A Swiss robotics company developing an artificial intelligence system designed to power the next generation of humanoid robots has raised fifty million dollars in new funding. Flexion, based in Zurich and founded only last year, claims its technology could mark a turning point for the industry by allowing robots to reason, adapt and operate with minimal human involvement.
The Series A round, led by DST Global Partners and joined by Nvidia’s NVentures, Redalpine, Prosus Ventures and Moonfire, brings Flexion’s total funding to more than fifty seven million dollars. The company plans to expand its teams in Switzerland and the United States, scale up its robot fleets and push its software toward commercial deployment.
A new approach to robot intelligence
Flexion’s founders, former researchers from Nvidia and ETH Zurich, argue that today’s robots remain limited by heavy scripting and human teleoperation. Many systems require vast numbers of demonstrations or remote operators to complete even simple tasks. According to chief executive Nikita Rudin, this leaves the field stuck with brittle, task-specific solutions that fail outside tightly controlled environments.
The company’s answer is a three layer autonomy stack built on recent advances in generative AI. A command layer uses language models to interpret natural language instructions and break them into subtasks. A motion layer combines vision, language and action models trained largely on synthetic data to generate movement plans. Finally, a transformer based control layer coordinates the robot’s full body in real time, drawing on a modular library of skills that can be recombined for new behaviours.
Flexion says this architecture avoids monolithic black box systems by keeping each layer modular and testable. Its heavy use of simulation and synthetic data also sets it apart from rivals who rely on thousands of hours of costly human teleoperation. Investors say this gives the company a significant advantage, allowing it to generate diverse training experiences without being constrained by human labour.
From simulation to the real world
A key claim is that Flexion’s technology can generalise across different robot types, from humanoids to wheeled or multiarm machines. Rather than building intelligence for a single model, the firm aims to provide an adaptable brain that manufacturers can license for a wide range of platforms. Early trials, including work with major equipment partners, suggest the system can already plan and execute tasks such as grasping objects or navigating around obstacles.
Central to this is what Flexion calls its sim to real strategy. The AI learns primarily in high performance simulations, then transfers that knowledge to real robots with selective fine tuning using real world data. The company says this approach produces motion generation and control systems robust enough to handle unpredictable environments and terrain.
Why investors are backing Flexion
The latest funding comes during a surge of interest in robotics, with global investment already exceeding ten billion dollars this year. Backers say Flexion’s technology addresses a missing piece in the sector: a scalable intelligence layer that allows robots to operate autonomously in everyday settings.
Analysts point to ageing populations, labour shortages and rising industrial demands as factors accelerating the need for more capable robotic systems. Flexion believes humanoid robots will play a central role in meeting that demand, potentially working in factories, logistics centres, disaster zones and even in future space missions.
For now, the company is focused on developing what it calls the brain of the robot, leaving hardware design to manufacturers. Its ambition is to create general purpose robotic intelligence that can support long horizon tasks and work safely alongside people.
Flexion’s founders describe the challenge as difficult but necessary. As Rudin puts it, the aim is not simply to build robots that move, but robots that think.
