NEURA bets on robot gyms to anchor Europe's physical AI

German robotics company NEURA is constructing a chain of training centres where robots practise under real-world conditions. The first two flagships are being built in partnership with leading German universities, betting that physical repetition, rather than raw computing power, is the missing ingredient for humanoid machines.

The company has announced a facility with RWTH Aachen University, covering roughly 3,000 square metres at the Hightech Campus Melaten and slated to open by late 2026. Around 20 of the university's institutes will be connected to the site. A second German facility, the TUM RoboGym, will open at Munich Airport later this year alongside the Technical University of Munich, supported by a €17m commitment.

These are the first nodes in a wider network. NEURA is developing ten gyms globally, with five expected to be operational by the end of 2026 across Europe, the United States and China. The facilities combine physical practice with high-fidelity simulation, feeding the resulting data into the Neuraverse, NEURA’s cloud platform for developers and clients.

The strategy targets a fundamental data gap. While software AI models train on trillions of web-scraped words, robots lack a comparable volume of physical experience. Gathering this physical data is inherently slower and more expensive than scraping text. Simulations also fall short of capturing the unpredictability of a real factory floor. “The biggest bottleneck in Physical AI is no longer intelligence, it’s experience,” said founder and chief executive David Reger.

For Europe, the siting of these initial flagships is a deliberate industrial play. RWTH rector Ulrich Rüdiger framed the Aachen centre as “securing Germany’s leading position in the global AI competition.” NEURA is explicitly pitching its network as physical AI built from Europe, offering a sovereignty narrative as the US and China heavily subsidise their own robotics sectors.

NEURA has significant financial backing for this push, having closed a $1.4bn Series C round in June with investors including NVIDIA, Qualcomm, Amazon and Bosch. However, the physical AI race remains highly speculative. Most of the planned gyms are still on paper, and practising in a controlled environment does not guarantee reliable performance on an active production line. Furthermore, the business model relies on industrial partners contributing proprietary operational know-how, even if NEURA insists each partner retains control of its own data. Until humanoids move beyond the demo stage, NEURA's gym network remains a high-stakes bet on who controls the physical foundations of future automation.