Flexion €43M Series A to build humanoid AI brains, scaling US ops

Flexion, a Zurich startup focused on reinforcement‑learning control for humanoid machines, has closed a €43 million (about $50 million) Series A round to accelerate development of its autonomy stack and expand into the United States.

The financing was led by DST Global Partners and includes participation from NVentures (NVIDIA’s venture arm), redalpine, Prosus Ventures and Moonfire. The cash follows a recent €6.3 million seed round the team secured earlier, giving Flexion substantial runway to scale compute, simulation and robot fleets.

Founded in 2024, Flexion positions itself as a software-first player: it is not building new hardware but aims to deliver the “brain” that can enable humanoid platforms to operate reliably in unpredictable, human-centred environments. The founders are engineers and researchers with backgrounds at institutions and firms such as ETH Zurich, NVIDIA, Meta, Google, Tesla and Amazon, and the stack combines language-grounded task parsing, synthetic vision-language-action training and transformer-based full‑body control.

The company outlines a three‑layer architecture: a Command Layer using large language models to translate goals into actionable plans; a Motion Layer primarily trained on synthetic data with real‑world exception handling; and a Control Layer that composes modular skills into behaviours. Together the layers are designed to detach cognitive capabilities from any single robot morphology, increasing portability across OEM platforms.

Flexion’s raise arrives amid a busy European robotics funding landscape. Notable disclosed rounds this year include NEURA Robotics (€120 million), mimic (€13.8 million), Unchained Robotics (€8.5 million) and several sub‑€15 million deals across autonomy and dexterous manipulation. Those rounds total roughly €165 million in public disclosures and underline investor appetite for software and AI layers that can unlock commercial use beyond labs.

The market case for adaptable humanoid intelligence is framed by demographic and labour trends: ageing populations and persistent skills shortages in manufacturing, logistics and field services are increasing demand for autonomous systems that can generalise across tasks. Analysts estimate that flexible autonomy could address capacity gaps in industries where bespoke automation has historically been too brittle or costly to deploy.

Challenges remain: translating promising lab demonstrations into reliable, certified field deployments requires robust sim‑to‑real transfer, safety validation and durable partnerships with hardware makers. Flexion aims to meet those hurdles with a hardware‑agnostic software stack and by scaling compute and fleet testing, betting that cognition — not mechanics — will be the differentiator in the next wave of robotic adoption.