German robotics company Sereact announced on Sunday that it has completed a $110 million Series B funding round, primarily aimed at pushing into the U.S. market and scaling production of Cortex 2.0 — what the company calls a "robotic brain."
This is not another humanoid robot startup. Sereact equips existing industrial robotic arms with a "cognitive layer." Its client list already speaks volumes: BMW, Daimler Truck, Mercedes-Benz, PepsiCo, Active Ants, and Austrian Post.
What Exactly Is Cortex 2.0?
In simple terms: it combines a vision-language-action (VLA) model with a world model, allowing a robotic arm to understand tasks and plan its own movements without engineers having to write grasping scripts for every type of box.
What sets Sereact apart from companies that only run demos is its training data. Cortex 2.0 is trained on over 1 billion real-world production "picks" — not simulations, not lab data, but picks accumulated by robotic arms working in factories.
CEO Ralf Gulde put it bluntly:
"You have to feed it with a data flywheel — push the model onto the production line, endure failures alongside it, and let the model learn from what actually happens on the shop floor."
In other words: no matter how impressive a lab demo looks, a model only truly learns to work when a robot is dropped into a BMW warehouse for eight-hour shifts, where a single failure can cost thousands of dollars.
A Number Worth Highlighting
Sereact currently has 200 systems deployed and running.
On average, human intervention is required only once every 53,000 picks.
To put that in perspective: a typical logistics robot handles roughly 2,000 to 5,000 picks per day. At 53,000 picks per intervention, that means a machine can run continuously for over 10 days before needing human help. In the warehouse automation industry, this is a top-tier figure.
For comparison, many warehouse automation startups still tout "100 consecutive successful picks" in demo videos as a selling point.
Why Germany, and Why Now
Germany and Europe have long been strong in logistics automation (with legacy giants like KUKA and ABB), but few have applied the latest AI techniques like VLA and world models. Sereact is one of the few bringing the Silicon Valley-style foundation model approach to European warehouses.
The U.S. market is the key target for this funding round. Amazon and Walmart warehouses are sitting there, and landing one or two major clients could immediately double the company's valuation.
The deeper logic: physical AI is moving from "can it run a demo?" to "dare it go on the production line?" The first phase was about model architecture and paper benchmarks; the second phase is about the flywheel — the more robots run in real factories, the more data they collect, and the stronger the next generation of models becomes.
Sereact has already carved out a strong position in this data flywheel race.
What to Watch Next
This round did not disclose valuation. But for reference: competitor Robot Era raised over $200 million in Series C at a $1.5 billion valuation, and Figure AI has already surpassed $39 billion. Sereact's $110 million Series B likely places it near the $1 billion unicorn threshold.
Key things to watch over the next 12 to 18 months:
- Can they sign 3 to 5 U.S. warehouse clients from zero?
- Can the intervention rate of 1 per 53,000 picks be reduced by another order of magnitude?
- Which approach proves more effective: the VLA + world model path, or Figure's end-to-end neural network?
Physical AI is burning far more capital in 2026 than last year. Sereact's $110 million is just the entry ticket.
Sources: Sereact Raises $110 Million to Scale AI Robotic Brain (PYMNTS); CocoLoop; Sereact $110M Series B announcement (company announcement).