Typhoon Saudel shutters Shanghai and Ningbo; ~90 vessels queued as Far East congestion persists

The Analysis

I have stood on the apron at Yangshan before daylight and watched a queue with nowhere to go. Ninety is the number this week, as many as ninety ships waiting more than a week for a berth in Shanghai, and from the shore it reads like a shipping problem. It is not. It is a terminal problem, then a yard problem, then a truck problem, and the order matters more than the figure.

Typhoon Saudel shut Shanghai and Ningbo for several days last week. The track reached far enough north to disrupt Busan, and by the end of the week the forecasts were pointing at Shenzhen. Those are the facts on the table. What costs you money is not the closure itself. It is the recovery, because a port that loses several days does not hand them back, and the fleet already at sea keeps arriving on schedule whether or not anyone is there to take the lines.

A closure is not a switch, and a terminal is not a car park. When Shanghai and Ningbo shut, what stops is a sequence. The pilot boarding ground closes first, then the berth windows collapse, then the yard stops receiving because nothing is being lifted off, then the gate appointments are cancelled and the trucks turn around and drive back up the highway. Run that sequence in reverse and it does not move at the same speed. The pilot can be back within an hour of the all-clear. The yard takes days, because a yard is a physical stack in a fixed number of blocks with a fixed number of machines, and past a certain density the crane cannot lift one box without moving three others first. Anyone who has stood in a congested yard knows the symptom: the ship is at berth, the plan says eight hundred moves, and the boxes still do not come out at the promised rate.

Here is the arithmetic the headlines skip. Take a port that loses five working days. Assume, generously, that for the following weeks the terminal can work ten per cent more moves per day than normal, with extra gangs, overtime, borrowed equipment and a shipping line willing to accept late windows. Five lost days divided by ten per cent of daily capacity is fifty days. Fifty days just to get back to where you started, before counting one box of new demand. That is not a forecast, it is division, and it explains something that otherwise looks like stubbornness: a storm that closed a port for several days is still visible in the berth queue two months later.

Since mid-July the coast has not had a clean window. That phrase is doing real work. A queue is a stock, not a flow, and a stock only falls when inflow stays below capacity for long enough. With storms arriving every few weeks, the recovery period of one is interrupted by the arrival of the next, and each interruption leaves a residual that the following week has to absorb on top of its own work. By the time a typhoon the size of Saudel arrives, the terminal is not clearing one five-day hole. It is clearing five or six of them, stacked, while the carrier decides which of its services eats the delay. That decision is not made at the port. It is made in an office, on a service rotation, and your booking is a line item in it.

Busan is where the corridor becomes visible. The storm track went far enough north to disrupt it, and it is also the place carriers send volume when Shanghai or Ningbo cannot take it. Both things happened in the same week, which is the part worth pausing on. A hub that absorbs diverted volume fills its own yard faster, and a hub that fills its yard starts delaying the very transhipment connections it was chosen to protect. Shenzhen, further south, was being flagged as the next node in line. Put the map on the table and what you see is not a port problem at all. It is a weather system walking along a coastline where the deep-water berths that can take the largest ships are concentrated in a handful of places, and every one of those places is inside the same envelope for a few days. An alternate gateway is only an alternate if it sits outside the weather. Most of the time, on this coast, it does not.

Transhipment is where this turns from an inconvenience into a lead-time problem. A direct box needs one berth window at each end. A transhipment box needs two at the intermediate hub, and it needs them to line up. When a hub is working through a backlog, the connection is the first thing to break, because a feeder waiting for a mother vessel is the cheapest thing on the quay to postpone. Your Shanghai direct booking becomes a Shanghai to Busan to Europe routing, or a Shanghai to Kaohsiung routing, and nobody asks you. Each extra leg is days you did not put into the promise you made to your customer. Worse, the hub is chosen by the carrier and can be changed after you have already committed a delivery date, so the number on your order confirmation has quietly become an output of somebody else's network decision.

The point I have not seen in the coverage is about the first ten days after a port reopens, and it is not about the quay. It is about the gate. A multi-day closure leaves two piles: the boxes that were in the yard waiting to load, and the boxes that were on the road or still at the factory waiting to get in. Both want the same gate at the same time, and the gate has a fixed number of appointment slots per hour, a fixed number of machines and a fixed amount of ground to put a box down on before it can be stacked.

