Schaeffler published its third-quarter numbers this week. Inside them was a sentence that deserves more attention than the headline figures: management said the hype has cooled somewhat, but that activity related to actual delivery remains quite brisk.
The sentence was supported by data. Two new Chinese customers placed mass-production orders in the quarter. The humanoid order book rose to about EUR 400 million. Five customers are now in mass-production cooperation. Sample orders, the stage before commitment, reached 83.
Why a supplier’s order book is the better read
Against the valuations circulating in the sector, EUR 400 million is not a dramatic number. What makes it informative is its structure.
Eighty-three sample orders means a large population of programmes are still being tested and qualified. Five mass-production customers means a small group has crossed the line between whether a part can be made and at what rate it must be delivered. The combination describes a supply chain in the middle of a ramp, neither in a boom nor in a bust.
The second detail is stickier than the value. Existing customers are buying more categories: from a single transmission component up to modules, force and tactile sensors, and controllers. When a customer moves from buying one part to outsourcing several subsystems to the same supplier, the relationship stops being transactional. That is what a moat looks like in components, and it is harder to build than a framework agreement.
The two ends of the same quarter
The same week brought the opposite data point from the whole-machine side. Unitree’s first report as a listed company showed third-quarter net profit down 17 per cent, despite revenue growth of 70 per cent and a 250 per cent increase in research spending. Separately, Tesla was reported to be recruiting night-shift supervisors whose job is generating training data for Optimus.
Read together, the pattern is familiar from every capital-intensive industrial cycle. The companies assembling the finished product are spending on capacity, engineering and data, and their income statements carry that cost. The companies supplying the parts are paid on delivery. During the build-out, the supplier’s accounts are a cleaner signal of real demand than the assembler’s.
That logic has a limit, and it should be stated. A supplier order book tells you what is being built, not what is being sold. If the machines accumulating in factories and laboratories are not eventually deployed into paid work, the orders stop, and the supplier is the second party to find out.
The test is dated
Schaeffler is due to disclose delivery years and revenue recognition schedules on 3 November. That disclosure is more consequential than the EUR 400 million headline.
If the order book converts inside twelve to eighteen months, and if the Chinese share of those customers is substantial, then the mass-production inflection has supply-chain evidence behind it rather than presentation slides. If delivery is spread across three to five years, the valuation of the whole-machine segment has further to fall, because the deployment curve it assumes does not exist yet.
There is a broader point in the way the quarter was described. Hype and delivery are usually treated as a single variable that rises and falls together. Schaeffler’s management separated them, and the separation is the most useful thing anyone said about this market in October. The enthusiasm may be cooling while the purchase orders keep arriving. Investors who only watch attention will miss the second part, and so will investors who only watch shipments, because shipments without a delivery schedule are a promise about a factory, not about a customer.
Editor’s note: This is an adapted translation of the original OFweek report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://robot.ofweek.com/2026-10/ART-898890-12003-30709274.html.