Shanghai and Ningbo Clear Typhoon Backlogs as South China Terminals Absorb Diverted Ships

The Analysis

Spread the map out and the first thing that matters is not the port that closed. It is the hinge that closed with it. Shanghai and Ningbo sit where the Yangtze's manufacturing belt meets the open sea, and when Typhoon Bavi forced both to shut for a stretch in July 2026, it did not take out two container terminals. It took out the seaward gate of an entire river system, and every schedule built on top of that gate had to be rebuilt while the storm was still moving north.

The queues are thinning now. Carriers are working through the vessel backlog at Shanghai and Ningbo, waiting times are coming down from their peak, and it is reasonable to start writing about recovery. Follow the boxes instead of the headlines, though, and a different picture appears. The pressure that left the Yangtze delta did not evaporate. It is building at Yantian, Shekou and Nansha, three terminals of the Pearl River delta, a thousand kilometres to the south, on a different river, serving a different set of factories, and now handling tonnage that was never proformaed for them.

Forwarders are telling shippers to confirm container gate-in and actual departure rather than trusting the booked ETD. It reads like housekeeping. It is actually the most useful sentence in the whole story, because it points at the one part of a typhoon recovery that almost nobody photographs. I want to take that sentence apart, then put the map back together around it.

Start with what happened, because the summary is short and the consequences are not. Bavi forced closures at Shanghai and Ningbo in July. Waiting times at both ports lengthened. Carriers responded by omitting calls, cutting ports out of a rotation rather than trying to serve every one of them late. A skipped call is never a neutral act. The vessel that does not berth at Ningbo does not load the boxes that were gated in for it, and it does not discharge the boxes that were meant to come off it. Those boxes do not vanish. They are re-fed into the system later, at the back of a queue that was never sized for them, on a schedule that was never written for them.

Put the two deltas side by side and you can see the design assumption under three decades of port building here. The Yangtze mouth and the Pearl River mouth each grew their own constellation of berths, their own feeder networks, their own inland links, their own labour pools and their own customs flows. Each was meant, on the whole, to mind its own hinterland: a factory outside Suzhou ships through Shanghai, a factory in Dongguan ships through Yantian. That assumption held for years because the two systems rarely had to cover for each other. Bavi is a test of whether it still holds, and the early evidence is that it holds less well than the brochures suggest.

Ten years ago this was a different story, and the difference is not the size of the ports, because they were already enormous. The difference is where the slack sat. There were more independent mid-size terminals in the mix, more services that overlapped, more ways a box could miss its intended vessel and still catch a workable one. Consolidation did what consolidation always does. It lowered the cost of the average move and raised the efficiency of the ordinary day, and it quietly retired the redundancy that only earns its keep on the extraordinary one. The extraordinary day is now. Bavi is what a system optimised for the average looks like when it meets the tail.

The mechanism runs through the schedule, not through the wind. A proforma schedule is a promise with a little hidden fat in it, a few buffer hours at each port, enough to absorb the ordinary friction of a berth running late or a pilot held up. A typhoon eats all of that fat at once. Once the buffer is gone, a carrier has two ways to protect the rest of the rotation. It can carry the delay forward and let it smear across every downstream port, or it can cut one port out of the rotation and keep the others on time. Carriers almost always cut, because a skipped call is a clean, contained loss while a spreading delay is a loss that compounds. That decision is rational for the vessel. For the cargo at the omitted port it means a new route, a new departure date, or a longer wait, and none of those three is free.

Which is where the geography turns. Yantian, Shekou and Nansha were not sitting half empty, waiting to lend a hand. They run their own rotations, their own feeders, their own chains of barges and trucks reaching deep into the Pearl River delta's factory floors. When diverted tonnage arrives, the terminal does not absorb it out of goodwill or even out of spare capacity. It absorbs it by tightening everything else. Berth windows compress. Yard blocks fill with boxes that will now sit longer than usual. Truck gates queue. And the feeder and barge slots, the small vessels that stitch the big deep-sea terminals to the smaller river ports, get squeezed first, because they are the least protected part of the system and the easiest to bump.

Who feels the squeeze depends on where you already stand. A large importer with a service contract and a named account team does not get a faster berth, because nobody gets a faster berth during a recovery, but it gets earlier information and it has the volume to act on it: split a booking, re-route a factory's output through a second loading window, carry a week's delay against inventory it already holds. A smaller exporter has neither the information lead nor the volume. Its only real lever is timing, and timing at a congested port is a cruel sort of lever. It rewards the shipper who can gate in early and punishes the one whose factory finishes late, and the factory's production calendar was set long before Bavi had a name.

