Warehouse Humanoid Robots: Deployments, Pilots and Evidence

Warehousing is the proving ground of humanoid robotics: structured racks, repeatable totes, measurable throughput, and chronic labor shortage make it the first economics that can plausibly close. That is why the field's most-cited deployments are logistics deployments. The honest ledger still has three columns — verified production, pilots, announcements — and the platforms below should be read against it, not against launch-event footage.

Why warehouses come first

The task set (tote moves, parcel induction, trailer unload assistance, kitting) tolerates current cycle times and grip reliability; the environment is indoors, mapped, and safety-fenceable; and the alternative — unfilled shifts — is a cost every operator already feels. A humanoid that holds even modest throughput across a full shift competes on a spreadsheet, which is the only stage that matters.

Reading a warehouse pilot honestly

Ask of every pilot claim: named customer or anonymous; independent observation or maker footage; duration and shifts; throughput stated or implied; what happens at pilot end. The profiles record what the public evidence establishes for each — and mark the rest INSUFFICIENT rather than assuming success.

What the evidence does not yet establish

Public records rarely disclose sustained cycle-time economics, intervention rates, or total cost per unit moved versus conventional automation. Absence of that data is not proof of failure — it is a gap the tracker will fill as disclosures arrive, and this site will not fill it with guesses meanwhile.

Frequently asked questions

Which humanoids run in warehouses today?
The deployment tracker holds the current verified list by platform — logistics is where several platforms have their strongest evidence; check states and dates there.
How many robots are deployed?
Where counts are publicly established they appear with sources; where only 'fleet' language exists, the record says the number is not established. Beware precise-sounding totals without provenance.
Should our operation pilot one?
Scope it as a project: one task lane, success metrics, exit criteria, and the verification questions from the for-sale tracker. The public record here arms the diligence.

Related on Physical AI Radar

Last reviewed: 2026-09-02 · Data as of the dates shown on each linked profile · Methodology · Corrections