Hai Robotics to Deploy 1,500+ Rack-Climbing Robots in One European Fulfillment Center

The scale is unusual even for large e-commerce facilities. Hai Robotics specifies a 30,000 m² automated operation, 1.2 million double-deep storage locations and designed system throughput above 24,000 totes per hour. The customer is not named. Hai only describes it as a leading European fashion retailer that already operates multiple HaiPick systems across Europe.

This is not a small pilot: the customer is expanding an existing relationship

The most important part of the announcement is not the robot count alone. The retailer is not starting with a dozen units in a laboratory. According to Hai Robotics, it already runs multiple HaiPick systems in Europe and is now expanding the relationship to more than 1,500 HaiPick Climb robots. That reduces the likelihood that this is merely a technology demonstration purchased as an experiment.

We still do not know the full go-live schedule, deployment phases or contract value. It is also unclear whether all 1,500+ robots will operate simultaneously from day one. RoboMorrow therefore describes this as a contracted large-scale deployment, not as an already proven throughput record.

How HaiPick Climb works

HaiPick Climb belongs to the ASRS/goods-to-person family. Robots move within the racking structure and bring totes to workstations instead of sending workers on long picking routes. For this project Hai Robotics specifies a double-deep configuration, storing totes two positions deep to increase storage density while making access logic more complex.

This is different from a conventional floor AMR carrying a rack. HaiPick Climb uses rack-climbing robots and is designed as an integrated storage-and-tote-flow layer. For buyers, individual robot speed is not the main comparison metric. Layout, ports, buffering, WMS/WCS integration, redundancy and peak behavior matter at least as much.

24,000 totes/hour is a system figure, not a per-robot number

Warehouse automation announcements often blur individual-unit performance and installation throughput. Here, the 24,000+ totes/hour figure refers to designed throughput for the complete center. It does not mean one HaiPick Climb processes thousands of totes per hour. Performance comes from the combination of a very large fleet, racking, workstations, task queues and orchestration software.

This distinction also matters when comparing the project with smaller installations. Hai Robotics markets HaiPick Climb as scalable, but actual results depend on the process. With 1,500+ robots, the real test will be maintaining high throughput through failures, maintenance, charging and order peaks rather than only reaching a maximum slide-deck number.

1.2 million locations: density may matter as much as speed

For European e-commerce, real-estate cost and access to large logistics sites are major parts of TCO. Double-deep storage can increase tote count within the same building volume, but it requires sophisticated access and slotting logic. At 1.2 million locations, poor slotting or queue management can affect the result more than a small difference in robot travel speed.

That makes this project interesting as a system benchmark. At this scale, value shifts from individual hardware toward orchestration, software and integration. It connects directly with RoboMorrow's AMR vs AGV 2026 guide.

What remains unknown

• The customer is unnamed. Hai Robotics says only “leading European fashion retailer.”
• The location and country are undisclosed.
• Contract value, CAPEX/OPEX and payback are unknown.
• The full go-live date and ramp schedule are unknown.
• There is no independent audit yet of 24,000+ totes/hour in live operation.
• The remaining manual share is undisclosed at goods-to-person workstations and exception handling.

Why this matters for Europe and Poland

Europe is already a mature warehouse-automation market, but more than 1,500 robots in one facility shows goods-to-person projects reaching a scale previously associated with the largest sortation and automated-storage installations. The story is relevant to Poland because Poland remains one of Central Europe's major logistics hubs and competes for large e-commerce and fashion-fulfillment investments.

That does not mean every warehouse should copy this architecture. At lower volumes, a simpler AMR fleet, conveyors or partial automation may deliver better TCO. The Hai Robotics project is useful precisely because it shows the high end of the market: a very large installation where density, throughput and central control must work as one system.

How to judge the system after go-live

The best test will not be robot count by itself but availability of the complete system. In a fleet above 1,500 units, individual failures are inevitable, so the architecture needs to isolate problems, redistribute work dynamically and keep the facility operating without cascading shutdowns. Useful metrics will include real uptime, technician interventions, peak throughput and recovery time after a workstation or racking-zone failure.

The second dimension is economics. Without contract value, there is no reliable way to calculate cost per storage location or cost per additional tote/hour. If the customer or Hai Robotics later discloses CAPEX, service cost or ROI, the deployment can be compared with conventional ASRS, shuttle systems and large AMR fleets.

What qualifies as the next material update

The next meaningful update should be a named customer or location, full go-live, real operating metrics or disclosed contract value. Another media outlet repeating the “world's largest” phrase would not be a new event.

For now, RoboMorrow's editorial decision is straightforward: publish as a standalone deployment-scale story, while clearly labeling the record claim as Hai Robotics' statement and the throughput figures as designed system parameters awaiting independent operating validation.

Sources

Hai Robotics — official 8 Sep 2026 announcement · Hai Robotics — HaiPick Climb · RoboMorrow — AMR vs AGV 2026