China rail freight hits 3.05bn tonnes in 9M; China-Europe trains up 15% to 16,000

The Analysis

Put the land bridge map on the table this time, the one with Alashankou and Khorgos on the western edge, Erenhot and Manzhouli up north, and a single gauge change between all of them and Europe. On 20 October the national railway group published its nine-month figures: 30.5bn tonnes of freight carried, 188,000 wagons loaded a day on average. Two numbers, and the second one is the one that decides whether your container moves in November.

I have stood on the Kazakh side of Khorgos watching a train come apart and go back together under a gantry, and I have watched the same thing happen at Brest on the other end of the corridor. A land bridge is not a line. It is a chain of nodes where something has to be lifted, transhipped, re-documented, or waited for, and the throughput of the whole corridor is set by the slowest of those nodes rather than by the fastest locomotive on it.

So the nine-month release gets read here as a dispatch sheet and not as a press release. 30.5bn tonnes. 188,000 wagons a day. 15.6bn tonnes of coal, 10.4bn of it power station coal. 88.96m tonnes of grain, up 15%, with a September of 10.25m tonnes up 26%. 348m tonnes of rail-road intermodal, up 54%. 16,000 China-Europe trains, up 15%, against 11,000 Central Asia trains, up 2%. More than 970,000 TEU on the new western land-sea corridor. And 1,629 logistics package contracts covering 1.32bn tonnes. Every one of those numbers points at a node, and the nodes are where the money is.

Start with the pair that nobody reads together. 30.5bn tonnes over nine months and 188,000 wagons loaded a day. Nine months to the end of September is 273 days, so 188,000 wagons a day is something like 51.3m wagon loads over the period, and 30.5bn tonnes across 51.3m wagons works out at roughly 59 tonnes per wagon. That single derived figure tells you more about the railway than any of the published percentages, because it says what the system is short of. A corridor that is moving 59 tonnes in a wagon is not short of track. It is short of wagons, and wagons are allocated, not bought.

Which is why the composition matters more than the total. Coal is 15.6bn tonnes of the 30.5bn, a little over half of everything the railway moved, and 10.4bn tonnes of that is power station coal. Grain is 88.96m tonnes, under 3% of the total. If you are moving manufactured goods in containers on this network, you are bidding for wagons inside a system that spends most of its capacity on energy and a sliver of it on your boxes. That is not a complaint about policy. It is the operating reality of a corridon that allocates by category before it allocates by customer, and it is the reason a rail booking that looked comfortable in August can turn into a wait in November.

Read the coal split once more, because it decides how hard the fourth quarter is going to be. Of the 15.6bn tonnes of coal, 10.4bn tonnes is power station coal, two thirds of the coal and roughly a third of everything the railway carried. Power station coal is not a discretionary flow that responds to price. It is a supply obligation with a winter shape to it, and the winter shape puts its peak in exactly the two months you are trying to book. Add the grain acceleration on top and the picture is plain: two categories that together account for well over half the tonnage are both peaking into the same window. September transit and September availability are not the numbers to plan a fourth-quarter rail booking against.

Now the grain line, because it is the one that tells you about the fourth quarter. 88.96m tonnes over nine months is 9.88m tonnes a month on average, and September alone came in at 10.25m tonnes, above the nine-month average and up 26% year on year against 15% for the period. A monthly figure that runs above its own nine-month average while the year-on-year rate jumps eleven points means the autumn movement had not finished accelerating when the quarter closed. North-to-south grain does not stop at the end of September, and winter power station coal does not ease into January. Whatever slack your container booking found in September is the slack those two flows are about to compete for.

Then the number that should reset your routing plan for 2027: 16,000 China-Europe trains, up 15%, against 11,000 Central Asia trains, up 2%. Run across 273 days, that is roughly 58 China-Europe departures a day and roughly 40 Central Asia departures, and the gap between the two growth rates is thirteen points. One corridor is being pulled hard by demand and the other is barely moving. This line I have walked, and a thirteen-point split of that kind is not a market signal you read and forget. It is a capacity allocation decision that has already been made somewhere else, and by the time it shows up in your forwarder's quotation you are paying for it rather than planning around it.

