Unmanned-delivery vehicles are suddenly everywhere in China, and the question for 2026 is whether this is real scale or another bubble pumped up by capital and industry hype. The trigger was not a 2025 technology break but an economic case that finally started to close.
Three years ago a single vehicle cost 200,000 to 400,000 yuan. Falling costs across the new-energy supply chain, domain controllers, lidar, batteries and e-drives, have since cut that sharply. One maker says per-unit cost is down over 50 per cent, below 10,000 US dollars. At the same time, road rights and rules are clarifying, letting vehicles move from campuses and parks into express, postal, supermarket, pharma cold-chain and tyre delivery.
Demand is real where it pays. Youjia Innovation says its fulfilment rate holds above 95 per cent and near 90 per cent even at peaks, because late fresh spoils and late medicine incurs penalties. Per-order cost runs 25 to 35 per cent below pure labour, and acceptance is higher at night, in bad weather and in remote spots. The customer mix is broadening from a few large logistics buyers to small merchants and individuals, and many are shifting from buying one vehicle to running fleets.
Map-less driving became the 2026 direction because high-definition maps once took months and hundreds of yuan per kilometre to prepare. Dropping them cuts map cost near zero and shortens city rollout. One firm expects over 40,000 kilometres of map-less L4 mileage by month end.
That maturity pulled in the carmakers. Chery made autonomous logistics a company-level track with six chassis platforms. Changan Kaicheng launched a logistics vehicle with JD Logistics. Wuling, GAC and King Long followed. Suppliers felt it first: ZhiXing Technology has shipped about 3,000 domain controllers for Geely’s Remote brand and serves Geely, Chery and Changan plus robot and logistics firms.
The new model is deep cooperation rather than mere licensing. Dongfeng and Jiuzhi launched the Dongfeng OpenVAN brand, with Jiuzhi supplying autonomy and Dongfeng building and selling. White Rhino and Xinyuan formed a joint venture after three months of mass operation. Reusing the new-energy supply chain lets hardware costs keep falling, and L4 software cost could drop to a tenth at scale, much as passenger ADAS pricing already fell to the 10,000 yuan level.
Not every carmaker commits. One executive says OEMs often deliver dozens of units, then stop when reality differs from the pitch. As more entrants arrive, competition shifts from algorithm to product stability, cost and operations. Pure-algorithm firms from passenger L2 and L3 must rebuild Robovan data from scratch.
The hardest problem is not only technical. Local standards vary, non-standard scenes are hard to adapt, and overseas compliance differs by region. Without a unified framework, a good solution cannot be copied quickly across cities and clients. Unmanned delivery looks closer to commercialisation than Robotaxi and easier to monetise than general robots, but it is not a light business. Vehicles running is only step one. Scale depends on cost, standards, operations and safety maturing together.
As Drucker put it, efficiency is doing things right, effectiveness is doing the right things. In 2026, co-build by many parties and a fresh split of labour between OEMs and L4 firms are turning unmanned delivery from startup experiments into a new commercial-vehicle category. The next test is whether unmanned transport can run safely on public roads for the long term, because no wind can hold up a fall in safety.
Editor’s note: This is a translated adaptation of a Chinese-language report from LeiPhone (leiphone.com). Figures, dates and direct quotations are reproduced as published.