Geek+ P800R shelf-to-person robot

What the machine is

The P800R is a flat drive unit that lives under a shelf: 1,095 mm long, 830 mm wide, 195 mm tall, 142 kg on its own, and able to lift a full 1,000 kg rack. It moves at 2.3 m/s empty and 2.0 m/s loaded, navigates by inertial sensors plus QR-code fiducials on the floor to ±10 mm, and runs 1.5 to 2 hours on a 51.8 V / 27.9 Ah lithium pack.

Its job is to bring a shelf to a person. That is the whole product, and it is exactly why it belongs in this catalogue.

Why the P800R and not something else

We intended to let the prospectus decide by product-line revenue. That split does not exist. Geek+ reports revenue by geography (mainland China versus non-mainland) and by a single top-level solution family, and has never disclosed a P-series versus RoboShuttle versus Sorting breakdown. Treat any source that claims to know that ratio with suspicion.

Two things are established. Warehouse-fulfillment AMR solutions were 90.3% of Geek+‘s FY2024 revenue. And Geek+‘s global share of shelf-to-person is 48.5% against 23% of order-fulfillment AMRs overall — its share in the category it actually dominates is more than double its share of the broader market. The P800R is that category’s workhorse. Geek+‘s own case-study index agrees: the large majority of named enterprise deployments carry the P800 / shelf-to-person tag, with roughly three for RoboShuttle and two for PopPick.

Rejected alternatives: RoboShuttle RS8/RS5 is a halo product that collects height records and awards but very few named deployments; PopPick has solid installations (340 units at S&S Activewear) but is a workstation and system brand rather than a robot model; and P40 could not be found anywhere in Geek+‘s current lineup, so we treat it as unverified.

Three spec contradictions worth stating

All three are on Geek+‘s own website, and anyone fact-checking this entry will find them.

First, the P-series family page publishes one identical spec block for all four models — 600 kg payload, 170 kg weight, 60 mm lift, 950 x 702 x 275 mm — which is the P500R’s data applied to the P800R and P1200R, on a page whose own header text says “maximum load of 1200 kg.” It is a stale template. This entry uses the per-model table on the shelf-to-person page.

Second, a used-equipment listing for a 2020-build P800R gives 193 kg and 274 mm height against the current 142 kg and 195 mm, while length and width match. That is consistent with a hardware generation change rather than a contradiction, but the two sets of numbers must not be mixed.

Third, the family page also lists a “P800R V6” and a “slim version” with no separate specs, and Geek+ does not state which hardware revision the published table describes. Separately, Geek+ and Intel launched a vision-only navigation system in November 2024; whether it ships on the P800R is not disclosed. We record navigation as QR-code plus inertial and footnote the vision variant.

Charge time is not published by Geek+. We leave the field empty rather than import a number from an aggregator.

Whose work it takes

A roaming picker walks eight to ten miles a shift, and most of that time is not picking. It is walking and searching. The P800R deletes that time entirely.

So the direct exposure is order pickers and warehouse operatives, where Geek+ claims a 30%-plus reduction in labour need alongside 2-3x picking productivity. Replenishment associates are exposed by the same mechanism running in reverse — the shelf comes to the replenishment station. Inventory and cycle counters are the most completely displaced: a shelf-to-person system maintains a perpetual system-of-record inventory, and manual counting largely stops existing at deployed sites. Peak-season temporary staff are structurally exposed, because Geek+ markets the system explicitly as a substitute for seasonal hiring: throughput scales with robot count, not with recruitment.

One job is created rather than destroyed. The goods-to-person station operator survives — but transformed: fixed position, narrower motion, pace dictated by robot arrival, output metered per line. That is not a quantitative disappearance. It is a qualitative downgrade.

Pallet-jack drivers and material handlers are also being reduced, but by Geek+‘s M/F/X series rather than the P800R, and the entry should not claim that credit.

Why we care for LostJobs

Because a company that has been the world’s largest AMR vendor for seven consecutive years, has 72,000 robots deployed, and holds 48.5% of its core category had no entry in this catalogue at all — while several startups with one or two named customers did. That is the exact shape of the failure this catalogue is trying to correct: a long tail with a hole at the head. A reader who spots one obvious absence stops trusting everything else on the page.

There is a second reason, and it cuts against instinct. The P800R is not exciting. Shelf-to-person AMRs are a decade-old category and this is an incremental hardware line. That is precisely the point. The machine reshaping warehouse employment is never the newest one; it is the one that has already sold 72,000 units. Novelty is not a virtue.

The last word belongs to the evidence. This catalogue’s claim is that a sceptical reader can check it. So: every labour figure above — the 30% reduction, the 2-3x productivity — is vendor-supplied, and we found no independent study, no customer-disclosed headcount data and no third-party audit behind any of it. The Robot Report’s measured 913 items per robot per day at an average 0.44 m/s sits well below the marketed peak. Mistaking a vendor’s multiple for operating reality is the most common error in this industry. We would rather flag it than repeat it.