IMPLICATOR
NEURA Robotics and Qualcomm Technologies announced a long-term partnership today to build reference architectures for cognitive robots that handle perception, reasoning and physical control on the device itself, not in the cloud.
The deal pairs Qualcomm's Dragonwing IQ10 robotics processors with NEURA's full-stack robot hardware and its Neuraverse software platform, targeting machines that can work safely alongside humans in factories, warehouses and homes. The announcement landed ahead of the Embedded World trade show in Nuremberg and days after Bloomberg reported that NEURA is closing a €1 billion funding round backed by stablecoin issuer Tether, valuing the Metzingen-based startup at roughly €4 billion.
Qualcomm has been looking for something bigger than phones. CEO Cristiano Amon stood on stage at Mobile World Congress in Barcelona a week ago and said robotics would become a "larger opportunity" within two years. Dragonwing, Qualcomm's first robotics processor, shipped in January. It goes straight at Nvidia's Jetson, which has owned this market for years.
Key Takeaways
- NEURA Robotics and Qualcomm will co-develop edge AI reference architectures using Dragonwing IQ10 processors for cognitive robots
- Qualcomm targets Nvidia's Jetson dominance with energy-efficient robotics chips that shipped in January
- NEURA reportedly closing €1B Tether-backed round at €4B valuation, roughly half initial expectations
- Partnership joins Schaeffler, Bosch and Zimmer Group deals building NEURA's manufacturing and AI training pipeline
Why Qualcomm is betting on robot chips
Ask anyone at Qualcomm about Dragonwing and they'll point you to the Snapdragon story. Speed wasn't what made Snapdragon the default Android chip. The full platform was, software tools, reference boards, a developer community that locked in manufacturers before competitors could catch up. Amon wants that same lock-in with robots.
"People have said just robotics alone could be a trillion-dollar opportunity in terms of market size," Amon told CNBC on March 3. "The reality is, we see now, because of physical AI, robots have become a lot more useful."
Wall Street loves these numbers. McKinsey sees three hundred seventy billion dollars in general-purpose robots by 2040. RBC Capital Markets sees nine trillion in humanoids by 2050. Those projections sit decades out, which makes them almost unfalsifiable. But Qualcomm doesn't need a nine-trillion-dollar market. It needs enough robot companies buying Dragonwing chips to cover the engineering cost, and that bar sits a lot lower.
Nvidia CEO Jensen Huang has been making the same bet from the opposite direction. His company's Isaac robotics platform and Jetson edge processors already power robots from Boston Dynamics, LG Electronics and NEURA itself. Qualcomm positions Dragonwing as the energy-efficient alternative, trading raw compute power for lower power draw. For a humanoid robot running on batteries all day, that trade-off matters a great deal.
The architecture inside the machine
NEURA and Qualcomm call their joint architecture "Brain + Nervous System." Not a metaphor. An engineering spec. One layer does the thinking, deciding what to pick up, where to walk, how to react when a person steps into its path. The other layer moves the body, sub-millisecond motor commands that keep a robot from dropping a crate or catching a foot on uneven concrete.
Why run both layers on the device? Two reasons. Speed and privacy. A robot working next to a person on a factory floor cannot wait 200 milliseconds for a cloud server to decide whether to stop its arm. And a manufacturer running proprietary production processes doesn't want that sensor data leaving the building.
Qualcomm's Dragonwing IQ10 processors handle heterogeneous edge computing, mixing general-purpose CPU cores with dedicated AI accelerators and real-time control units on the same die. NEURA layers its own embodied AI software stack on top. That stack handles 3D vision, natural language processing and tactile sensor feedback from NEURA's patented Touchless Safe Human Detection system, which lets robots sense humans nearby and stop before contact occurs. No safety cage required.
The partnership also plans a standardized runtime interface so that applications built for one robot form factor can run on another without rewriting the software. A perception model trained on a factory arm, for instance, could deploy onto a humanoid with minimal reconfiguration.
NEURA's Neuraverse platform sits at the center of this approach. It functions as a simulation, training and deployment environment for physical AI workloads running on Dragonwing hardware. One feature stands out from a competitive standpoint. When any single robot in the network learns something new or encounters a novel situation, the Neuraverse pushes that knowledge across every connected machine. NEURA calls this "shared intelligence." A fleet of a thousand robots effectively learns faster than one robot working alone. That's the logic, anyway.
Join 10,000+ AI professionals
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
NEURA's partnership blitz
The Qualcomm deal is the latest in a string of partnerships that NEURA signed over the past five months, each one targeting a different bottleneck in scaling humanoid robots from working prototype to factory-floor product.
Start with the joints. Nobody has figured out how to build actuators that fit inside a humanoid shoulder, generate enough torque to lift heavy objects, and survive 24-hour shifts without burning out. Schaeffler stepped in last November to co-develop high-torque versions for those joints. The deal goes both ways. Schaeffler plans to put a mid-four-digit number of NEURA humanoids into its own factories by 2035, generating the movement and interaction data that feeds back into the Neuraverse for AI training.
