JD Logistics unveils 'SuperBrain + Wolf' robot army: 300m units in five years

The Analysis

JD Logistics used its 9 September JDD conference to show the full SuperBrain plus Wolf fleet: nine product lines, 11 robots, and five new models launched the same day, from the warehouse-picking Canglang to the minus-20C cryogenic Zhiwolf, the 24/7 unmanned pharmacy Mulang, the L4 delivery van Dulang and the fully autonomous drone Feilang L05. Under the SuperBrain 3.0 model the fleet moves from single-unit operation to cluster coordination.

The number that matters is the procurement plan behind it: 3 million robots, 1 million unmanned vehicles and 100,000 drones inside five years, which is 4.1 million units across three categories. Read that line carefully. A five-year procurement target is a capital plan, not a service you can book. Nothing in the announcement says which nodes get which equipment, on what date, at what tariff, or how much integration burden is passed to the merchant.

The five briefs below take different decisions. Exporters need a quote that survives a handling-cost change and know which side of the Incoterm carries it. Import e-commerce sellers need a real cost per order, not a launch date. Factories need to know which of the 11 robots can survive contact with their cartons. Brands need a delivery promise that does not rest on a capability dated years out. Procurement needs contract language that converts a five-year plan into dated, priced milestones.

For Exporters

• By 15 September: get a node-by-node list marking which automated lines serving your SKUs are live today and which are not, with installation dates.
• By 20 September: obtain fee schedules split manual versus automated, quoted per carton and per order, with each node's dispatch cut-off.
• By 30 September: re-paper the packaging spec for automated lines, covering carton tolerance, long-side barcode, no protrusions and a stated pallet footprint.
• By 10 October: finish ASN, order and inventory integration and push a 200-order test file before peak.
• Keep 30-day quote validity through the fourth quarter and re-quote only against a written fee-schedule change.
• Where you sell DDP, show the handling fee as a named quote line so the automated saving stays with you.

The fleet JD Logistics showed on 9 September is the largest automation commitment announced in Chinese e-commerce logistics this year: nine product lines, 11 robots and five new models, from the warehouse-picking Canglang to the minus-20C Zhiwolf and the L4 delivery van Dulang. The number that reaches your quotation is the five-year procurement plan of 3 million robots, 1 million unmanned vehicles and 100,000 drones, or 4.1 million units across the three categories. If your goods pass through a JD-operated node, the handling fee and the dispatch cut-off inside your quote are the two lines that plan changes. Everything else in the announcement is a specification.

Cost the difference before you quote it. Assume you ship 3,000 cartons a month into a JD-operated node and the handling fee is RMB 3.60 a carton on a manual line against RMB 2.90 on an automated one. The gap is RMB 0.70 a carton, RMB 2,100 a month and RMB 25,200 a year, and it belongs to whoever pays the handling fee. Under DDP into the Chinese channel that is you, and the saving is yours to keep. Under FOB or FCA it is your buyer's, and you should refuse any re-pricing that passes a manual-line rate to you while your buyer collects the automated one. Now check the timing before you build this into a quote. Four point one million units over five years averages about 68,000 units a month across the whole network, so the node serving your SKUs may not be automated for years.

By 15 September, ask your China logistics provider for a written list of which nodes serving your SKUs run automated lines today and which do not, node by node, with the installation date for any line not yet live. By 20 September, get the fee schedule split into manual and automated handling, quoted per carton and per order, with the dispatch cut-off for each. By 30 September, re-paper the packaging spec to what an automated line needs: carton dimension tolerance, barcode on the long side, no protruding staples or straps, and a stated pallet footprint. By 10 October, complete system integration for ASN, order and inventory data, and push a 200-order test file through before peak locks your volumes. Hold 30-day quote validity through the fourth quarter and re-quote only against a written change to the fee schedule.

