Germany's New Robot Billionaire Is Betting Big On Humanoid Gyms

By Nina Bambysheva | Forbes | October 18, 2026

David Reger’s $7 billion Neura Robotics wants to be theApple of physical AI, building not only an army of industrial and humanoidrobots, but also their operating software system and the “gyms” where they willtrain. Nvidia, Amazon and Tether are onboard to the tune of $1.4 billion.

“I want Neura to become the largest taxpayer in Europe oneday,” enthuses 38-year-old robotics entrepreneur David Reger.

His ambitious startup Neura Robotics was founded in 2019 inMetzingen, Germany, a town 20 miles south of Stuttgart. This region gave riseto automotive powerhouses like Mercedes-Benz, Porsche and Bosch. Neura buildscognitive or “thinking” robots, including humanoids, as well as the operatingsoftware for them. In June, Neura raised up to $1.4 billion in a financinground led by Tether, the controversial company behind the world's largeststablecoin, a cryptocurrency pegged to the U.S. dollar. Nvidia, Amazon,Qualcomm, Bosch, Schaeffler and the European Investment Bank also invested.

The round valued Neura at about $7 billion, according topeople familiar with the deal, putting it among Europe's most valuable roboticsstartups. The company says most of the capital has already been transferred,with a smaller portion still dependent on performance milestones. Reger ownsabout 20% of it, making his stake worth nearly $1.5 billion.

One of Neura’s biggest boosters is Stephen George, whose investment firm InterAlpen Partners made its first investment in the company, worth $16 million, in 2023. An ex-Goldman banker, he previously ran eBay billionaire Jeff Skoll’s family office, Capricorn Investment Group. “Neura is Apple meets Mercedes-Benz,” says George, with a nod to the car maker headquartered half an hour up the road.

The least conventional part of Neura’s strategy is whatReger calls robot gyms, where his clients can rent dedicated training cellsequipped with simulation environments and tracking technology to train therobots it buys from Neura for specific jobs. Whatever one robot learns can thenbe stored on Neura’s cloud-based platform, NeuraVerse, and transferred to othermachines the customer owns. Reger wants companies to be able to trade or leasethose skills, too.

Of course many of Neura’s grand ambitions have yet to berealized. Most of its actual revenue, which the company declined to disclose,still comes from more conventional collaborative robots (known as cobots),including mechanical arms, mobile carts and their combinations. Neura’s latestavailable financial statement from Germany’s Company Register dates back to2022, where it reported a loss of close to $7 million. A spokesperson for thecompany says that its orderbook now exceeds $1.1 billion, half relating to itsexisting cobot fleet. Neura also claims it has $11 billion in its pipeline,“comprising signed agreements with varying levels of commitment.”

According to the spokesperson, approximately 80% of thislong-term pipeline relates to humanoids, Neura gyms and NeuraVerse.

On its website, Neura markets its 4NE1 humanoid robot as“Your Next Gen Humanoid Teammate." The white and black cyborg, which thestartup is pinning most of its future on, stands 5 feet 11 inches tall, weighs176 lbs and looks like it came out of central casting for a dystopian sci-fithriller.

Chinese robot makers have already shipped 19,100 humanoids,or 97%, of all humanoids sold worldwide in the first half of the year. Figure,the Silicon Valley humanoid maker, was valued at $39 billion last year, morethan five times what Neura is worth now. German-born Reger still believes hiscompany can put Europe back in a technology race it has mostly watched from thesidelines.

His motivation goes back to San Francisco, 2009. That iswhen his family urged him to do social work before starting a career, so at 21,just after the financial crisis, he moved to California and became a casemanager for homeless people. He had planned to stay in the U.S., but what hesaw on the job sent him home after a year and a half with a newfoundappreciation for Germany’s social safety net.

That safety net depends on workers, and Germany is runningout of them. As baby boomers retire, the country’s workforce is set to shrinkby about 4.3 million people by 2036, according to the German EconomicInstitute. Robots will have to fill part of that gap. If Germany is going todepend on them, Reger argues, it should build them itself.

With the country currently dependent on critical resourcesfrom oil to semiconductors from abroad, Reger is looking for a rebalancing inAI-driven technology. “Europe cannot preserve its prosperity, its socialsystems and its relevance in the world if we only consume the next generationof technology from somewhere else,” he says. “We have to create enormous newvalue here again. Physical AI gives us that opportunity.”

Born in 1988, Reger grew up near Rottweil in southwesternGermany, the eldest of 11 siblings. His father works at a nearby automotivesupplier and his mother is a homemaker. As a boy, Reger spent his Saturdays atthe plant cleaning machinery. He skipped university for a three-and-ahalf-yearapprenticeship as a mold maker, learning to build the tooling used to castindustrial parts.

