Teradyne Robotics and Sereact team up to put AI-driven robot arms in warehouses

Teradyne Robotics, the company behind Universal Robots (UR) and Mobile Industrial Robots, partnered with Sereact, a Stuttgart-founded physical AI company. The announcement came on October 21 in Berlin.

Physical AI is software that lets a robot understand the messy real world and act in it, not just repeat a fixed motion. Sereact's version is called Cortex. Under the deal, Cortex runs on UR robot arms.

The target is what the companies call unstructured tasks, such as handling returns and picking single items out of a mixed pile.

Why it matters: The companies say they have already shipped hundreds of UR robots in dual-arm systems to live customer sites across Europe. This is not a lab promise.

Teradyne Robotics makes UR's collaborative robot arms, often called cobots, which are built to work near people. UR says its platform is open, connected and proven through more than 100,000 industrial deployments.

Sereact was founded in Stuttgart in 2021 by Ralf Gulde and Marc Tuscher. Its software lets robots grasp, identify and handle objects they have never seen before.

The Robot Report notes Sereact opened a US headquarters in Charlestown, Massachusetts this summer. The company's own release describes it as a Boston headquarters.

Sereact's funding gives a sense of its size. The Next Web reports a $116 million Series B this year, and customers such as IKEA, DHL, BMW and Mercedes-Benz, according to the release it saw. FashionNetwork reports Zalando joined that round as a strategic investor in July.

Cortex is what the companies call a Vision-Language-Action model. In plain words, it looks at a scene through cameras, takes an instruction, and decides how the arm should move.

At the NVIDIA GTC Berlin 2026 show, Cortex ran on two UR12e-1300 arms in the NVIDIA booth. They showed a returns-handling job that is already running in live logistics centers.

Sereact CEO Ralf Gulde puts the pitch this way: "One brain serves every robot, with no retraining for each new item or task. And what one robot learns, every robot benefits from."

In real life Think of one very good warehouse worker who shares everything they learn with every other worker the moment they learn it. That is the idea behind one brain on many arms.

The joint system is deployed with CEVA Logistics for Zalando, the European fashion and lifestyle platform. It says it connects 62 million active customers with more than 7,000 brands across 29 markets.

At Zalando fulfillment centers in Greven, Germany, and Swiebodzin, Poland, the robots now process returned fashion items on their own. The companies call this the first phase of a broader plan to scale returns automation across Europe.

The Next Web reports the systems are sold as robotics-as-a-service, where the customer pays for the robots as a service instead of buying them. CEVA's Chris Walton says the model sorts items faster and with fewer manual steps.

Supply Chain Digital adds that Zalando says about 50% of its online sales are returned. That makes returns a very large pile of mixed items.

Gulde explains it directly: "Returned products almost never arrive in a robot-friendly sequence. Items differ in shape, orientation and packaging, and the mix changes continuously."

A robot programmed in advance expects a known item in a known spot. A crumpled jacket next to a shoebox does not fit that script.

Sereact's pitch is that Cortex needs no training for each product and no fixed list of articles, the companies said. Gulde told The Next Web that a system working only on items it has seen before "does not work in a returns operation at all."

The big picture: Sereact also says that by quickly reconditioning returned goods for resale, the solution helps retailers follow European Union rules that ban destroying or landfilling returned merchandise.

Returns are the first showcase, not the whole plan. The companies list a long set of jobs for the joint system:

• Picking and sorting: piece-picking and sortation, and order induction.
• Stores and stock: store replenishment and bin consolidation.
• Factory work: machine tending and kitting.
• Shipping: packing and goods handling.

Gulde says every one of these runs on the same production model. Sereact also gains reach: it says partnering with Teradyne Robotics and its global distribution and integrator network helps bring the system to customers around the world.

NVIDIA technology sits under the whole stack, The Robot Report reports. Cortex is trained in NVIDIA Isaac Sim, an open simulation framework, using NVIDIA Cosmos open world models, which means AI models that predict how the physical world behaves.

After that it is post-trained with reinforcement learning, a method where software improves by trying things and getting scored, using the NVIDIA Isaac Lab framework. It is benchmarked on Teradyne Robotics hardware before any model is deployed.

On the warehouse floor, Cortex runs on NVIDIA Jetson Thor at the robot cell, where decisions must happen at line speed. NVIDIA DGX Spark handles site-level development, validation and model updates.

Teradyne Robotics also said it is contributing SimReady OpenUSD robot assets for UR and MiR products to the SimReady Central Library on Hugging Face. That helps developers reuse robot models in simulation.

Sereact's own numbers, as Industrial Production reports them, are 200 systems in use across Europe and more than one billion picks. Around one in 53,000 operations needs human help.

Industrial Production lists Active Ants, Austrian Post, BMW, bol., Daimler Truck, DHL, Mercedes-Benz, PepsiCo and the Rohlik Group among its customers.

The company's Series B was announced as $110 million, led by Headline. FashionNetwork reports Zalando's investment brought the round to $116 million and total funds raised to over $145 million.

Cortex 2, the next version, is meant to first imagine possible actions with a world model, then pick the most promising one before the arm moves. Co-founder Marc Tuscher describes it as giving robots "a form of imagination."

CEVA's Chris Walton calls returns one of the toughest challenges in logistics, citing unpredictable volumes, manual quality checks, seasonality and a disproportionately high cost per unit, according to Supply Chain Digital.

The companies say the robots take over repetitive handling and reduce physical strain on workers. Staff move to supervising stations, handling exceptions and checking quality.

The next step is geography. Sereact opened its US headquarters this summer to bring the joint solution to North America, and the partnership gives it Teradyne's sales and integrator network.

The Zalando rollout is also only phase one. The companies say it will grow across Europe.

What neither company has published are prices, the number of robots per site, or targets for how many systems they plan to ship. Those numbers will show how fast this grows.

• Online shoppers: returned items may get sorted and back on sale faster, though the companies have not promised any change for customers.
• Retailers and warehouses: a robot that handles unknown items makes returns easier to automate than before.
• Robotics fans: physical AI is moving from demos into paid, live work.