German scale-up Neura Robotics, which announced a Series C round of up to $1.4 billion in June, making it Europe’s most heavily funded humanoid and cognitive robotics company to date, is aiming to become a global champion. In October, it launched its own robot foundation model (Neuron AI), a fourth-generation humanoid (4NE1 Gen 4), and a new training gym for robots in collaboration with the Technical University of Munich, with plans to open others in the U.S. and China. Its strategic backers and partners include Nvidia, Amazon, Qualcomm, Bosch, Schaeffler, and the European Investment Bank.
“Europe’s industrial knowledge gives us a real opportunity to lead the next era,” Neura Robotics CEO David Reger said in written responses to questions from The Innovator. “We have to act on it.”
Humanoid robots are scaling fast, from a tiny base. Factories installed about 603,000 conventional industrial robots in 2025 (up 11%), taking the global installed base past 5 million, according to the International Federation of Robotics’ World Robotics 2026 report. Humanoids are a rounding error by comparison: around 17,000 shipped in 2025. But growth is steep. According to the Smart Analytics Global (SAG) Robotics 360 Report, global humanoid robot shipments reached approximately 19,100 units in the first half of 2026, up 272% year on year. It forecasts close to 60,000 for the full year, with industry revenue of about $1.6 billion and expects shipments to reach 500,000 units a year by 2030. Industrial and commercial uses accounted for more than 70% of shipments, up from about 50% a year earlier.
Globally, robotics companies had raised $55.8 billion in 2026 by early June, according to Dealroom. Valuations rest largely on company-reported metrics and conditional orders. The next six to 12 months will be a proof phase, showing whether pilots turn into repeat paid orders and whether costs fall enough to compete with fixed automation and human labor.
Neura’s Pitch
Neura’s pitch is that the binding constraint on useful robots is no longer hardware but real-world training data. Its strategy has three parts. First, a range of robot “embodiments” (arms, mobile platforms, humanoids) sharing one software layer. Second, the Neuraverse, a cloud platform for deploying and updating skills across fleets, which also lets customers fine-tune models on their own tasks. Third, a network of “Neura Gyms”: physical centers where robots learn real industrial tasks, combined with simulation and human demonstrations captured in sensor suits. The TUM RoboGym near Munich Airport, opened October 22, covers 2,300 square meters, involves roughly €17 million of investment over three years (about €11 million from Neura), and already hosts companies from more than 10 industries. Neura says further gyms are planned in Europe, the U.S., and China.
The newest piece is Neuron AI, launched on October 14. Neura describes it as a multimodal foundation model that fuses 3D vision, touch and force, audio, robot state, and language, and adds memory so a robot keeps track of objects and the consequences of its actions when they are out of view. The company calls the architecture “Perception-Language-Action,” a deliberate contrast with the vision-language-action models used by most rivals. It says one model runs across five Neura robot types and that the key metric is “Time-to-Skill”: how quickly a robot can learn a new task and perform it reliably. At Nvidia GTC Berlin, which took place October 20–22, it showed 4NE1 running an 11-minute laundry sequence, from sorting to loading the dryer, without human intervention.
“Physical work demands an understanding of what happens when you touch and move things,” says Reger. “You hear when a component clicks into place. You feel when an object starts slipping. Neuron AI is our Perception-Language-Action foundation model, designed to combine vision, hearing, touch, language and memory across different robotic bodies.” The German scale-up trains Neuron AI on real robot experience from the Neura Gyms, human demonstrations, and simulation. “Building it ourselves lets us develop the intelligence and the robot together,” Reger says. “We judge progress by reliable task performance and time-to-skill: how quickly a robot learns something useful. That is where any advantage must prove itself.”
Building The Ecosystem
Neura has been buying established hardware businesses that come with products, customers, and manufacturing capacity:
- ek robotics (October 2025): automated guided vehicle maker, renamed Neura Mobile Robots with the ek brand retained.
- Huber Automotive development division (October 2025): control units, battery management, and energy storage, from an insolvent automotive supplier.
- ACTIVE Shuttle from Bosch Rexroth (August 2026): an autonomous intralogistics platform.
