The robot rental boom is real, and so is the catch

On 30 June 2026 CNN published a piece on China’s humanoid rental market. The headline framing was sceptical: the robots have captivated the world, and renting them out is exposing their limits. We think the article is right, and we think it makes the case for our own model rather than against it. Here is why.

What is actually happening

There are now more than 153,000 robot rental businesses in China, according to state media. Agibot has launched a rental arm of its own, SHAREBOT, and expects the rental market to reach roughly 1.5 billion dollars by the end of 2026. In its first three months SHAREBOT logged over 5,500 orders.

A humanoid rents for around 3,500 yuan — about 517 dollars — per day. Cheaper packages start near 3,000 yuan a day for exhibitions, events and, CNN notes, the occasional marriage proposal. Demand is not the problem. Demand is enormous.

The detail everyone skipped

The day rate includes shipping and a human operator who helps control and programme the machine.

That single line is the whole story. These robots are not being rented as autonomous labour. They are being rented as a capability that arrives with a person attached, because at this stage of the technology the person is what makes the capability reliable.

CNN reads that as evidence the technology is oversold. That reading is defensible. We read the same fact differently: it tells you what the actual product is. Nobody is selling a machine. Everybody is selling a machine plus the expertise to make it do something useful, and the companies pretending otherwise are the ones who will disappoint their customers.

We rent these machines too

It would be dishonest to write this without saying so. We rent humanoids by the week ourselves, with an operator on site when the deployment calls for one. The Chinese market and ours are selling the same thing at different price points.

What we try to be careful about is the difference between two uses that look identical on an invoice:

• Rental for attention. A trade fair, a store opening, a campaign. The machine needs to be present, hold an interaction and stay upright. This works today, reliably, and it is most of the market.
• Rental for output. A station, a cycle time, a measurement. This also works, but only after somebody has defined the task, and it belongs in a pilot with numbers attached rather than a week’s hire.

Selling the first as though it were the second is how this industry loses trust. It is also, we suspect, a fair description of some of what CNN found.

What it means in the Benelux

If you want a humanoid for an event, rent one. It is a solved problem, the machine is a catalogue product and the week is the right unit.

If you want one to do work, the honest sequence is different: define the station, run a pilot with a cycle time and a success rate, and only then look at a 36-month contract. Skipping that order is how a robot ends up in a corner of a warehouse being politely ignored.

The caveat

CNN’s conclusion is that this technology is still years from replacing human labour on a factory floor or in a household. On the household, we have no quarrel. On the factory floor, the six-day production-line livestream from Longcheer published the day before suggests the picture is more uneven than that: some tasks are ready now and most are not. Both things are true, and anyone selling you only one of them is selling something.