By Hamza Faisal

Today's robots are amnesiacs. They can see, hear, and feel — but they forget everything the moment a task ends. NEURA Robotics, the German company building cognitive robots, wants to fix that, and it has brought in NXAI to do it. The two companies announced on October 20 that they will jointly develop a memory component for cognitive robots, a partnership they frame as a European answer to the physical AI race.

Why robots need to remember

The problem is deceptively simple: a robot that genuinely helps in everyday life and industry must be able to remember, as NEURA founder and CEO David Reger put it. Today's machines treat every situation as if it were the first. They cannot track a task across many steps or act on accumulated experience — the exact opposite of how a human works on a factory floor or in a warehouse.

The companies say two distinct kinds of memory matter: working memory, the short-term state a robot holds while executing a task, and episodic memory, the longer record of what it has already seen and done. Without both, robots stay stuck in demo mode — impressive in a video, helpless in a real building.

xLSTM meets Neuron AI

The partnership has two parts. First, the companies will integrate NEURA's Physical AI stack with xLSTM, NXAI's recurrent neural-network architecture designed to hold and update state efficiently across long sequences. The work builds on Neuron AI, NEURA's multimodal foundation model for physical AI, which already combines 3D vision, touch, audio, robot state, and language with a memory function.

The promise is speed and efficiency: xLSTM is built to maintain state over long sequences without the heavy compute costs of alternatives, which matters when a robot must keep track of a multi-step task in real time. "Memory and experience are what allow robots to solve tasks reliably across many steps," said Prof. Dr. Sepp Hochreiter, founder of NXAI — the researcher who co-invented the original LSTM architecture that xLSTM extends.

A second model on the Neuraverse

Second, NXAI plans to make its own xLSTM-based model available on the Neuraverse, NEURA's platform where third-party AI approaches can reach real robots. It will sit alongside Neuron AI, giving robot builders a choice of models — and, the companies say, the ability to customize the NXAI model for specific tasks.

The announcement leans into geography as much as technology. "Two European strengths, one goal: physical AI from Europe, together for the world," Reger said — NEURA is headquartered in Metzingen, Germany, where it was founded in 2019, while NXAI is based in Austria. The framing echoes the theme of the moment: the next wave of AI is physical, and Europe is determined not to watch it happen elsewhere.

The second half of the deal is still a plan rather than a finished product — NXAI's model is not yet live on the Neuraverse. But the first half is concrete engineering: a memory layer that could let cognitive robots do what today's machines cannot — remember yesterday's job while doing today's.

By Hamza Faisal