Robots Atlas>ROBOTS ATLAS
Robotics & Hardware

NEURA Robotics opens NEURA Gym at RWTH Aachen for physical AI

NEURA Robotics opens NEURA Gym at RWTH Aachen for physical AI

NEURA Robotics and RWTH Aachen University are opening the NEURA Gym RWTH Aachen — a physical AI training facility for robots. It is one of ten such 'gyms' planned worldwide. The announcement came on July 24, 2026. The goal is simple: gather the real-world data that robots badly lack today.

Key takeaways

  • NEURA Gym RWTH Aachen will occupy about 3,000 sq m at the Hightech Campus Melaten.
  • It is one of 10 planned gyms across Europe, the U.S. and China; half operational by end of 2026.
  • The facilities combine real-world training with simulation to generate training data.
  • Data is scaled through the cloud-based Neuraverse platform.
  • The gym network is backed by a Series C round of up to $1.4 billion.
  • A second German site is being built at Munich Airport with TUM/MIRMI.
10 facilities for physical AI training, aimed at robotics' biggest gap: the shortage of real-world data.

The problem: robots lack data

Large language models learn from vast amounts of text scraped from the internet. Robots have no such internet. Physical data on how objects slip, bend and resist is vanishingly rare online. That is the bottleneck for the whole field of robot learning.

NEURA Robotics wants to close that gap. Simulation alone is not enough — as the company admits, it cannot reproduce the friction, variability and unpredictability of a real environment. So NEURA Gym combines two sources: physical training of robots on the floor, and high-fidelity simulation that multiplies the data collected.

Robots have access to a fraction of the data that large language models train on.

That is how the company frames the whole effort, as cited by The Robot Report.

How a robot 'gym' works

NEURA Gym RWTH Aachen is roughly 3,000 sq m at the Hightech Campus Melaten. In practice it is a hall where robots perform real tasks while the system collects multimodal data: Data from several sources at once — vision, force, motion, contact — captured in sync as the robot acts. from them — vision, force, motion and contact.

The collected data flows into Neuraverse, an open cloud platform. There it is scaled through simulation, and industrial partners can train and validate robots for specific use cases before deploying them. It is a 'data first, deployment later' model.

Ten gyms and Series C money

Aachen is one piece of a bigger puzzle. NEURA is building ten facilities across Europe, the U.S. and China, with half expected to be operational by the end of 2026. A second German site is going up at Munich Airport in partnership with the Technical University of Munich (TUM) and its MIRMI institute — over 2,300 sq m, testing fleets of cognitive and humanoid robots.

Money backs the scale. The global gym network is supported by a Series C: A startup's third major funding round, usually used to scale an already-working business. round of up to $1.4 billion. Interest has come from manufacturing, automotive, healthcare and industrial automation.

Real training versus simulation alone

Many players lean mostly on simulation because it is cheap and fast. NEURA takes a different route — it treats real-world training as a necessary condition and simulation as a multiplier, not a substitute. That is more expensive, but closer to the physics a robot will actually face.

The European angle is worth noting too. The facilities are being built at German technical universities, and the RWTH rector speaks plainly about securing Germany's position in the global AI race. It is a bet that the edge in robotics will come not from the model itself, but from access to data and the infrastructure that underpins European physical AI.

Why it matters

The shortage of physical data is one of the main limits on learning robots today. Whoever solves the problem of collecting and scaling such data gains an advantage that is hard to copy. Building a dedicated network of 'gyms' is an attempt to turn that bottleneck into an asset.

The choice of partners matters too. Tying the facilities to leading technical universities gives access to talent and research while building European data infrastructure. In a race dominated by the U.S. and China, it signals that Europe is trying to play its own card — not the model, but the data and the place where it is made.

What's next?

  • Half of the ten facilities are meant to be operational by the end of 2026 — the nearest measurable milestone.
  • Bringing the Munich Airport site with TUM/MIRMI online will show whether the model can be replicated beyond Aachen.
  • To watch: whether industrial partners (manufacturing, automotive, healthcare) turn stated interest into real deployments via Neuraverse.

Sources

Share this article