Sereact raises $110M to scale Cortex 2 model

According to Bloomberg, The Next Web, Tech.eu, and EU-Startups, Stuttgart-based robotics software company Sereact raised $110 million in a Series B funding round led by Headline. The Next Web and Tech.eu list new investors Bullhound Capital, Felix Capital, and Daphni, and Tech.eu and EU-Startups report participation from returning investors Air Street Capital, Creandum, and Point Nine. Tech.eu and The Next Web report that Sereact did not disclose a valuation. Tech.eu and EU-Startups state the company intends to use the funding to develop and scale its latest model, Cortex 2, and to expand operations in the US, including a new Boston office planned for the coming summer (Tech.eu, EU-Startups). EU-Startups quotes CEO Ralf Gulde with deployment metrics: 200 systems, one billion picks, and an intervention rate of one intervention per 53,000 picks, which the article presents as the company's performance claim (EU-Startups).

Tech.eu and The Next Web describe Sereact's approach as a software-first stack built around Vision Language Action Models (VLAMs) that combine visual perception, language understanding, and action planning into a unified model. Tech.eu reports that Cortex 2 trains on diverse physical behaviours so a robot can select the most promising action for a given manipulation task, extending beyond simple pick-and-place to contact-sensitive work such as assembly and kitting (Tech.eu). The Next Web adds that Sereact's systems run on a range of hardware form factors, from single-arm and dual-arm setups to humanoid platforms, and that real-world deployments feed a data flywheel used to improve models over time (The Next Web, Tech.eu).

Editorial analysis: Companies commercialising physical-AI for manipulation increasingly prioritise closing the sim-to-real gap by collecting production data from warehouses and factories. Observers following the sector note that a data flywheel fed by real deployments can materially improve robustness on edge cases compared with models trained predominantly in simulation. For practitioners, this pattern emphasizes operational engineering (sensor integration, safety, intervention logging) as much as model architecture when moving manipulation models into production.

Editorial analysis: A $110 million Series B led by a major cross-border VC like Headline signals investor conviction in enterprise robotics software rather than hardware-heavy automation players. For ML and robotics engineers, the notable parts are the reported scale of real deployments (customer names and the numbers quoted by Ralf Gulde in EU-Startups) and the explicit claim that Cortex 2 targets contact-rich tasks. Those two elements matter because contact-sensitive manipulation is substantially harder than bin picking and is where software-level generalisation can create differentiation if it holds up in diverse production settings.

Editorial analysis: Observers and practitioners should track three indicators: 1) independent benchmarks or third-party demonstrations of Cortex 2 on contact-rich tasks; 2) customer case studies from the named partners (BMW, Daimler Truck, Bol, Active Ants) showing deployment scope and maintenance burden; and 3) how Sereact documents safety, intervention rates, and tooling for supervised corrections in production. Tech.eu and The Next Web report that a US office in Boston is part of the expansion plans; EU-Startups places that timeline in the coming summer, which will be an observable milestone for hiring and local deployments (Tech.eu, EU-Startups).