the physical body. This tripartite architecture is not just an engineering choice - it has direct implications for data governance that those operating in Europe must be able to read: the division between local and cloud computing determines where operational data is processed, with what latency, and - a non-trivial question for a European operator - under what jurisdiction.

The Chinese advantage in hardware is not limited to Agibot. China now accounts for around 80 per cent of global humanoid robot production and leads in patent filings. This position is no coincidence: the 15th Five-Year Plan (2025-2030) explicitly names 'embodied intelligence' among the six industries of the future to be developed. The State of Hubei has earmarked 10 billion yuan for the sector; Wuhan has opened a 200 million yuan robot training facility. The model is that of a coordinated industrial policy: the data generated in the Robot Academies are shared, benefiting the entire industry. For those used to analysing comparative AI governance, this is not simply a competitive advantage: it is a digital model that has no equivalent in other systems.

Shi: "In Europe we think of adaptive models

Shi frames this strategy precisely: 'In China, training takes place through Robot Academies. It is a model that works because we have the scale, the infrastructure and the coordination needed. In Europe we have to think about adaptive models: the data generated in Europe stays here, and the local training will feed a local ecosystem'.

On the application sectors, Shi is explicit in identifying the priorities for the European market: 'The events and trade fair sector are the two natural outlets. The planned sequence - first controlled environments such as trade fairs and events, then a progressive industrial environment - follows the same step-by-step logic that has characterised Chinese expansion: building operational reliability before tackling the complexity of a real production line.

On the regulatory front, Shi does not see it as an obstacle: 'Robots will only be placed in an industrial context when all the relevant certifications and licences have been obtained and, above all, the right trust in the machines has been established'.

For a European company considering the adoption of humanoid robots, the questions that matter are not yet the technical ones. They are governance questions: who controls the data generated in production, where is it processed, under which jurisdiction, and with what right of access by the supplier? These are questions that standard contracts do not yet cover, and which require a cross-industry, legal and geopolitical reading together.

That this awareness is already present in Agibot's management is, in itself, a strategic signal.