Three alternatives and three traps. If your SKUs are non-standard, stay on a manual node with a later cut-off and a higher per-carton fee, which is usually cheaper than re-tooling packaging for a few hundred orders. If you sell DDP, show the handling fee as a named line on the quote rather than burying it in the total, so the automated saving lands with you when it arrives. The traps are the same shape. Treating the 9 September launch as capacity you can buy today is the first: it was a product launch with a five-year procurement plan attached, not a service-level commitment. Packaging to suit a gripper rather than a parcel network is the second, because thinner board that survives a robot cell can collapse in the last mile. The third is an automated rate quoted on a manual line, which appears in the accounts as a variance nobody can explain.

For Cross-Border E-commerce

• By 18 September: obtain order-level fee breakdowns for every node, separating pick-and-pack from consumables, storage and returns handling.
• By 25 September: split the catalogue into automation-ready and manual-only SKUs and price each list on its own rate.
• By 30 September: re-price the top 20 SKUs on corrected handling costs and rebuild landed cost per unit.
• By 10 October: run 500 live orders through the automated node and compare the actual fee per order against the quote.
• Hold 21 days of cover at the node nearest your fastest movers and 30 days on cold-chain lines.
• Do not put a same-day dispatch promise on the drone, L4 van or cold-chain robot lines until your own node is certified.

For a seller running parcels through a JD-operated node, the 9 September launch is a cost-per-order story before it is a technology story. Nine product lines, 11 robots and five new models went on show, and the minus-20C Zhiwolf and the 24/7 unmanned pharmacy Mulang speak directly to cold-chain and regulated categories. The plan behind them is 3 million robots, 1 million unmanned vehicles and 100,000 drones in five years. Where your order passes an automated pick line, the pick-and-pack fee and the cut-off are what change. Where it does not, nothing changes at all, and that is the part to establish first.

Assume 80,000 orders a month and a pick-and-pack fee of RMB 1.15 an order on a manual line against RMB 0.80 on an automated one. The gap is RMB 0.35 an order, RMB 28,000 a month and RMB 336,000 a year. Set that against an average order value of RMB 120 and a 30% gross margin of RMB 36: the RMB 0.35 is 0.29% of revenue and just under 1% of gross margin, real but not decisive. The cut-off is the bigger number. Assume 22% of orders arrive between a 16:00 manual cut-off and a 21:00 automated one, which is 17,600 orders a month moving from next-day to same-day dispatch, and a same-day dispatch is worth roughly a full day at the far end. Five hours of cut-off buys a day of transit, and that is where the money and the reviews sit.

By 18 September, get the order-level fee breakdown for every node you use and separate pick-and-pack from consumables, storage and returns handling. By 25 September, split your catalogue into two lists: SKUs that clear an automated line on dimensions and barcode placement, and SKUs that do not, then price each list on its own rate. By 30 September, re-price the top 20 SKUs on the corrected cost and rebuild the landed cost per unit. By 10 October, run 500 live orders through the automated node and measure the true fee per order against the quote before peak locks your volumes in. Hold 21 days of cover at the node closest to your fastest movers, and 30 days on the cold-chain lines, where minus-20C capacity is thinner than ambient.

Two alternatives. Moving only fast movers to the automated node and leaving slow movers on a manual line close to demand keeps the cheap per-order rate where the volume is and avoids re-tooling packaging for a long tail of low-volume SKUs. For cold-chain and pharmacy-adjacent categories, the Zhiwolf and Mulang lines show where capacity is going, but the certification date of your node, not the launch date of the robot, is what you can promise against. The traps are a manual-node rate billed on non-standard cartons after you have already quoted the automated rate, a same-day promise built on a five-year drone and L4 plan, and a safety-stock plan drawn at 21 days when a five-hour cut-off shift is a quarter of a day of cover rather than a whole day.

For Manufacturing Plants

• By 16 September: audit your packing line against the automated spec for carton tolerance, long-side barcode, flap security, no staples and pallet overhang.
• By 22 September: run a 200-carton test load and measure the scan failure rate at the receiving node, fixing the most frequent defect first.
• By 30 September: write the automated spec into your packing work instruction and your supplier packaging standard.
• By 10 October: agree the peak appointment and unloading window and confirm the node's fourth-quarter maintenance shutdown calendar.
• Move low-volume non-standard lines to a manual node rather than re-tooling, and price the higher per-carton fee into the product cost.
• If you are buying your own picking cell, lock the slot and the price now: 4.1 million units of demand sit in front of you in the queue.