He had never touched a robot when MABI, a family-owned Swissmachinery maker, hired him in 2011 and gave him a choice between working onrobotics and 3-D printing. Reger picked robots first, figuring they mightsomeday do the printing anyway. “I had no clue about robots,” he says. “Ipretended I had a clue.” He taught himself at night from YouTube videos. Withina year, he says, his small team was demonstrating how to use a large roboticindustrial arm at Automatica, Munich’s big automation fair.

In 2018, after seven years at MABI, Reger took a bigger ideato its board. Most factory robots then sat behind fences, repeating one taskover and over. He wanted machines that could be taught new jobs, with what theylearned stored on a shared software platform so engineers would not have torebuild the automation every time. The board liked the idea but told him he wasfive or six years early. Reger thought waiting would make him late.

So instead, he partnered with Han’s Robot, a Shenzhen cobotmaker backed by automation giant Han's Laser Technology, to establish aGerman-based collaborative robotics venture, Han's Robot Germany. He recruitedabout a dozen robotics scientists and engineers, including his colleagues atMABI, with expertise spanning AI, hardware, software and sensor systems. TheChinese company took a 70% stake in the venture, which combined Han’s robotichardware with the German team’s expertise in software, sensors and developingmachines for customers’ specific needs. It was eventually renamed Neura.

In June 2023, Reger arranged for a group of Europeaninvestors to buy out Han's ownership for $11 million. By then, ChatGPT had madeAI the hottest trade, and investors were looking for ways to put it to work inthe physical world. A month after the Chinese sold their stake, Neura announcedanother $55 million raise from a group of European investors at an undisclosedvaluation.

Why would China’s Hans sell its stake at such an apparentdiscount? Perhaps because Neura continued to buy a bulk of its collaborativeindustrial robots and components from Han’s Robot, which was since renamedHuayan Robotics. According to an IPO prospectus filed by Huayan in March 2026,Neura was also Huayan’s largest customer as of September 2025 and in turnsupplied Huayan with high-performance reducers, the precision gearboxes thatmove a robot’s joints. Huayan built them into the robots it made for Neura andwas free to put them in robots for other customers. Despite the cozyrelationship, Reger insists that “neither Han’s Robot nor Han’s Laser retainsequity or rights to Neura’s IP.”

To date, Neura has mostly prospered by selling robot armsthat list for about $18,500 to $44,000 at European resellers, wheeled robotsthat carry parts around a factory, which sell for about $50,000, and machinesthat combine the two. Reger is vague about how many machines he has sold since2021, when Neura launched its first robot. He estimates 5,000 to 8,000 of themand says many are sold under the brands of other robot makers, who are itspartners. Among them: Japan’s Kawasaki Heavy Industries, Omron Robotics andTaiwan’s Delta Electronics, which invested in Neura in 2024 via its Dutchsubsidiary.

Why would a robot maker sell someone else’s technology? Whenthe partnerships began, nothing comparable was for sale, answers Reger. Neura“owns the stack for everything that’s core,” he says, including motors,encoders and safety detection sensors.

Owning doesn’t mean making. Some of the parts for Neura’shumanoids come from German suppliers like Bosch and Schaeffler, which providesNeura with actuators, a.k.a. “the muscles of robots”. These devices allowhumanoid joints to move and can amount to a half of a robot’s component costs.

The startup’s other partners include Nvidia, which suppliessimulation technology and compute behind Neura’s AI, Amazon for cloud-basedNeuraverse, and Qualcomm, whose processors Neura is building into the robots’“brain and nervous system.” Tether, whose dollar-pegged stablecoin USDT iswidely used for trading and payments in emerging markets but has also become apopular vehicle for illicit finance, wants to put its wallet software insideNeura’s robots. Eventually robots could be able to transact with each other.

“Autonomous machines need the ability to process informationlocally, make decisions, and transact without relying on centralizedintermediaries,” says Tether’s billionaire chief executive, Paolo Ardoino.Reger shares his vision. “The future will be that every robot has its ownwallet,” he says.

Demand for humanoid robots could be immense. Reger claims12,000 companies contacted Neura in the last year, mostly about its humanoids.He figures it currently has the ability to make only enough robots for about100 customers. “Companies are willing to pay $300,000 to $550,000 just for “anice demo,” says Reger, referring to the cost of a single humanoid robot itself($110,000 for Neura’s latest model) plus the annual fees associated with itstraining, deployment and maintenance. Manning humanoid pilot programs forcustomers is resource intense. It typically requires Neura to send its ownengineers to someone else’s factory, which disrupts any inhouse progress onimproving humanoid technology.

So Reger decided that Neura would build facilities, or gyms,where companies could rent space and have their own workers train the robotsthemselves. Unlike large language models, which learn largely from informationavailable online, robots need to learn how to interact with the physical world.To pick up a car part, for instance, a robot must learn where to reach, howfirmly to grip it and how to put it in the right place. That requiresdemonstrations and repeated practice, explains investor George.

