Generative Bionics plans Europe's largest physical AI data factory with AMD

Three sites, starting in Italy

Generative Bionics presented the plan at CES Unveiled in Milan on 21 October 2026. The first hub, planned in Italy, is due to go into operation by the fourth quarter of 2027. A second facility at another European location and a third site serving global markets are to follow.

According to the company, the three facilities together will recreate real industrial environments, generate high-precision physical data and link that data to large-scale digital twin simulations. The 5,000 square metre Italian hub is meant to become Europe's largest training facility for humanoid robot physical AI. It will replicate selected environments together with industry partners and let fleets of GENE.01 robots perform tasks continuously under real conditions.

Why humanoids need their own data

The company draws a comparison with language models. Large language models learn from vast amounts of information available online, but no comparable data source exists to show humanoid robots how to behave in the real world. The data factory is meant to create one, by collecting data while GENE.01 carries out industrial tasks and extending that experience through digital twin simulations.

Generative Bionics says industrial humanoids must learn from the conditions they will face at work. That requires data covering contact, force, movement, proximity, temperature, task constraints and human interaction. For GENE.01, the dataset will also include signals captured by its multimodal whole-body skin, alongside vision and motion data.

The process is described as a cycle from reality to simulation and back. Every physical environment and robot configuration will also be rebuilt as a digital twin. The same tasks can then be repeated across thousands of variations in layout, objects, materials, operating conditions and failure modes that could not be recreated physically. Models trained and tested in simulation return to the robots for validation, while data from real operation feeds the next training cycle.

AMD as infrastructure partner

AMD will serve as Generative Bionics' AI infrastructure partner, and the network will be powered by AMD Helios. The company describes the rack-scale architecture as the compute foundation for processing physical and synthetic data, training physical AI models and running thousands of digital twin scenario variants in parallel. According to the release, AMD Ventures invested in Generative Bionics' seed round, and AMD compute is on board GENE.01.

Generative Bionics plans to bring robotics, AI and data processing teams together with the infrastructure needed to manage the full chain from raw multimodal robot data to AI-ready datasets, validated models and deployment on robot fleets.

Daniele Pucci, CEO of Generative Bionics, said: "Wir bilden die Welt physisch und digital nach, um humanoide Roboter in großem Maßstab zu trainieren, bevor sie jemals in sie eintreten." Keith Strier, SVP, Global AI Markets at AMD, said AMD Helios will power "Italiens erste physische KI-Datenfabrik".

What it means for industrial customers

The infrastructure is intended to train World Action Models and to validate complete robot configurations before customer deployment, covering perception, control, safety behaviour and task execution. As fleets accumulate operating hours, operational data is meant to improve performance further.

For industrial customers, the stated benefit is that GENE.01 can be configured and validated for a specific environment before it arrives on site. The company's broader goal is to move humanoid robotics from laboratory demonstrations to reliable industrial use, and to strengthen European sovereignty in humanoid robots. The release gives no investment figures or customer names, and the timeline names only the first hub's 2027 start.