Then there are the cargoes that cannot be flexible. Reefers need plugs, and plugs are finite, so a reefer that misses its vessel in a recovery is not merely late, it is in a queue for a scarce resource. Out-of-gauge and project cargo need specific handling windows, and a compressed schedule is exactly the schedule that does not want to give windows away. And then the empties, which are the cheapest thing in the system right up to the moment there are none of them. A backlog that swings loaded tonnage from one delta to the other swings the empty pool with it, because empties chase the exports. Move the loading port and you have moved the demand for boxes, and boxes do not teleport.

Here is a small piece of arithmetic to make the shape of this visible, and I will state the assumptions plainly because they are mine and not the port's. Assume a large terminal normally receives three deep-sea calls a day and turns each of them in roughly a day at berth. Assume Bavi closed that terminal for three days. In an ordinary three days, nine calls would have berthed. During the closure none do, so nine calls stack up. Now assume that once the port reopens it can reach four calls a day, one extra turn a day, through longer shifts and tighter yard work, which is already a stretch and not a pace anyone can hold for long.

The net catch-up is exactly one call a day. Clearing the nine-call hole therefore takes nine days of running above normal, before the port has done anything about the new arrivals still streaming in behind it. Read that again: a three-day closure buys roughly a nine-day recovery, a three-to-one amplification, and that is in a single-port model with every knock-on effect stripped out. The amplification is not a failure of management. It is what a queue does when you stop feeding it and then start again.

Now run the same shape on the diverted cargo. Every box that could not wait for the Yangtze delta to recover had to go somewhere, and the somewhere is the Pearl. Those boxes land at Yantian, Shekou and Nansha as extra demand on terminals that are already working through their own rhythms. Stack the two curves and the second delta's busiest fortnight lands, very roughly, two to four weeks after the first delta reopens, far enough away to look unrelated, close enough to be the same event. That time gap is the reason the story feels like it is ending just as the second chapter begins.

Timing matters more than any single number here, so let me lay out the clock as I read it. The sea-side queue at Shanghai and Ningbo clears over days to a couple of weeks, depending on how much recovery capacity the terminals can actually sustain. Vessel schedules take longer to look normal, because a rotation only resets when the whole loop, every port and every feeder, realigns, and that is measured in weeks, not days. The empty pool takes longer still, because empties have to be repositioned by the very network that is already working flat out. The commercial effect lands last: the inventory that runs short, the delivery promise that slips, the retail window that closes, all of it arriving a month or more out, which is exactly why it draws the least attention at the moment the visible queue starts to drain.

If you want to know how fast this is really clearing, watch three things and nothing else. Watch the anchorage count off the river mouth, because it is the first number to move. Watch berth waiting time, because that is the number that actually costs the cargo. And watch gate turn time at the terminal, because that is the number that tells you whether the queue is moving from the sea side to the land side. A falling ship queue with a rising gate queue is not recovery, it is relocation. Two of those three you can get from your carrier or your forwarder without asking for anything special. The third you usually have to ask for, and in a recovery the asking is part of the job.

Here is the point I have not seen made, and it is the one that matters most to anyone actually moving cargo. Everyone watches the sea side of a recovery: anchorage counts, waiting times, the line of ships off the river mouth. But the binding constraint is almost never the quay. A terminal can usually find a way to put more ships alongside, because that is what cranes, berths and overtime are for. What it cannot do is conjure, in the same week, more trucks, more chassis, more yard space, more barge slots or more empties. When the headlines say the backlog is clearing at Shanghai and Ningbo, what is really happening is that a sea-side queue is draining into a land-side queue that no camera is pointed at.

The reason the forwarders' advice, confirm gate-in and actual departure, is the smartest line in the story is that it is the instruction that tells you which side of that interface your box is stuck on. A booked ETD is a sea-side number. A gate-in record is a land-side fact. When the two disagree, the land side is right, and the land side is the thing that will actually make you late.

There is a second layer under that one, and it is about the network's memory. A box that leaves through a different port than the one it was booked through does not simply rejoin the system where it should have been. It rejoins somewhere else, on a different rotation, in a different sequence, and the feeder and empty loops that were calibrated around the original path are now out of phase with reality. Nothing about that box shows up as an error until the downstream leg fails to receive it. This is the quiet cost of a diversion: the network carries a map of where things ought to be, and every diverted box makes that map a little less true for several weeks afterwards. Multiply it by a few thousand boxes and you get a system that is technically recovering while quietly misconnecting.

