The first thing you notice inside a cube storage warehouse is that there is nothing to see. There are no aisles, no forklifts, no people walking the floor. There is just a wall of aluminum grid and the sound of robots moving underneath it.
In a traditional distribution center, a broken forklift slows one aisle. The facility degrades gracefully. When you replace those racks with a high-density cube storage grid, the facility no longer degrades gracefully. You trade the localized failures of a traditional warehouse for the binary state of a single machine. A software lockup in the grid controller stops every pick in the building.
The business case for these systems is usually driven by real estate. Operations shows a slide demonstrating that the new grid will hold three times the inventory in the same square footage. The capital expenditure is approved based on construction savings. Then IT is handed the project to integrate the system, which is when everyone discovers they haven’t just bought new shelves. They have purchased a monolithic, proprietary appliance.
The storage calculation is the easy one. The difficult calculation is how many orders can leave the building when the robots are charging, replenishment is running, and half the requests are for bins buried low in the stacks. A bin-count guarantee is not a pick-rate guarantee.
Throughput in a cube system is a software property, not a hardware one. The grid capacity is just a number. Actual throughput depends on the warehouse control system scheduling, the charging policy, robot traffic density on the top rails, and how the order stream is batched. Two sites with identical hardware can run at wildly different rates.
The warehouse management system no longer knows where the product is. It only knows that the product is inside the block. The grid’s internal software decides exactly where a bin lives, constantly reshuffling it based on pick frequency. If the WMS continues to group orders poorly—asking for distinct, low-velocity items across disparate areas—the grid bottlenecks on digging operations, starving the human pickers at the ports. The WMS wave plan has to change. If you do not change the WMS, you will not see the throughput.
Adding robots to a saturated grid is like adding cars to a traffic jam to make the commute go faster. Every robot occupies a cell and needs a clear path. Eventually, they spend more time waiting for intersections to clear than moving bins.
Charging is another throughput variable that often gets treated as an energy decision. The policy that minimizes electricity cost is rarely the policy that maximizes throughput. When someone changes the charging policy to save on the utility bill, throughput can drop fifteen percent the next day, and it might take a week for operations to connect the two events.
When the system works, it is invisible. When it fails, the blast radius is the entire building. Inventory on hand and inventory accessible are different measures. Stock remains physically present during a grid-control outage, but staff have no practical way to retrieve a bin near the bottom of a sixteen-deep stack. You cannot throw bodies at a peak season problem when the problem is a solid cube of plastic and aluminum.
Redundancy in the warehouse control system is not the same as resilience. If the primary controller fails, the redundant system takes over, but failover is not instant and rarely clean. Reconciling the physical location of in-transit robots with the controller’s database delays recovery. Restarting the software does not reconcile inventory. The interface contract needs to describe what happens after a retry, not just the happy-path message. Ask the vendor to demonstrate recovery from a stopped shift, not only a clean startup.
Fire protection in a dense grid is a design-phase conversation, not a commissioning-phase one. Dense plastic bins, lithium batteries on the robots, and no aisles for a hose stream create a scenario that local authorities and insurers struggle to classify. A design accepted for standard shelving does not cover a dense grid. The fire marshal will ask what happens when a bin of aerosol cans ignites in the middle of the structure. A PowerPoint slide from the vendor will not satisfy them. The suppression design requires review by an independent fire engineer before the concrete is poured.
Cold chain operations make the physics worse. Battery performance drops, condensation forms on the grid, lubricants behave differently, and robot maintenance in sub-zero temperatures is a fundamentally different job. A temperate-zone reference site does not translate to a minus-twenty cold store.
You are not buying a warehouse. You are buying a fifteen-year dependency with a floor plan. The physical grid will outlive several software releases and hardware generations. The robots, the firmware, the spare parts, and the service contract all come from one vendor. There is no second source for a robot that only fits this specific grid.
The vendor’s roadmap becomes your roadmap. When the vendor announces the robot model in your grid is end-of-service in eighteen months, you buy the new one or you stop operating. The replacement might be a different form factor, requiring a partial grid rebuild. The original business case assumed a fifteen-year lifespan, but the exit plan is rarely part of that initial math.
The switching cost is the cost of demolishing the grid. If you cannot describe how you leave, you have not finished the analysis. What does decommissioning cost? Who removes the grid? What is the cost to revert the building to standard racking?
The warehouse control system is the hub. It talks to the WMS above and the robots below, plus the programmable logic controllers for conveyors, safety systems, and the building management system. Everything else is a spoke, and the hub is entirely vendor-controlled.
The WCS generates telemetry on robot utilization, charge cycles, and fault codes. Usually, the vendor owns this data. Negotiate export rights and a data schema before signing. You will need it to manage performance and to argue about service level agreement credits.
The integration surface is where projects stall. The WMS-to-WCS integration is often a custom middleware layer built by the vendor’s professional services team. When the vendor’s lead engineer leaves, the knowledge leaves. Require the interface specification, the error handling model, and the source code for the middleware, or at least a clear path to maintain it.
The storage density business case is easy to build. The availability, integration, and exit business cases are hard. Put the warehouse control system in the architecture review, the disaster recovery plan, and the change management process. Treat the vendor relationship as a permanent marriage, because replacing the vendor means closing the building.
The building may keep the grid longer than it keeps the software that first ran it.
Take it with you
Takeaways
- Throughput in a cube system is a software property, not a hardware one
- Charging policy changes can silently cut throughput by fifteen percent
- Fire protection requires independent engineering review before construction
- The exit plan and decommissioning cost belong in the original business case
Asked and answered
Questions
Why does throughput vary between two sites with identical cube storage hardware?
Throughput depends on the warehouse control system scheduling, charging policy, robot traffic density on the top rails, and how the order stream is batched. Two sites with identical hardware can run at wildly different rates.
What happens to inventory access when the grid controller fails?
Inventory remains physically present, but staff have no practical way to retrieve a bin near the bottom of a sixteen-deep stack. You cannot throw bodies at a peak season problem when the problem is a solid cube of plastic and aluminum.
Why does the WMS need to change when a cube storage system is installed?
The WMS no longer knows where product is, only that it is inside the block. If the WMS continues to group orders poorly, the grid bottlenecks on digging operations, starving the human pickers at the ports.