Singapore waits hit 3-5 days and Shanghai 3.51 as Asian port congestion persists

The Analysis

Put the map on the table and start where the ships are standing still. In the week ending 15 October, vessels at some PSA terminals in Singapore waited three to five days for a berth, and the ships carrying the heaviest transhipment volumes waited longest of all. Shanghai's seven-day average berth wait sat at about 3.51 days, with delays of three to four days reported at individual terminals. Antwerp averaged about 1.37 days, with barge delays running near 24 hours; Rotterdam averaged 1.39 days, with feeder and barge schedules slipping 12 to 24 hours. Tokyo's Oi terminal stopped loading empty containers because of a container backlog, and that suspension ran to 10 October. On 9 October the United States Port of Mobile closed for a hurricane. Six facts across three oceans, and one question underneath all of them: which of these is a pulse the network simply absorbs, and which is the network quietly changing shape.

Ten years ago this map would have been read another way. Congestion then was understood as a shortage of berth and crane at the port in front of you, and the remedy was understood locally: longer quay, taller stacks, another gantry on order, a productivity clause written into the concession. A wait of five days today is very rarely produced at the place where it gets measured. It is produced three thousand miles away, in decisions about which hub a string rotates through, how much cargo a single sailing is asked to carry, and how many onward services that cargo now belongs to.

Read the six numbers as one picture rather than six local stories, and the datum that matters is not any individual day count. It is the spread. Asian waits of 3.51 days and three to five days set against European waits of 1.37 and 1.39 days is neither weather nor luck. That gap tells you where the slack has gone in this network, and the routing decisions of the next two months get made on exactly that gap.

Take them in the order the map suggests, east to west, and start with Tokyo, because it is the strangest of the six. Oi suspended the loading of empty containers because it had run out of room, and the suspension ran to 10 October. Empties are the part of the box cycle that appears on no route map and yet decides whether tomorrow gets stuffed at all: no empty, no booking, no export. A terminal refusing empties is not telling you that boxes are scarce in the world; it is telling you that boxes arrived there faster than trucks and trains could carry them away, which is a landside failure wearing marine clothing.

Assume you load ten boxes a week at that gateway, forty a month. Assume the suspension costs you three days during which empties cannot be picked against this terminal: forty boxes of monthly scheduling moved sideways, and the only answers available are to pull every pickup forward by those three days or to find another depot, which recovers the wait in trucking hours and spends it again in invoices.

Shanghai's 3.51 days asks a different question, because most of the cargo there is yours rather than somebody else's in transit. Assume you quote your own customers 45 days door to door on this lane; the number is yours to correct, but pick one and hold it. Just over three and a half days of berth wait consumes 7.8% of that promise before a single mile is steamed, and individual terminals are reporting three to four days, so plan against four and it is 8.9%. The habit of quoting at port granularity, five weeks from Shanghai, was formed when a port was one place with one queue.

It is now a set of terminals keeping different clocks, and the distance between the port average and the terminal tail is precisely the distance between a plan and a surprise. Insist on the terminal rather than the port appearing on your booking confirmation, Waigaoqiao or Yangshan, because two sets of gates inside one port do not behave the same way in the same week.

Singapore deserves the longest look, because whatever happens here propagates. Assume you move 40 containers a month through Singapore as transhipment cargo, which is 1.33 containers a day across a 30-day month. Assume the planning assumption buried in whichever spreadsheet produced your safety stock is one day of berth wait, because that is roughly what a well-behaved month used to cost. Three to five days means two to four days of unplanned elapsed time per box. Covering the upper end takes four days of cover at 1.33 containers a day, which is 5.3 containers, about 13% of a month's volume standing still instead of moving. Assume 22 pallets to a box and you are asking for roughly 117 pallets of additional stock somewhere in your system. Nobody announces that. It arrives as a nervous purchasing decision taken one container at a time, and if you do not take it deliberately you will take it anyway, later, at a worse price, in a hurry.

Why do the heaviest transhipment ships wait longest? Because a hub like this one does not serve a single queue. It serves several at once, and a mother vessel belonging to one string waits while boxes belonging to another are lifted off the feeder that brought them in. The container that has travelled furthest is often the one waiting on its last three hundred miles. Delay in a network is not the sum of its links; it is what their interdependencies multiply into, and the busiest junction is where every upstream wobble gets absorbed.