There is also a lane question that gets flattened in every summary. The two deltas do not carry the same traffic. The Yangtze delta is heavy on Europe-bound and trans-Pacific boxes that move in large, well-organised blocks. The Pearl River delta carries a great deal of trans-Pacific volume too, but it also carries an enormous amount of intra-Asia and feeder traffic that never leaves the region, plus the electronics and consumer-goods flows that have their own timing. So when a shipper says he will just move it south, the honest answer is that it depends which lane your cargo is on. If your service calls both deltas, a diversion is a scheduling inconvenience. If it does not, diversion does not move your cargo to a different berth on the same ship. It moves it to a different ship, a different booking and a different contract, and that is a much bigger decision than it sounds.

What would change my reading? If the same pattern repeats inside this same season, another closure and the same downstream spillover into the Pearl, I will stop calling Bavi a pulse and start calling it a pattern, because one typhoon is weather and two in a season, with the network answering each one the same way, is a fact about how little slack the system now carries. On the other side, if Yantian, Shekou and Nansha absorb the diverted tonnage and settle back into their own rhythm within a normal rotation, then I have been too gloomy and the network has more give than I am crediting it with. The single variable that flips the whole conclusion is time. If the sea-side queue clears in under a week, this is a story about schedules. If it drags past a fortnight, it becomes a story about where the network's redundancy went.

On the ten-year view I do not think Bavi redraws anything. The two deltas will still be the two deltas; the Yangtze will still carry the export machine of the interior down to the sea, and the Pearl will still feed the southern coast. But a pulse is how you take a system's pulse, and this one is informative. A network with real resilience takes the hit at the extremity and barely notices it at the centre. A network that has traded redundancy for efficiency transmits the hit inward, port to port, delta to delta, until somewhere a factory discovers that the problem it thought was a shipping problem is actually a capacity question that has been ten years in the making. I do not yet know which network this is. Bavi is one reading. The next storm will tell us whether it was an outlier.

The bigger question, and the one I would rather end the decade on, is where this map grows next. For twenty years the answer was always more: more deep water, more cranes, more automation, more throughput per hectare. That answer still works on the average day. What the last month has been quietly asking is whether the next round of investment is about more capacity or about more slack, a second loading port that a factory can actually reach, a barge or rail link that gives a box a real alternative, an empty pool that does not have to travel across a thousand kilometres to find a hull. Those things look expensive right up until the week they save the shipment. Then they look cheap.

Which brings me to the alternatives, and their costs, because just move it to South China is a sentence that hides a great deal of work. Changing the loading port does not change one box; it changes a chain. The inland leg has to be re-quoted and re-booked, because a thousand kilometres of drayage does not move itself. The empty pickup moves with the loading port, which is fine if empties are available there and a genuine problem if they are not. The feeder or barge connection into the deep-sea terminal needs a slot, and feeder slots are the first thing to be squeezed. And the documentation chain, the booking, the shipping instruction, the whole sequence of cut-offs that hangs off the port of loading, has to be rebuilt around the new port, a task that always takes longer than anyone budgets for and always lands on the shipper's desk at the worst possible moment.

The traps are predictable. Reading backlog cleared as normal when it only means the queue stopped growing. Reading the booked ETD as a departure time when it is a hope. Reading the second delta as a substitute for the first when on some lanes it is not even the same service. And assuming empties will follow your cargo when they follow demand, not intent.

What to do about it, then, in the specific and unglamorous terms that actually help. Get the facts before the schedules, and get them per box. For every open booking with a loading date at Shanghai, Ningbo or anywhere downstream of them, go to the actual gate-in record and the actual departure of the named vessel, not the booked ETD, and do it this week rather than at month end. Anything with no confirmed loading inside 48 hours of its planned cut-off should be flagged now, while there is still a booking window to move it into.

If your cargo is going out through the Pearl instead, verify the inland leg, the empty pickup and the feeder slot before you congratulate yourself on finding a berth, because a berth without a truck behind it is not a booking, it is a wish. And build the delivery promise you give your buyer around a queue allowance, two weeks rather than two days, for the rest of August, and say so out loud, because a buffer you explain is a plan and a buffer you hide is an apology.

Put the map away slowly. The river mouths will still be there, the two deltas will still carry most of what this country sells to the world, and in a year or two this July will be a footnote in a schedule file nobody opens. That is not nothing. It is the sound of a system doing its job. But the month it takes a three-day storm to finish working its way through the network is worth remembering, not because the storm was large, but because it took so little to set two of the busiest gateways on earth, and the three ports that cover for them, waiting on the same tide.

↑ Back to the key points

— By Tomás Ferreira