Then came the training data problem. Workers in Bosch production facilities now wear sensor suits that capture motion, workflow sequences and environmental readings as part of a January 2026 technology partnership. That information trains NEURA's AI models for locomotion intelligence and situational awareness. Bosch also contributes hardware components and manufacturing know-how to help scale production beyond prototype volumes.
Smaller deals filled remaining gaps. Zimmer Group brought industrial gripping technology into the Neuraverse in February. Drees & Sommer signed on to develop buildings with embedded sensors that communicate with NEURA's robots in real time. NEURA also bought ek Robotics, a 300-person industrial automation outfit, right after closing its €120 million Series B in early 2025. Less flashy than the Bosch or Schaeffler headlines. But ek Robotics came with a fleet of warehouse logistics robots and the engineers who know how to keep them running.
The money and the question it raises
Partnerships don't pay for themselves. The reported €1 billion round from Tether, at a €4 billion valuation, would supply it. But the valuation itself tells a story. When rumors of the Tether deal first surfaced in November 2025, sources told the Financial Times that NEURA could be valued at €8 billion to €10 billion. The actual figure came in at roughly half that.
NEURA's order book reportedly sits near $1 billion, with customers including Kawasaki Heavy Industries and Omron. The company employs more than 1,200 people across 45 countries and eight offices. Its product lineup spans robotic arms, mobile logistics platforms, the MAiRA collaborative robot, a consumer assistant called MiPA, a quadruped designed for rough terrain, and the flagship 4NE1 humanoid. The 4NE1 stands 5.9 feet tall, carries up to 220 pounds and runs NEURA's proprietary AURA AI system. A smaller variant, the 4NE1 Mini at 52 inches, debuted at CES 2026 in January alongside the quadruped.
CEO David Reger has moved aggressively to build manufacturing capacity outside Germany. NEURA shifted its humanoid R&D to Zurich, positioning engineers near ETH Zurich's robotics cluster. Manufacturing partnerships with Sona Comstar in India and a €45 million subsidiary in Hangzhou give the company production footholds in the two largest growth markets for industrial automation. Reger has also turned down defense industry contracts, insisting NEURA's robots serve civilian applications only.
Tether brings patient capital and nothing else. The stablecoin issuer earned more than $10 billion in net profit during Q3 2025 and has been investing in AI data centers and energy infrastructure. For NEURA, the appeal is simple. Tether can finance capital-intensive hardware scaling over years without the exit pressure that typical venture funds apply after year three or four.
NEURA can sign partnerships. Manufacturing humanoid robots at scale while simultaneously building the AI software that makes them worth buying is a harder trick. Five million robots by 2030, that's NEURA's stated target. Tesla, with decades of assembly-line experience and supply chains that cost billions to build, ships roughly 1.8 million vehicles a year. Five million humanoids from a company that barely existed seven years ago.
Qualcomm's Dragonwing processors give NEURA a compute platform it doesn't have to design from scratch. Schaeffler's actuators solve the joint hardware problem. Bosch's factories supply training data. But assembling all of these pieces into production lines that reliably output humanoid robots at industrial price points is a different kind of problem. You can hear the gap between ambition and execution in the valuation itself. Investors are cautious. A company with a billion-dollar order book that lands a €4 billion price tag, not the €8 billion to €10 billion its backers initially wanted, is a company the market believes in conditionally. The condition is delivery. The factory floor, not the partnership announcement, is where that gets tested.
Frequently Asked Questions
What is Qualcomm's Dragonwing IQ10 processor?
Qualcomm's first dedicated robotics processor, shipping since January. It combines CPU cores, AI accelerators and real-time control units on one chip. Qualcomm positions it as an energy-efficient alternative to Nvidia's Jetson, trading raw compute for lower power draw, which matters for battery-powered humanoid robots working full shifts.
What is NEURA's Neuraverse platform?
A simulation, training and deployment environment for physical AI workloads. Its standout feature is shared intelligence: when one robot learns something new, Neuraverse pushes that knowledge across every connected machine. The platform runs on Qualcomm Dragonwing hardware and handles 3D vision, natural language processing and tactile feedback.
Why did NEURA's valuation come in at €4B instead of the rumored €8-10B?
The gap suggests investor caution about execution risk. NEURA claims a billion-dollar order book and 1,200 employees across 45 countries, but manufacturing humanoid robots at scale while building AI software simultaneously is unproven. The lower valuation reflects conditional market confidence pending actual delivery.
How does edge AI differ from cloud-based robot control?
Edge AI processes perception, reasoning and motor commands directly on the robot's onboard chip rather than sending data to remote servers. This eliminates network latency, critical when a robot arm must stop within milliseconds to avoid hitting a person. It also keeps proprietary manufacturing data inside the factory.
What role does Tether play in NEURA's funding?
Tether, the stablecoin issuer that earned over $10B in Q3 2025 profit, reportedly backs NEURA's €1B round. The appeal is patient capital. Unlike venture funds that pressure startups for exits after three to four years, Tether can finance capital-intensive hardware scaling over longer timelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: [email protected]
