--- title: "China's newest sorting hub can move 10 million parcels a day. The robots were the easy part" date: 2026-11-03 category: Humanoids & Robotics site: NeuroAI canonical: https://neuroai.site/a/na-robot-parcel-hubs-sorting-automation language: en --- # China's newest sorting hub can move 10 million parcels a day. The robots were the easy part > J&T's Jieyang hub can move 10 million parcels a day on cross-belt sorters and six-sided scanners. Parcel volumes rose 4.3%; automation is now about margin. At two in the morning the loudest thing in the building is not a machine. It is cardboard sliding on steel, again and again, as a belt carries boxes past a scanner that reads six faces at once and flicks each one down a chute. Vans back into the bays. People stand at the induction points, lifting sacks, cutting ties, feeding the belt. Everything after that happens without them. ## Key takeaways - J&T's Jieyang hub, which opened on 30 October, is designed for up to 10 million parcels a day and averages about 7 million, on 258 mu (≈ 17 hectares) of land. - It runs 13 classes of equipment — six-sided scanners, cross-belt sorters, conveyor lines — 24 hours a day. - Parcels grew 4.3% in the first nine months of 2026 while revenue grew 7.2%, so margin now comes from cost per parcel. - The jobs that automated first are those where the object is a box and the destination a barcode. - Humanoids are being trialled in the same buildings, but the work that pays today is done by fixed conveyors and robot arms. ## A hub built for the peak, not the average J&T Express (极兔速递) opened the Jieyang transfer centre on 30 October, in Jiedong district, Guangdong. It is the company's largest smart logistics project in eastern Guangdong: three double-deck operating platforms, a cloud warehouse and a fully automated sorting system, with peak capacity of 10 million parcels a day and average throughput of about 7 million. J&T also says its rebuilt Chongqing campus, reopened on 25 October, runs at "100% automation coverage", with labour efficiency up 20% and space efficiency up 220%. Those are company figures, but the logic is visible in the building. A cross-belt sorter does not get tired, does not need a night-shift allowance, and does not slow down in bad weather. It also cannot pick a bottle off a shelf, which is why the two ends of the parcel business look nothing alike. ## Why parcels automated before everything else Sorting is the friendliest possible problem for a machine. The objects are mostly rectangular, arrive in a predictable orientation, and carry a barcode that states where they should go. The environment is fenced, lit the same way all night, and free of customers. Compare that with what humanoid makers advertise: unloading a mixed pallet, folding a shirt, restocking a shelf. Each requires deciding what the object is before deciding what to do with it. That gap explains this year's pattern: fixed automation is being installed at scale while general-purpose robots are still being demonstrated. JD Logistics says its "Yilang" (异狼) dexterous arm has handled more than 15 million parcels in warehouse sorting, with 99.9% tracking accuracy — an arm on a fixed base doing one narrow job, and the closest thing the industry has to a proven deployment. Cainiao's climbing robot, ZeeBot, works the same seam: it runs at up to 4 metres per second and climbs to the top of a five-storey shelving unit in about 10 seconds, using space previously left empty. Cainiao delivered two such apparel warehouses, in Tianjin and Suzhou, before Double 11. ## The money moved from volume to cost For a decade, Chinese express delivery was a volume game. Companies cut prices, and parcels arrived. That arithmetic has stopped working. State Post Bureau data released on 27 October shows 151.3 billion parcels in the first nine months of 2026, up 4.3%, and revenue of 1.16 trillion yuan (≈ US$164B), up 7.2%. In September, average revenue per parcel was 7.81 yuan (≈ US$1.10), up 3.5% year on year. Revenue growing faster than volume is unusual here, and it is partly regulatory: Beijing has spent two years suppressing below-cost pricing. But it also changes what a sorting hub is for. When you cannot win on price, you win on cost per parcel, and cost per parcel is set by how many people you need at three in the morning. ## What still needs hands The automated part of a hub begins after induction and ends before the last mile. Everything between — unloading, opening, feeding, handling damaged or unlabelled items, loading outbound cages — still involves people, and that is where the labour cost sits. A parcel that arrives crushed, wet, or with a torn label has no barcode to read and no fixed shape to grip. A worker handles it in four seconds; a general-purpose robot must recognise the problem, choose a grasp, and often ask for help. Humanoid pilots at logistics sites are worth watching, not yet counting. Several Chinese makers have put bipeds into parcel halls, mostly on night shifts and defined stations. The commercial question is whether a machine costing more than a year of a worker's wages can run enough hours to pay that back. At a hub that already sorts 7 million parcels a day, a handful of humanoids is a rounding error. The conveyors are the business. ## Honest limitations Throughput and efficiency figures here are company statements. J&T does not publish headcount at Jieyang, so the labour saving cannot be calculated from outside, and "100% automation coverage" is not a defined term. I have not seen audited cost-per-parcel data for any of these hubs. The claim that automation is now a margin defence rather than a growth play is my reading of the State Post Bureau numbers, not a disclosed company position. The humanoid pilots in parcel halls are described only in aggregate; no company discloses unit counts, uptime or cost per hour, so that section stays qualitative. Warehouse picking, line-haul trucking and last-mile delivery have different economics; last-mile is where most of the labour still is. ## What readers can do now - **If you are sizing a robotics market**, separate fixed automation from general-purpose robots. The first is already a large, boring, profitable business; the second is a bet. - **If you run a fulfilment operation**, measure your induction and outbound labour before buying anything. That is usually where the cost is, and it is the part vendors do not quote for. - **If you are following Chinese logistics**, watch unit revenue per parcel rather than volume. It tells you whether automation is being funded by growth or by cost pressure. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-robot-parcel-hubs-sorting-automation Free to quote with attribution and a link to the original.