- ADLATUS Robotics (August 2026): professional cleaning robots.
Key partnerships include Schaeffler, Bosch, Qualcomm, AWS, Dassault Systèmes, Kawasaki (whose CL cobot series runs on Neura’s platform), SECO (September 2026), Austria’s NXAI (October 2026), and Gyms with RWTH Aachen and TUM. The Schaeffler partnership, announced in November 2025, covers joint development, actuator supply, and deployment of Neura humanoids in Schaeffler plants. German press reported a framework order worth about €300 million, and Schaeffler, which makes high-precision components, bearings, and complete system solutions for the automotive, aerospace, and industrial sectors. has talked of a mid-four-digit number of humanoids in its plants by 2035. Schaeffler is also spreading its bets, however, by investing in two other humanoid suppliers: Neura rivals Humanoid and Agility Robotics. Schaeffler recently extended its humanoid robot ecosystem by establishing a technology partnership with Swiss-based Hexagon Robotics to supply components and said it is planning to deploy at least one thousand humanoid robots from Hexagon Robotics in its global production system within the next seven years.
Integrating Humanoid Robots Into The Workplace
Challenges in integrating humanoid robots into the workplace include safety, getting labor unions on board, and convincing companies to share their training data.
Neura has not yet announced a date for unrestricted operation of its 4NE1 humanoid robot alongside employees. “Safety has to be demonstrated for the specific task and environment, including the tools and objects the robot handles,” says Reger. “Our initial deployments take place in controlled environments with partners, and we expand as the relevant assessments and validation are completed.”
Reger says Neura is making the case for humanoids to works councils and unions by starting with the people doing the work and involving their representatives early. The German scale-up is asking workers which tasks wear people down, where positions are going unfilled, and what training employees will need. “Demographic change makes those questions urgent,” Reger says. “Experienced workers are retiring, often taking decades of practical knowledge with them. Robotics can help preserve that expertise and relieve physical strain. People should have a voice in how the technology enters their workplace and a clear opportunity to benefit from it. Acceptance has to be earned through experience.”
Neura also needs to convince companies, including Germany’s Mittelstand, to share training data. “The right partnership makes a company’s expertise more valuable while keeping it under that company’s control,” Reger says. “Partners bring their processes and knowledge; we provide the robots, training infrastructure and platform. They decide which skills remain private and which they want to license to others.”
Private training carries additional infrastructure and computing costs, he says. Sharing can reduce those costs and create licensing opportunities, with commercial terms agreed with each partner.Reger frames this as a way for the German Mittelstand to turn generations of process know-how into robot capabilities while retaining control over it.“Our ambition is that a company can eventually export its expertise as a robotic skill, alongside the products it already sells,” he says.
Competitive Landscape
Neura competes on two fronts: humanoids and general-purpose physical AI, where the field is global and very well funded, and the more established cobot and intralogistics markets, where it faces incumbent automation suppliers. The humanoid race is where its valuation is being set. Competitors include Chinese robotics companies AgiBot and Unitree; U.S. companies Figure AI, Tesla (with Optimus), Agility Robotics, and Boston Dynamics; as well as the U.K.’s Humanoid.
Neura claims an order book and pipeline above $1 billion and targets millions of robots by 2030. Chinese robotics companies ship about 97% of the world’s humanoids, but U.S. and European labs lead on robot brains.
Neura’s differentiator is that it “brings together the robot, its intelligence and the infrastructure through which it learns,” says Reger. The company’s hardware provides the senses and movement. Neuron AI connects perception with action. In the Neura Gyms, partners teach robots real tasks, and the Neuraverse connects training, skills, and deployment. All of these things give Neura “a direct link between what happens in the physical world and what we improve in the technology,” he says. “Partners contribute the process knowledge that makes it useful. A manufacturer knows its own production better than any robotics company ever could. We want to help turn that expertise into capabilities it can deploy across its business. That is how Physical AI becomes an industrial tool.”
Neura’s advice to companies? Start experimenting with humanoid robots now and help shape the industry and the standards.
