The Years After the Sale
The years after a robot is sold are a business of their own. Whether that business pays for itself is decided by how hard the machine is worked and how it is sold, and the audited filings of companies that already run such businesses straddle the line.
In one paragraph
A durable aftermarket annuity is the story everyone tells about robots: sell the body once, then earn on spares, service and software for the machine’s life. This study tests that story against the audited filings of companies that already run such businesses. The record does not support it as the default. The dominant sale instrument is an outright capital sale on a public procurement register, with a lot value, a delivery month and no term, and no recurring fee. The parameter that decides whether the years pay is utilisation: about 57 operational hours per robot-month observed at a named automotive deployment, against 520 assumed on page 41 of a vendor’s investor deck, two readings the study keeps strictly apart and never averages. The verdict is a location, not a value: the machine’s years land on a bench of resolved service businesses, on one side or the other of the line where a service line covers its own direct cost.
What it establishes
- 01
The razor-and-blades prior is named and rejected on the record. The modal instrument is an outright capital sale on a government procurement register, with a lot value and a delivery month, no term and no recurring fee; the largest disclosed order book records no lessor income at all, and the largest shippers keep rental below half of one per cent of revenue.
- 02
Utilisation decides whether the years pay. The best publicised humanoid pilot divides to about 57 operational hours per robot-month on the two-robot reading; a vendor’s SEC-furnished investor deck assumes 520. One is measured arithmetic on published inputs, the other a model’s assumption, and the gap is the study’s central parameter.
- 03
The recurring line is genuinely straddled by resolved businesses. AITX’s device-rental line runs at a 71.4 per cent group gross margin and ABB’s service margin beats its product margin by 5.91 points; Knightscope’s service gross margin is negative 54.7 per cent for a second consecutive filed year, Serve Robotics posts a company gross loss of about 580 per cent of revenue, and Geek+ deliberately scaled its rental line from 14.1 per cent of revenue to 0.08 per cent.
- 04
The verdict is a location, not a value. In year one, sold outright and still under warranty, there is no service line to have a margin. In a later year, worked harder and sold on recurring terms, the position sits between ABB and CYBERDYNE, and the margin percentage falls out of that position.
- 05
A reader finishes able to make two calls: whether to build, buy or contract for a humanoid field-service estate, and whether to negotiate parts, diagnostics, firmware and calibration access as a condition of the component design-in. The second call is the one this record prices most clearly.
Download the full study.
20 pages, general industry report. No form, no address, no mailing list — the file downloads directly.
Run the study’s own world.
Set the machine’s hours a month between the observed and the assumed anchors, choose the sale instrument and whether its fee is flat or indexed to hours delivered, price a wear event and a technician hour, and drag the warranty’s edge across the machine’s five years. The model returns the cost of serving against what the instrument bills, the service margin, and where that year lands on a bench of resolved service businesses, or that there is no service line to place.
Both the study and its simulator are open. Nothing on this page is behind a form, because a form that hands over a PDF to anyone who types an address is a lead-capture device and this is not that.
Read it here
The whole study, inline. Every factual statement carries a superscript reference to a verified source list, and each figure and table carries the claim ids behind every value it prints.
This browser will not display the PDF inline. The file itself is unaffected.
Download the PDFThe Best Temperature Is Taken
Its world opens on a definite verdict — 900 °C on iron ore, parcel C-105, opening about 2036 — where this study’s opens on an undefined one. One world’s difficulty is making “there is no answer yet” inviting; the other’s is making a definite answer that must never harden into a legal conclusion.
Both worlds were built on the same discipline: parameters load at evidenced central values, one gesture returns to them, range ends are published extremes, and any other configuration is labelled as the reader’s scenario.
They behave nothing alike, because their subjects do not. That contrast is the strongest thing the pair can show.
This study is a showcase of Quorum, not the whole of it.
It summarises a small part of a much larger body of research. Every factual statement in it carries a reference to a verified source, and every figure carries the claim ids behind each value it prints. If your matter is adjacent to this one, Quorum can be pointed at it.