Stuttgart-based Sereact has closed a $110 million Series B led by Headline to launch Cortex 2.0, a vision-language-action model that lets industrial robots simulate multiple outcomes and choose the action most likely to succeed , a capability that moves robotics from scripted execution to genuine physical reasoning.
The distinction Sereact is drawing matters more than it might sound. Most industrial automation is brittle precisely because it is programmed: a robot arm trained to pick one type of box will fail on a slightly different box, different lighting, or a crushed corner. Warehouses and logistics environments are full of exactly those edge cases. Sereact's Cortex 2.0 trains on real production data from more than 100 live deployments , actual picks at BMW, Daimler Truck, Bol, and H&M, which the company has confirmed is in active talks , rather than lab simulations. That real-world training loop is what enables the consequence-prediction architecture: the model has seen enough physical variation to simulate plausible outcomes before committing to a movement, rather than pattern-matching to the closest training example and hoping the execution holds.
CEO Ralf Gulde, who co-founded Sereact with Marc Tuscher in 2021, has described the core principle in consistent terms: robots that act situationally rather than following rigidly programmed sequences. The name Sereact is a compression of sense, reason, act, and Cortex 2.0 is the first model that makes the reason step genuinely consequential. The company is building across robot arms, mobile systems, and humanoid platforms , a deliberate decision to avoid being tied to one hardware form factor as the physical AI market evolves. Its investors agree. Creandum, Air Street Capital, and Point Nine returned for the Series B alongside new entrants Headline, Bullhound Capital, Felix Capital, and Daphni. Nico Rosberg, the former Formula 1 world champion, remains a backer. Total disclosed funding is now above $140 million.
Grishin Robotics' analysis of the round captures the sharper strategic thesis: robotics software becomes more valuable than the robot body. That argument has been made about enterprise software generally for three decades, and it is now being applied to physical systems at scale. If Cortex 2.0 can sit as the operating intelligence across different hardware , existing installed arms, new mobile systems, future humanoids , Sereact occupies a platform position rather than a product position. Platform positions in enterprise software produce durable margins and compounding switching costs. A logistics operator that has trained Cortex on two years of production data has a switching cost that is not easily replicated by a competitor with a fresher model and fewer real-world hours.
The Boston office opened alongside the Series B announcement, which signals that Sereact is positioning for US enterprise contracts rather than just European expansion. BMW and Daimler Truck as existing customers give the company automotive references that transfer directly to American manufacturing procurement conversations. H&M signals the retail and e-commerce logistics opportunity, where order volumes, SKU variety, and return processing create exactly the unpredictable physical environment that brittle automation fails and consequence-predicting systems thrive.
Europe's physical AI moment
The raise lands at a specific moment in the European AI investment landscape. European investors and founders have watched large language model investment concentrate in the US and China. Physical AI , robotics, autonomous systems, industrial automation , is a domain where European manufacturing depth, engineering talent, and industrial customer relationships create genuine competitive advantage. Sereact is not alone: Swiss and UK-based robotics startups are drawing comparable capital for similar architectural bets. But Stuttgart is not a coincidence. Germany's automotive and logistics sectors provide the densest concentration of industrial robotic deployment in the world outside Japan and South Korea. Building and training a consequence-prediction model in that environment produces a dataset advantage that is genuinely hard to replicate from a San Francisco lab.
The $110 million will go primarily toward Cortex 2.0 development and the US expansion. The model launches today. The commercial test of whether consequence-prediction at scale produces the reliability improvement that investors are pricing in will play out over the next eighteen months in warehouses that process millions of picks a year. That is a real-world benchmark that no laboratory can substitute for, and it is exactly the benchmark Sereact has built its architecture to win.
Also read: AI agents can now cost more than the humans they were supposed to replace • Hipfire is a Rust-native AMD inference engine that beats llama.cpp on consumer GPUs • China's Manus intervention rewrites the rules for cross-border AI deals