Congestion at one hub is somebody else's opening, and networks are honest about that even when press releases are not. Every week a ship loses three days at one gateway, a planner somewhere else looks at the same map and prices the identical box through a different one: a call fifty miles up the coast, or the port the string already sails past on its way east. Assume you move 40 containers a month through Singapore and that 20% of them are time-flexible, eight boxes. Assume routing those eight through a secondary gateway adds two sailing days to the transit but removes three days of variable berth wait at the hub: expected elapsed time falls by a day, and what matters more, the variance around it collapses, because a secondary hub is secondary precisely because its queue is shorter. The practical lesson here is not to pick a winner among ports. It is to keep one alternative lane alive, priced and booked occasionally at all times, because an alternative you have never used is worth nothing in the week you need it.

There is a second thing the numbers omit, which is what happens after the wait ends. A ship that loses four days does not lose four days, it loses its slot at the next port, which means boxes destined onward arrive on whichever sailing will still take them, which means your container misses the connection it was routed on and takes the following one a week later. Assume you move 40 containers a month through Singapore on connection-dependent routings and that one in five of them is booked onto a tight onward leg: eight containers a month sit exposed. A missed connection is not worth three days in Singapore; it is worth those three days plus the schedule interval of the service that should have carried it onward, and that interval is usually seven.

Cross now to the North Sea, where the numbers look almost dull. Antwerp about 1.37 days, Rotterdam about 1.39: two neighbours separated by 0.02 days and roughly a hundred kilometres of water, behaving as much like a single system as two competing ports have ever looked. The interesting entries in that column are the ones underneath. Roughly 24 hours of barge delay against Antwerp, 12 to 24 hours of feeder and barge slippage at Rotterdam. The quay is fine. The plumbing behind it is not. When the berth is open and the barge is late, the constraint has moved inland: to a lock, to a barging slot, to an empty depot with no space left, to whichever part of the chain nobody photographs. Assume you move 120 containers a month by barge off Antwerp, four a day. Twenty-four hours of slippage is those four containers arriving late every day, which either gets planned forward or gets covered by stock, and the choice belongs to whoever noticed first.

Then Mobile, closed on 9 October for a hurricane, which the network will absorb in a week or two and remember for a season. Weather closures are the purest form of pulse: they arrive with warning, they end on a schedule set by the atmosphere rather than by commerce, and everybody can see them coming, so the network routes around them. Set that beside Tokyo and Singapore and you have the discipline worth keeping here. A closure you can predict is an operational nuisance. A wait of three to five days that you can only measure after the fact is a structural signal, because it is the visible edge of something the network has not yet rerouted, repriced or invested its way past. Conflating the two is how companies end up panicking about hurricanes and shrugging at plateaus.

Set this week against the way the same question was handled ten years ago and the change is not in the ships, it is in what buyers used to take as given. Back then the working assumption was that a published rotation would hold, and transit was quoted as a number rather than as a distribution. The operators who come through a season like this one intact have quietly replaced that assumption with two habits: they plan against a band instead of a date, and they price variability as a line item instead of absorbing it as misfortune. Assume three quarters of your Asian volume sits on routings reporting waits between 3.51 days and five days. Quoting internally on four days with a two-day band either side turns a surprise into a parameter, and parameters can be resourced, argued about in advance and handed to somebody rather than discovered on a Monday morning.

Now the thing every one of these headlines leaves out, and the one that costs the most in practice. Each of the numbers above is an average taken at the scale of a port, and none of them is distributed evenly inside that port. Inside Shanghai's 3.51 days sit terminals running three and terminals running four. Inside Singapore's three to five days sits a rule nobody prints, which is that your wait depends on how much transhipment cargo your particular ship is working and whose string it belongs to.

Two containers booked on the same service, into the same port, in the same week can wait two days apart, because one of them is sitting under the other in a stack that came off a feeder which was late. The average erases all of it. Vivian Zhao, who prices capacity for a living rather than describing it, would put the same point into language about numbers: the distribution matters more than the mean, because nobody ships a mean, they ship a particular box on a particular rotation.

