Sereact Zalando Returns Robots Go Live Amid Thin Revenue

What went live

Zalando has moved AI-driven robots into one of the messiest parts of online fashion: returns. Dual-arm robotic systems from Stuttgart-based Sereact are now in live operation at logistics sites run by CEVA Logistics in Greven, Germany, and Świebodzin, Poland, handling items that customers send back to the retailer.31 The three companies describe the go-live, announced October 7, 2026, as the first operational milestone in a strategic partnership meant to scale automated returns handling across Europe.3

The robots run on Cortex, Sereact's "physical AI" platform. They grasp, identify and sort returned garments without per-item training or fixed SKU profiles.31 That detail is the core of the pitch. Returns arrive in unpredictable shapes, packaging and condition, which makes them a harder problem than outbound fulfillment. Zalando had already used AI-powered robots for millions of fulfillment picks before extending the technology to returns.1

The systems are delivered under a Robotics-as-a-Service model. The companies frame the robots as taking over repetitive, physically straining work, while warehouse staff shift toward station supervision, exception management and quality control.3

The money behind the launch

The partnership has a financial side as well as a technical one. Zalando joined Sereact's Series B as a strategic investor in July 2026.3 An independent teardown of the company puts total funding at roughly $146 million:2

• a $5 million seed round in August 2023
• a €25 million Series A in January 2025
• a $110 million Series B in April 2026
• an extension to $116 million when Zalando came in

No valuation has been disclosed.2

Revenue is the thinner part of the picture. Sereact has not published figures. One third-party estimate puts 2025 revenue at about $6.4 million on a 68-person team, and the teardown flags that number as unverified and only directional.2 Headcount has since grown to around 93 by June 2026. The company is also building a US presence out of the Boston area, starting with about 10 hires and aiming for around 60 within a year.2 Sereact says more than 200 systems are live across Europe, and it counts BMW and PepsiCo alongside Zalando as deployments.2

Taken together, the numbers describe a company raising money faster than it is booking revenue. That is common for venture-backed robotics firms. It also means a lot rides on whether flagship deployments like the Zalando returns lines can turn into repeatable, high-volume contracts.

The foundation-model bet

Sereact positions itself as a robot "brain" company, not a hardware maker. Its products include PickGPT and Cortex 2.0, which it describes as a vision-language-action (VLA) foundation model.2 The thesis is that one hardware-agnostic model can generalize across products and environments.

The teardown names the central open question: can such a model handle an ever-changing mix of SKUs without the company being dragged into costly per-site services work?2 Fashion returns are a good stress test. If Cortex really sorts unfamiliar garments without bespoke training, as the launch materials claim,3 the economics of RaaS improve sharply. If each new site still needs heavy engineering support, margins will look more like a systems integrator's than a software company's.

Where the accounts agree and differ

The trade-press account and the joint announcement tell the same operational story: same sites, same partners, same no-per-item-training claim.13 Both emphasize capability and workforce benefits. Neither offers performance metrics such as throughput, accuracy rates or the share of returns handled autonomously versus routed to people.

The teardown adds the commercial context the announcements leave out. It shows a modest, unverified revenue base against substantial capital raised, and an unresolved question about services intensity.2

Our read

This deployment matters more as a signal than as a scale event. Two sites are a beginning, not a network. Still, Zalando putting money into Sereact and then putting its robots on returns is a meaningful endorsement from a demanding customer. It also gives Sereact a showcase for a high-variability task that rule-based automation struggles with.

The risks are financial and competitive. Robot foundation models are drawing heavy investment industry-wide, and a company with roughly $146 million raised and single-digit-million reported revenue will need to prove its model generalizes before better-capitalized players close the gap. The next things to watch are how quickly the Zalando partnership expands beyond Greven and Świebodzin, and whether Sereact starts publishing hard performance data.