Your exposure is the carton and the pallet, not the robot. JD Logistics showed nine product lines and 11 robots on 9 September and stated a five-year plan for 3 million robots, 1 million unmanned vehicles and 100,000 drones. If you ship components, spares or finished goods into a JD-operated distribution centre, an automated line accepts a narrower carton than a person does. A carton that fails the scan does not stop the line; it drops to manual handling at the manual rate, and your delivery date moves with it. That is the whole risk in one sentence.

Assume you send 12 truckloads a month, 26 pallets a truck and 40 cartons a pallet, which is 1,040 cartons a truck and 12,480 cartons a month. Assume 4% of cartons fail the automated scan, which is 499 cartons. At RMB 4.20 a carton for manual handling against RMB 0.90 automated, the fallback costs RMB 1,647 a month and RMB 19,764 a year, spent on a defect rate nothing on your own line would notice. A rejected load is worse: one truck turned away at the gate costs the full round trip plus the appointment slot, and a second one in the same week consumes the annual saving.

By 16 September, walk your packing line against the automated spec and record carton dimension tolerance, barcode position on the long side, flap security, the absence of staples and straps, and the pallet overhang limit. By 22 September, run a 200-carton test load and measure the scan failure rate at the receiving node, then fix the highest-frequency defect first rather than all of them at once. By 30 September, put the automated spec into your packing work instruction and into the supplier packaging standard, because a supplier that downgrades board grade to save cost can put you back on the manual line. By 10 October, agree the appointment and unloading window for peak and confirm the fourth-quarter maintenance shutdown calendar at the node, since a 24/7 automated line has a narrower window for anything unplanned.

If your product genuinely cannot meet the carton spec, the alternative is a manual node, and the cost is a higher per-carton fee plus slower reconciliation at month end. That is workable and usually cheaper than re-tooling for a low-volume line. There is a second reason to read the five-year plan carefully: 4.1 million units of robotics demand will lengthen the order book for comparable equipment, so if you are buying a picking cell of your own, lock the slot and the price now rather than at the point you need it. The traps are a carton that passes your own check and fails the receiving scan, a pallet that meets the footprint limit and overhangs at one corner, and a just-in-time delivery plan built on a node that goes fully automated next year rather than this one.

For Brand Owners

• By 20 September: map every customer promise to its delivering node and channel, flagging the ones that depend on a 16:00 manual cut-off.
• By 30 September: list which SKUs sit in nodes that are automated today, with node names and installation dates for lines not yet live.
• By 10 October: re-tier the promise by node capability rather than by city so two order paths do not produce two promises.
• Brief customer service with the exact wording of any changed promise, including the reason, by 15 October.
• Keep temperature-controlled promises on certified cold capacity, not on the minus-20C robot launch.
• Before writing a date, ask the node for installed units today, the installation date and the fee line they activate.

The robot fleet changes one thing for a brand: the cut-off behind the promise. JD Logistics set out nine product lines, 11 robots and five new models on 9 September, including the minus-20C Zhiwolf, the 24/7 unmanned pharmacy Mulang and the L4 delivery van Dulang, under a five-year plan for 3 million robots, 1 million unmanned vehicles and 100,000 drones. A later cut-off converts orders from next-day to same-day dispatch at the node. A cold-chain robot converts temperature-sensitive categories from exceptions to standard. Neither exists on the launch date, and both are worth pricing correctly when they do.

Assume 60,000 orders a month and a cut-off that moves from 16:00 to 21:00. Assume 22% of orders land in those five hours, which is 13,200 orders a month that ship a day earlier. If a missed delivery on your promise carries an assumed RMB 8 credit and earlier dispatch removes a third of those incidents, that is about 4,400 credits, RMB 35,200 a month and RMB 422,400 a year, all of it bought with five hours of cut-off rather than a faster courier. Now the margin line. Assume the pick-and-pack fee falls from RMB 1.15 to RMB 0.80 an order, which is RMB 0.35 against a RMB 200 order at a 30% margin of RMB 60. That is 0.58% of gross margin on one order, and it compounds across 60,000 of them.

