Sereact Raises $110M Series B for Robots That Predict Their Own Actions

Sereact, a German robotics software company that builds AI brains for warehouse and manufacturing robots, announced a $110 million Series B funding round on Monday, led by Headline with participation from Bullhound Capital, Daphni, and Felix Capital. Existing investors Air Street Capital, Creandum, and Point Nine also wrote follow-on checks.

The round funds two parallel pushes: scaling Cortex 2.0, the company's new robot foundation model, and entering the United States, where Sereact will open its first US office in Boston this summer. Customers already running Sereact's stack in production include BMW, Mercedes-Benz, Daimler Truck, PepsiCo, and Austrian Post.

What Cortex 2.0 Actually Does

Most current warehouse robots are programmed for narrow tasks: pick this part, place it here, repeat. Cortex 2.0 takes a different approach. It combines a vision-language-action (VLA) model with a world model that simulates the physical consequences of an action before the robot executes it.

In practice, that means a Sereact-equipped picking robot can look at a bin of unfamiliar objects, hear a plain-language instruction like "put all the fragile items in box A," mentally run several candidate motions, score which is least likely to drop or damage anything, and then move. The "world model" part — the internal simulator the robot uses to predict outcomes — is what Bloomberg's headline calls "predicting consequences."

The company says more than 200 of its systems are deployed across Europe, completing over one billion production picks. Roughly one in 53,000 picks requires remote human intervention. That intervention rate is what's pulling enterprise customers in: at warehouse scale, anything above one in 10,000 typically forces a human-in-the-loop architecture that wipes out most of the labor savings.

Why World Models Matter for Robotics

Robotics has been the laggard of the AI boom. Language models scaled by training on internet text. Image generation scaled by training on image-caption pairs. Robotics has nothing comparable — physical interaction data is expensive, slow, and varies enormously across hardware.

World models are the current best bet for breaking that bottleneck. The idea is that if a robot can run an internal simulator of its environment, it can practice in simulation, learn from imagined experience, and reduce the amount of real-world data needed for new tasks. Google's Genie 3, Wayve's GAIA, Nvidia's Cosmos, and now Sereact's Cortex 2.0 all sit in this category.

Sereact's specific edge is integration. Most world-model work has been published as research demos. Sereact has run its system in BMW factories and Austrian Post depots for over a year and built up the operational data — pick failures, edge cases, environmental noise — that turns a research idea into a deployable product.

Who This Affects

The most direct competitive pressure is on Covariant (acquired by Amazon in 2024), Symbotic, and Boston Dynamics' picking platform. All three target the same warehouse and 3PL customers Sereact wants to take in the US.

The deeper threat is to the closed, single-vendor stacks that warehouse robotics has historically run on. Sereact ships its software as a model layer that can run on top of robot arms from multiple manufacturers — closer to "robot operating system" than "vertically integrated robot." If that approach wins, Cortex 2.0 becomes infrastructure the way AWS became infrastructure.

For US labor markets, scale-out timing matters. Warehouse worker shortages have been the stated motivation for major robotics deployments at Amazon, Walmart, and FedEx. Sereact's planned Boston office puts it within recruiting distance of MIT robotics talent and within sales distance of Massachusetts and New England 3PL operations.

What Industry Watchers Are Saying

Air Street Capital's Nathan Benaich, an early investor and one of the most-cited voices on European AI, wrote that Sereact represents "the first European robotics company that can credibly compete with US peers on both technology and deployment scale." Bloomberg's coverage led with the "robots that predict consequences" framing, which is likely to become the company's marketing hook in the US launch.

The $110 million round size is notable. European robotics has historically been underfunded relative to US peers — Sereact's Series B is among the largest robotics rounds ever raised in Europe and signals that European generalist VCs are now willing to write US-scale checks for category leaders.

What's Next

The Boston office opens this summer. Expect US customer announcements through the second half of 2026, particularly in 3PL, automotive parts, and food and beverage — the verticals where Sereact's European customer reference list translates most directly.

Cortex 2.0 itself will see further model iterations through 2026. The company has hinted at expanding beyond pick-and-place into assembly and kitting, both of which require longer-horizon planning that pushes the world model harder.

For developers, the more interesting watch is whether Sereact opens any of Cortex 2.0 to third parties via API. The company has not committed to that path, but several investors on the cap table — particularly Air Street and Creandum — have publicly favored open ecosystems.

The Bottom Line

Robotics is finally getting the AI tailwind that everyone has been predicting for two years, and Sereact is one of the few companies with both deployment scale and a credible world-model approach. The $110 million round buys it the runway to compete in the US against Amazon-owned Covariant and Symbotic — and tests whether European robotics can become a global category leader rather than a research showcase.