The ten-thousandth unit came off the line on a Sunday, and the company put out a short announcement that evening. The factory was ramping, yields were good, and two models had reached 10,000 units since the September launch.
The same company's third-quarter report contains a sentence that reads oddly beside that number. Through the end of September, the consumer-robot business had recognised no revenue.
Both are accurate. They count different things, and the distance between them is the most useful thing a foreign reader can take from China's consumer-robot rush.
Key takeaways
- Swancor Advanced Materials (上纬新材) announced on 8 November 2026 that its Qiyuan (启元) Q1 and T1 robots had reached 10,000 units off the line.
- Swancor's third-quarter report shows nine-month revenue of RMB 1.209 billion (≈US$170m / HK$1.33bn) and a net loss of RMB 422 million (≈US$59m / HK$464m).
- Consumer-robot R&D reached RMB 360 million of a RMB 384 million (≈US$54m) total; the robot business had booked zero revenue through 30 September.
- Customers had prepaid RMB 210 million (≈US$30m / HK$231m) by the end of June; a prepayment is a liability, not revenue.
- The company's 2026 incentive plan requires new-business revenue to reach 25% of the group total this year.
Three counters, three questions
Units off the line (下线) is a production number: what the factory finished. Delivered is a logistics number. Recognised revenue (确认收入) is an accounting number, and it waits until control of the goods passes to the buyer.
Prepayment (预收货款) sits to one side: cash received before delivery, booked as a contract liability (合同负债). A large prepayment balance is evidence that people want the product, not that the product works or that the money will stay.
The sequence lets a company report 10,000 units, a nine-figure deposit base and zero revenue in one period without any being false. Consumer robotics is unusually exposed to that gap: the unit ships with an app, a warranty and a promise of future skills, and acceptance can drag.
What the filing shows
Nine months to 30 September: revenue RMB 1.209 billion, down 5.51%, with the third quarter alone at RMB 406 million, down 18.04%. The attributable net loss was RMB 422 million.
The legacy business is not the problem: new materials — corrosion-resistant resins, wind-blade materials, composites — earned RMB 63.5 million in net profit, up 4.91%, on a gross margin that improved 0.2 points to 14.42%. The loss comes from the new line: research spending rose 1,182% to RMB 384 million, of which RMB 360 million went to consumer robots. By 30 September the robot team numbered 295 people, with 135 patent applications filed, 67 of them inventions.
Contract liabilities had already jumped from RMB 4.32 million at the end of 2025 to about RMB 188 million by June.
At the base list price of RMB 19,999 (≈US$2,800), 10,000 units would be worth about RMB 200 million if every unit sold at list and every sale were recognised. That is about 16.5% of nine-month group revenue, against a full-year target of 25%.
Why the count runs ahead of the cash
Production is the easy part to publish. The company claimed a two-and-a-half-minute cycle at launch, daily capacity up from "more than ten" to "more than 200" by 25 October, and 240 units a day by the 9 November ceremony. Preorders opened in August, the first store opened in Shanghai on 28 September, and stores now run in Guangzhou, Shenzhen and Xi'an.
None of that is revenue until a unit is delivered and accepted, and the company said deliveries began only in October, after the reporting period closed. That is the whole gap: a timing artefact, not a contradiction.
Where it goes next is partly a channel question: units can sit in stores and with education partners before a customer accepts one. The company has been signing distribution, including an agreement with the iCar group at the IFA trade fair in Berlin in September 2026 for overseas channels (International Financial News, 9 November 2026).
What the company is measured on
On 14 August 2026 the company published a restricted-share incentive plan whose targets concern only the new business: revenue at no less than 25% of the group total in 2026, up at least 160% in 2027 and 700% in 2028 against 2026, with 100, 200 and 300 new invention applications. The first grant went out on 18 September at RMB 91.85 a share to 345 recipients.
A target written as a share of revenue is a deadline in disguise, and it explains why a unit count is the number a company reaches for: it is the only figure publishable without waiting for accounting.
For the wider market, IDC is the corrective: the firm counted about 18,000 humanoid units shipped worldwide in 2025, up 508%, with 36.8% going to entertainment, 24.6% to research and education, 17.0% to data collection and 9.2% to manufacturing (IDC, cited by Beijing Business Today, 22 September 2026). Nearly four in five went to a stage, a classroom or a data-collection centre. A consumer product changes who the buyer is, not what the count measures.
Honest limitations
The 10,000-unit figure is a company production count, published by announcement and not audited; it counts units leaving the line, not units in use. The 240-a-day rate comes from an executive at a ceremony, not from a filing, and the same line was described as "more than ten" a day a month earlier, so the ramp is real but thinly documented. No sell-through, return or field-utilisation data has been disclosed, so channel inventory is unknown. The 25% target is an incentive condition, not a forecast, and the company says it cannot predict the robot business's effect on 2026 results.
What readers can do now
- When a Chinese robot maker quotes units, pull the contract-liability and revenue lines for the same period. A rising prepayment balance with flat revenue means the constraint is delivery, not demand.
- For procurement, ask for field-use hours after 90 days and a repair record rather than a shipment certificate; the IDC split shows most units so far are demonstrated, not deployed.
- For analysts, track the gap between units off the line and recognised revenue across two consecutive quarters. A widening gap is inventory in the channel; a closing one is a product.