By 20 September, map every customer-facing promise to the node and the channel that actually delivers it, and mark which promises depend on a manual line and a 16:00 cut-off. By 30 September, list which of your SKUs sit in nodes that are automated today, with the node name and the installation date for any line that is not live. By 10 October, re-tier the promise by node capability rather than by city, so a customer in a city served by two nodes does not receive two different promises from two order paths. Brief customer service with the exact wording by 15 October, including the reason for any shortened promise, and put the routing rule in writing so the choice is not made at the dock under pressure.

Two alternatives are worth keeping on the table. Keep a slower promise on high-value orders that already carry a delivery-window commitment, and use the automated node only for orders where a day of speed changes the outcome. Hold temperature-controlled promises on certified cold capacity rather than on a robot announcement, because the minus-20C line has to be certified at your node before it carries your brand name. The traps are a chain-wide promise written from one automated pilot, a cold-chain promise written before the node is certified, and a packaging change made to suit a gripper that then fails in the last mile, where the customer feels it. Ask three questions before you write a date: how many units are installed at this node today, on what date, and which fee line they activate.

For Procurement Teams

• By 25 September: model the automation category order book and put a written lead-time and price assumption on every line you buy.
• By 30 September: issue an RFQ on a 24-month frame with the first 12 months price-fixed and lead-time liquidated damages.
• By 15 October: amend the logistics agreement with an automation step-down at a named node and date, with verification stated.
• By 31 October: qualify a second source for spares and consumables before the 4.1 million unit plan concentrates the market.
• By 15 November: add a change-of-capability clause that reopens the fee schedule if the automated line slips.
• Do not treat 3 million robots as committed capacity in a negotiation: it is a plan, not a purchase order.

Read the procurement plan as a demand signal and it becomes a lever in two different contracts. JD Logistics will buy 3 million robots, 1 million unmanned vehicles and 100,000 drones in five years, which is 4.1 million units across the nine product lines and 11 robots shown on 9 September. If you buy warehouse automation, that order book sits behind the lead time and the price you are quoted. If you buy third-party logistics services, the same plan is the automation your provider says is coming, and it belongs in the fee schedule as a dated step-down rather than as a slide in a quarterly review.

Assume you are buying 300 picking robots over three years at RMB 180,000 a unit and a 12-week lead time. If the category order book extends lead time to 20 weeks and lifts price 8%, the unit becomes RMB 194,400 and the 100 units you take in year two cost RMB 1.44 million more. A 24-month frame with the first 150 units price-fixed avoids RMB 2.16 million of that exposure. Now the services side. Assume an annual third-party logistics spend of RMB 40 million with handling at 35%, or RMB 14 million. An automation step-down clause that cuts the handling rate 12% at a named node on a named date is RMB 1.68 million a year, and it costs the provider nothing it has not already announced publicly.

By 25 September, model the automation category order book and put a lead-time and price assumption on every line you buy, with the assumption written down where the negotiator can see it. By 30 September, issue an RFQ on a 24-month frame with the first 12 months price-fixed and a lead-time commitment carrying a liquidated-damages line, because a 12-week promise that delivers in 20 weeks is a schedule risk you carry alone. By 15 October, amend the logistics agreement with an automation step-down: a named handling-rate reduction at a named node on a named date, with the fee schedule attached and the verification method stated. By 31 October, name a second source for spares and consumables, since a 4.1 million unit plan concentrates the market. By 15 November, add a change-of-capability clause that reopens the fee schedule if the automated line is not installed by the agreed date.

Two alternatives. Instead of a price freeze, index the price to a published component index with a collar, which keeps the frame alive if the market moves the other way. Instead of a step-down tied to a calendar date, tie it to a verified capability, so the fee changes when the node confirms the automated line in writing. The traps are treating 3 million robots as committed capacity in a negotiation when it is a plan and not a purchase order; accepting an automation promise with a launch date and no node date; signing a committed volume with a shortfall penalty that costs more than the discount you won; and leaving spares and service outside the frame, where a single source can price the uptime of the whole contract.

↑ Back to the key points