In my experience the export side surges first, because exporters who missed a sailing rebook onto the next one and the cargo is already sitting there. Import boxes clear comparatively quickly, because consignees want them and the paperwork is already done. So the binding constraint in week one is not the ship and not the crane. It is a trucking appointment, and the haulier who tells you he is fully booked for four days is telling you the truth.

That has a consequence people miss. Your vessel being at berth is not the same as your box moving. A shipment can be on a vessel that is working, at a terminal that is open, and still miss the sailing because the gate slot was not there on the day the empty had to go back or the loaded box had to come in. If you take one thing from this piece, take that: in the ten days after a reopening the scarce item is the gate slot, and it is booked like a hotel room, first come and non-transferable.

Reefer cargo is a separate problem and a harder one. A terminal has a finite number of plugs, and they are laid out in blocks with power and monitoring, not scattered wherever there is space. When a port shuts for several days, the reefers already in the yard keep drawing power and keep needing monitoring, and the reefers arriving from the road have nowhere to plug in. That is not a delay you can absorb with a buffer in the purchase order. Cargo with a temperature regime has a clock that does not stop when the terminal does, and the decisions taken around it in the first seventy-two hours are usually about whether to keep it, move it, or sell it where it stands. Dangerous goods are the same story with a longer shadow: segregation rules mean a closed block cannot simply be replaced by another block, so a DG box caught in a closure can sit outside the terminal for a week while everyone waits for a compliant slot to open.

There is an empties layer that almost never makes the news and always shows up in the invoice. The export surge after a reopening needs empty containers, and the depots that hold them are on the same road network as the terminals, staffed by the same drivers. When sixty boxes want to be gated in on a Tuesday morning and there are four hundred chassis in the district, the shortage is not the vessel and not the berth. It is a metal box on wheels with a driver in it. I have watched exporters pay for a sailing they never made because the empty arrived at the factory on Thursday instead of Monday, and nobody in the chain felt responsible, because everybody had done their own job correctly.

The exposure is not the same for every kind of cargo, and the ranges are wide enough to plan against. Seasonal retail is the most brittle: a container of autumn goods that lands two weeks late has not been delayed, it has been written off, because the shelf date was the product. Industrial inputs are more forgiving on time and less forgiving on sequence, since a missing component stops a line even if the other nineteen pallets arrived early. Project and out-of-gauge cargo is a category of its own, because it needs a berth window with no other ship alongside, special handling and often a road permit, and a queue of ninety ships means the special window is the first thing the terminal gives away. Any of these three, sitting inside a storm window, should be treated as a different problem from a container of consumer goods that can wait a fortnight.

One structural detail deserves naming, because it explains why your box seems to lose every tie. When a terminal comes back from a closure, it does not work the queue in the order the ships arrived. It works to a plan built around berth length, crane productivity, the exchange size of each call and the commitments in the alliance's proforma. A service with four ships in the rotation and a fixed weekly window has more leverage than a one-off call, and a ship that can be turned in eighteen hours gets worked before a ship that needs thirty. That is not injustice, it is throughput math. But it means the small shipper's container is riding on somebody else's priority, and the only way to know which priority you are on is to ask your forwarder a direct question about the service string, not about your booking.

One more thing about the ninety. It is a count of ships, which is the easiest number to report and the least useful one to plan with. Ninety ships is not ninety times a fixed delay, because the ships are not the same size and do not work the same exchange. The number that matters to you is the terminal's delay against proforma, which is a different measurement and usually a smaller-looking one: if the average call is running three days behind its window, the queue can be long and your box can still make its connection. Ask for the delay in days, per service string, rather than the queue in ships. A forwarder who can answer that question has been inside the terminal's plan for the week; one who answers with the headline number has been reading the same article you have.

Let me put numbers on the smaller case, with assumptions stated so you can throw them out if they do not fit. An exporter ships forty containers a month out of Shanghai, mostly to Europe. Assume a quarter of that volume, ten boxes, falls inside a storm window. Assume each slips seven days, which is the low end of the contingency I would use. That is seventy box-days of extra inventory in transit, plus storage at destination for whatever lands together instead of in sequence. Assume three of the ten are transhipped through a hub and pick up five more days each. Eighty-five box-days does not sound dramatic until you convert it into the only number that matters, which is whether the customer's commitment date moves. For ten boxes it probably does not, if you told them early. For the two boxes feeding a production line, it does, and that is where the cost lives, in the line stop rather than in the freight.