The temptation is to read the Central Asia figure as cheap spare capacity and move volume onto it. Resist that for a moment and look at where the trains actually go. Central Asia services terminate in Central Asia. If your cargo is bound for Duisburg, Lodz or Milan, a Central Asia path adds a border, a second gauge change and a road leg at the far end, and the two-point growth rate tells you why that path is thin: the onward connections have not been built out at the same pace as the western corridor's.

Cheap capacity on a corridor that does not reach your destination is not cheap. It is a detour with a discount attached. The test is simple and it costs one phone call: ask your operator which terminal the train terminates at, how many borders it crosses to reach your delivery point, and what the road leg costs once it is there. If nobody can answer all three in one go, the 2% is telling you something.

Which brings us to the alternative that is genuinely on the board. The new western land-sea corridor moved more than 970,000 TEU of rail-sea intermodal boxes in nine months, roughly 108,000 TEU a month. That is the southern route: rail down to the Beibu Gulf coast and a ship from there, rather than rail all the way across Kazakhstan, Russia and Belarus. It is a different set of nodes entirely, and it is the credible second option for anyone whose European timing has slack in it. The price of it is a sea leg and the port call that goes with it; the prize is that you are no longer bidding for the same wagons as the coal and the grain.

Size that southern option properly before you commit to it, because the two corridors are not comparable in scale. Roughly 108,000 TEU a month is a substantial flow and a growing one, but it is a port-fed flow, and a port-fed flow has a different ceiling from a train-fed one. Its constraint is berth windows and vessel rotations at the coast rather than wagon allocation in Xinjiang. That is a genuine advantage in a quarter when the northern corridor is short of wagons, and a genuine risk the moment a shipping line reshuffles its schedule. Treat it as a hedge against one specific failure, the wagon shortage, and not as a general replacement for the land bridge.

On the documentation side, the release notes 244 new rail-sea one-document projects on the 95306 platform and 8.9% growth in container rail-sea intermodal volume. Derek Xu reads that line from the documents and customs end rather than from the timetable end, and his question is the one that decides whether the figure is worth anything to you: how many of those 244 projects cover the specific port-to-rail pair you actually ship on, and does the single document follow the box across the customs boundary or only up to it. A one-document project that stops at the port gate is a shorter form, not a shorter transit.

Here is the thing the reports did not spell out, and it is the one that hurts in practice. Rail-road intermodal came in at 348m tonnes, up 54% year on year, and it was printed as an achievement. Read it as a warning instead. Rail-road intermodal is the road leg that connects the rail yard to the door, and a 54% increase in it means the railway is feeding more and more tonnage into a road interface that does not grow at the railway's command. Trains can be added by timetable. Trucks, drivers, yard slots and gate appointments cannot. The fastest-growing line in the whole release is the line that describes the bottleneck, and it is growing faster than anything else on the page.

That has a direct consequence for anyone signing a package contract. The release counts 1,629 logistics package contracts signed between January and September covering 1.32bn tonnes, which works out at a little over 81,000 tonnes per contract and represents over 43% of all freight moved. Those contracts are being written in a year when the road leg is the fastest-growing and least controlled part of the journey. If your package contract is written station to station, as most of them are, then the 54% is happening outside your contract, on your account, at spot road rates, in the same weeks your rail slot was supposed to guarantee you certainty.

On the package contracts, one derived number should change how you negotiate. A little over 81,000 tonnes per contract across 1,629 contracts, covering more than 43% of everything the railway moved, says that package contracting is no longer a novelty for a handful of large shippers. It is how most of the tonnage on this network moves now. If you are outside one, you are buying spot inside a market where nearly half the volume is already committed, and you are doing it in the quarter when coal and grain peak together. If you cannot fill 81,000 tonnes on your own, the practical route is to enter a contract through your forwarder or through a shipper group on your lane, and to do it before the 2027 allocations close rather than after they are printed.