Neura’s robot gyms are expensive. The company and itspartners have committed $19 million to a facility at the Munich Airport. Andthat didn’t even include staffing costs. The gym is already operational, andcustomers are already booking training cells—some as many as 20—for at least ayear, says Reger. Rent varies depending on the equipment and data sharingarrangements, but can run into the tens of thousands of dollars.

Universities are another part of Reger’s plan. Researchinstitutions are short of funding and losing top PhDs to startups, he says. Arobot gym on campus would give them access to machines they might otherwisestruggle to afford, while providing Neura with researchers to help train itsrobots. The Technical University of Munich is a partner in the airportfacility, and a 3,000-square-meter site is being developed on RWTH AachenUniversity’s Melaten research campus.

Neura expects to have five gyms running by year-end: threein Germany, one in China and one in the U.S., and has 10 more in development.The U.S. plans include facilities in Detroit and San Diego.

Data collected during training can be stored on Neura’s“brain accounts” on Neuraverse, allowing companies to transfer what theirrobots have learned to other machines in their fleets or eventually sell thoseskills to other businesses. “Ultimately, our goal is that once a skill has beenlearned, it does not have to be engineered again for every individual robot,”says Reger.

Neura’s robot gyms will have competition. Investor andsupplier Schaeffler operates its own humanoid gym in Erlangen, Germany, whereit is training Hexagon’s AEON human robot for factory work. It has alsoinvested $10 million in another Neura rival, Agility Robotics, and is using afew of its Digit humanoids to handle materials in its U.S. factories.

Neura’s robots and their promised capabilities areimpressive, but when it comes to actual functionality its humanoids are stillfinding their footing. At the high-profile unveiling of 4NE1 at Automatica inJune 2025, Reger had only a motionless prototype to show. A companyspokesperson said an unexpected highway closure had delayed the arrival of theworking robot. A year later, at the Computex technology show in Taipei in June2026, a newer version of the same robot made it onstage carrying a tray ofQualcomm hardware, only to collapse backward moments later. Qualcomm blamed acommunication glitch, saying it had triggered the robot’s built-in“safe-collapse” sequence, designed to protect people and equipment nearby.

Neura’s latest humanoid, Gen 3.5, costs an estimated$110,000, or according to the company’s website, $62,000 if you purchase 20robots or more. Neura refuses to say how many humanoids it has actually builtand deployed, but Reger says his startup will be able to deliver 5 millioncognitive robots by 2030, including humanoids.

The International Federation of Robotics counted just about7,000 full-size humanoid robots sold worldwide last year. Many were bought notto work but to generate training data for AI models.

China is leading the humanoid race. Shanghai-based AgiBotaccounted for 44% of global humanoid shipments in the first half of 2026, orabout 8,400 robots, according to the market research firm Smart AnalyticsGlobal. Runner-up Unitree, based in Hangzhou, sells its basic G1, astripped-down humanoid that can walk but lacks dexterous hands and cannot beprogrammed by its owner, for $13,500, roughly an eighth of the 4NE1’s price. In2025, Unitree booked $253 million in revenue and shipped more than 5,500humanoids, though research firm SemiAnalysis estimates that only about 250 ofits humanoids are doing real work. After Unitree listed in Shanghai in August2026, investors briefly valued the company at $66 billion. Its market value hassince plunged to $24 billion, as the gulf between enthusiasm for humanoidrobots and the machines’ actual commercial use has become harder to ignore.

American humanoid makers have plenty left to prove, too. SanJose, California-based Figure has built about 1,000 humanoids, but it’s notclear how many have been deployed and where. Tesla is now building severalhundred Optimus humanoid robots per week, but the robots can’t yet reliablyhandle a wide range of tasks, according to a recent report from TheInformation.

One serious headwind in Neura’s near future comes frominternational trade barriers. In July, the U.S. Federal CommunicationsCommission added foreign-made “advanced robotic devices” to its Covered List, anational-security roster of equipment barred from the FCC authorization requiredfor sale in America. The category covers humanoids, four-legged and autonomousmobile robots. Fixed industrial arms, a good chunk of Neura’s business, andpreviously authorized models are exempt. A foreign maker can still bring a newrobot to market if it receives “conditional approval” granted after theDepartment of War reviews its manufacturing and supply chain. Neura told Forbesit is “developing manufacturing capacity in Detroit for wheeled robots andhumanoids, targeting 2027,” but until then it will need that clearance for itsnew machines.

Undoubtedly, Reger’s bigger ambition is at home. For morethan a century, the world has driven machines built in this corner of Germany.Now he wants it to build the machines that do the world’s work. “Europe has theengineering talent, industrial knowledge, manufacturing base and thousands ofcompanies that understand the physical world,” he says. “If we combine thatwith AI and move fast enough, I believe we can create an industry that becomesbigger than automotive.”

‍