I am deliberately not putting a rate on this. Rates on this corridor have been repricing on every congestion scare this year, and a number quoted today would be stale by the time you read it. The date arithmetic is stable. Use that instead; it survives the next storm.

Now scale it. A mid-sized importer brings in two hundred containers a month, half of it on a single Asia-Europe string. Assume thirty per cent sits inside the window, sixty boxes, and assume the same seven days. That is four hundred and twenty box-days, and at that size the problem changes character. Sixty boxes do not arrive late one at a time. They arrive in a lump, three weeks of demand landing in one week, and a distribution centre sized for a smooth flow has to absorb the spike through the same number of dock doors. The labour is not there, the yard is not there, and the appointments you needed at the DC were taken a fortnight ago by somebody else. So the delay you measured in days at the port gets measured in weeks at the DC, and the second measurement is the one your customer experiences.

The transmission has a shape, and it helps to know which wave you are standing in. The first wave is the closure itself, five days or so, and it is the cheapest, because everybody knows about it and everybody plans around it. The second wave is the berth queue, roughly from day five to day twenty, when vessels wait, sailings get rolled and the gate becomes the constraint rather than the quay. The third wave is the transhipment tail, from about week three to week six, when boxes surface at a hub you did not choose and the connection has to be rebuilt. A fourth wave is quieter and belongs to the network rather than to your shipment: capacity tied up in queues is capacity unavailable for the next fixture, and the corridor reprices for everybody, including shippers whose own port never closed.

What to do about it, in the order I would do it. When a closure notice lands, the first call is to the forwarder, and the question is not whether the shipment is affected but which sailing, with the actual vessel cutoff in writing. Ask for the omission status of the service over the past two weeks. Ask what the contingency routing is and who decides it. Do that inside seventy-two hours, because the answer degrades; after a week the forwarder is repeating what the carrier told him, and the carrier is repeating the rotation plan. Then book the landside, haulage, gate appointment and empty return, before you book anything else, because in the ten days after a reopening the gate slot is the scarce item and it does not wait for your purchase order to be amended.

Then amend the promise. Every downstream commitment that depends on that box gets seven to ten days added, in writing, with the number in the order rather than in the apology. And tell the customer on the day the notice lands, with the revised arrival and the routing, because a customer who hears it from you on day one is a customer, and a customer who finds out from a late container is a claim.

The alternatives each have a price, and I would rather name the prices in kind than pretend one of them is free. Moving to an alternate gateway sounds clean until you notice that the escape route uses the node that is congested: to get your box to Busan you first need a feeder berth in Shanghai, and Shanghai is precisely the place that cannot take a berth. Trucking to another Chinese port works for a slice of cargo and costs a domestic move, a second set of handling and a re-plan on the export side; it also puts your box on a road competing with everyone else who had the same idea in the same week.

Rail is a genuine option for some Europe-bound volume and no option at all for a container that has to be in a shop in three weeks. Air takes a fraction of the volume at a multiple of the cost, which is the right answer for two pallets and the wrong answer for forty boxes. None of these is a strategy. Each is a tool with a limit, and the limit is usually landside rather than on the main leg.

Two counterfactuals, because the conclusion is conditional and I would not want it read as a prediction. If the next storm does not arrive within three weeks, the queue clears, the contingency you paid for looks wasted, and somebody in finance will ask why you spent it. The answer is that you bought an option, not a forecast, and the option cost less than the launch you did not miss. If instead a storm lands inside ten days, the queue never clears, because the residual from the first is still there when the second arrives, and seven to ten days of buffer is no longer enough; you are in the fifteen-to-twenty range and the conversation with your customer becomes a different conversation. The variable that flips the conclusion is not the size of the storm. It is the interval between storms, and everything else, including how bad the queue looks on any given morning, is downstream of that interval.

I have watched this coastline long enough to be suspicious of the word unprecedented. The storms are not new and the queue is not new. What changed is the concentration: the deep-water capacity that can take the largest ships sits in a very small number of places on this coast, and when a weather system walks along it, the whole network feels it in the same week. That is the thing to plan against. Not the wind.

↑ Back to the key points

— By Tomás Ferreira