So book at the granularity of a terminal and a service rather than a port. Ask for the berth terminal to be named on the confirmation, ask for the string rotation rather than the transit someone advertised, and keep your own ledger of actual waiting by terminal and service, because after three months that ledger is worth more to your planning than any published weekly table.

There is a version of the same story that shows up at your own gate rather than anybody else's, and importers tend to meet it first. Assume you receive 120 containers a month from Asia through one European gateway, four a day across thirty days. A wait band of three to five days does not arrive as a smooth delay; it arrives as bunching, because the ships that were meant to spread across a week come in over two days instead. Assume your receiving dock turns eight boxes a day at standard cost and everything beyond that is overtime or weekend work.

Twelve arriving in one day means four roll to tomorrow, tomorrow's own four arrive on top of them, and you are three days behind before anybody has taken a decision. Lay Antwerp's 24 hours of barge delay on the same dock and the pile grows in a way no single-day number predicts. What begins as waiting at sea finishes as a rostering problem on land, and it is your roster rather than the terminal's.

There is a piece of this that lands on your own account rather than on the port's, and it is what the wait does downstream of itself. Assume your monthly 40 containers land at one European gateway and that 30 of them feed a plant running two shifts. Assume each box carries a week of one critical input and that your buffer stands at five days. A three-day berth wait with 24 hours of barge delay behind it is four days against a five-day buffer, which leaves the plant working on one day of cover for however long the next sailing takes, and turns any further wobble into a schedule intervention rather than a stock movement. Run the same arithmetic for a barge-fed customer base and the answer changes shape without changing much in size. This is why two companies facing an identical berth wait report entirely different quarters: the same wait sat on very different architectures.

There is one queue in this picture that you actually control, and it is the one most companies leave unmanaged. Assume you book 40 containers a month across four services into the same Asian hub, ten boxes on each, spread evenly at two or three a week per service, which is what an allocation looks like when tidiness is the objective. Concentrate instead: put all ten on one service and keep the other three for genuine overflow. Four unpredictable waits become one predictable one, and you end up with a single relationship, a single escalation path and one legible history instead of four thin ones nobody can read. Assume that concentration buys one working day of notice on the bad weeks: at 40 boxes a month that is 480 box-days a year of elapsed time you plan around rather than discover.

So is any of this a turning point? Set the ten-year lens on it and the honest answer is no, and the more interesting answer is that the question is being asked at the wrong scale. Hubs do not decline because they congest. They decline because the trade they were built to serve found another path, because the ships outgrew the water they sit in, or because a neighbour finished a deep-water project that changed the arithmetic of an entire coast. Three to five days at Singapore is painful and it is none of those things. What it exposes is a property networks have that balance sheets do not: the concentration that makes a hub efficient is the same concentration that makes it fragile. Every line in Asia deciding to relay here is what produced this level of efficiency, and every line deciding it in the same week is what produced this queue. Efficiency and fragility are the same property read on different days.

Where the network grows from here follows directly from that. It grows away from single-point dependence, slowly, and in decisions that will look routine rather than strategic. When a hub's queue becomes steady enough to be expensive, the lines start buying optionality: sister loops using another gateway, slots at the terminal that used to be the fallback, capacity booked at the port nobody argued about last year. Shippers do the same at smaller scale and in exactly the same shape, naming and qualifying a second gateway while the first one is still working, because the cost of qualifying it is a rounding error against four days of unplanned dwell on 40 containers a month.

Ten years ago this map was read by counting cranes. Read it now by counting the alternatives each hub has quietly handed its customers. Put yours back on the wall and update it every quarter, read the waits, and above all read the spread between them, because the spread is where the next decision has already been taken.

Fold the map away afterwards with one thought in hand: not one of the six numbers above was caused by a decision taken at the port where it was measured, and not one of them will be cured there either. Waits are symptoms, and the network is both the disease and the medicine. Ten years ago this was a different story, and the reason is not the ports: what changes, today just as it did then, is the network.

↑ Back to the key points

— By Tomás Ferreira