Humanoids get the headlines; the robots that already pay for themselves get a paragraph in a 10-K. Symbotic's FY2025 annual report, pulled via EdgarBeast and filed at sec.gov, describes the actual mechanism: incoming pallets are transferred to 'singulating' robots that deconstruct the pallet layer down to the case level.
That word — singulating — is the whole business. A pallet arrives as a stacked, mixed block of cases. Before anything can be stored densely or assembled into a store-ready outbound pallet, the system has to break that block into individual, addressable cases. Singulation is the step that converts a warehouse from a place that moves pallets into a place that moves cases, which is where fulfillment economics actually live.
Notice what the filing does not do: it does not promise general-purpose dexterity. The robots described are purpose-built for a constrained task — deconstruct a pallet to the case level — in a structured environment the company controls. That constraint is a feature. The ROI of warehouse automation has always come from doing one repeatable material-handling task reliably, not from a robot that can do anything.
This is the right frame for the whole humanoid-versus-incumbent debate. The case-level singulation Symbotic documents is shipped, deployed, and revenue-generating today, in a tightly bounded problem. A general humanoid in an unstructured warehouse is a much harder problem that the structured-robotics incumbent has quietly routed around by reshaping the environment instead of the robot.
Read the sec.gov description and the takeaway is unromantic and correct: the warehouse that works does so because it turned a hard, open-world manipulation problem into a narrow, structured one. Backlog is the only honest demo, and the mechanism behind it is singulation. Filing surfaced via EdgarBeast.