Put a real figure on it. Assume you are a shipper moving 20 TEU a month on the China-Europe corridor into a European inland point, and assume your door-to-door transit is quoted as a single number. Split that number into the rail leg, the gauge change, and the road leg at each end, and ask the forwarder to price the road legs separately. On the released figures, the rail leg is on a corridor growing 15% and the road leg is inside a category growing 54%. Which of those two do you think will be re-priced first when the fourth quarter fills, and which one is your contract silent about?

And read the 15% from the receiving end, because that is where the year is decided for an importer. A corridor delivering 16,000 trains a year into Europe is delivering them into Malaszewicze, Duisburg and a handful of inland terminals whose capacity did not grow 15% alongside it. The European half of a land bridge is a terminal and road network, and when the eastern half accelerates, the congestion does not stay in Kazakhstan. It migrates to the last node before your door. If you are the receiver, the question for your operator is not what the train costs. It is how many days your box is expected to sit at the European terminal, and who pays for the dwell when that number moves.

The gauge change deserves its own paragraph, because it is where a land bridge either works or does not. Alashankou and Khorgos take the container off the standard-gauge train and onto a broad-gauge one; Erenhot and Manzhouli do the same on the northern route; Brest and Malaszewicze do it again on the way into Europe. Every one of those places is a crane, a yard, a customs house and a queue. Do not underestimate that crossing. A corridor can add trains every week and still move no more boxes, because what it added was trains and what it needed was lift capacity and yard space at the two places where the wheels have to change.

A word on how a crossing actually fails, because it never fails the way a timetable suggests. At Alashankou and Khorgos the work is split, and in a heavy week the queue does not form on the track, it forms on the ground, in stacks of boxes waiting for a spreader. A crossing running at its lift limit shows it first as dwell: your container sits two days, then four, then a week, while the trains keep arriving to schedule. Anyone who has walked this line knows the tell. The tracking app still says departed, and the box is still in Kazakhstan.

This is why the 15% on China-Europe trains should be read together with the wagon number rather than on its own. Sixteen thousand trains up 15% is a timetable achievement. Fifty-nine tonnes per wagon and 188,000 wagons a day is a fleet constraint. If the corridor is adding services into a fleet that is already fully allocated, the additional trains compete with coal and grain for the same wagons and the same paths, and the constraint simply moves from the timetable to the yard. The symptom for you is not a rate increase. It is a booking that gets confirmed for a train that then leaves without your box.

So the practical question for the rest of the year is how much of your fourth quarter is exposed to that. Take the three flows the release identifies as accelerating, power station coal into winter, grain running 26% above last September, and China-Europe trains up 15%, and ask which of your own shipments sit on the same wagons, the same paths or the same yards as those three. Anything that does should be booked earlier than usual this quarter, and anything that can be shifted to the southern corridor or to a different crossing should be priced now, while the 2027 allocations are still being discussed rather than after they are published.

Set the substitution up the way you would set up any route change: with a test. Move one shipment, not a programme, through the crossing or the corridor you have never used, with your own documents and your own customs broker, and time it end to end including the yard dwell at the gauge change and the road leg at the far end. A corridor you have only read about is a brochure. A corridor you have shipped on once is an option, and an option is the only thing that gives you leverage when the allocation meeting happens without you in the room.

Then put a trigger on it rather than a date, because the corridor does not move on your calendar. Move volume to the southern rail-sea routing or to an alternative crossing when your China-Europe rail booking is rolled twice in a month, or when quoted transit on your lane stretches by more than a week against the transit you saw in September. Move it back when the roll stops. Written down now, that rule costs you nothing; worked out in December, when the coal and the grain are both peaking and the yard at Khorgos is full, it costs you the season.

And watch the one-document projects rather than admiring them. Two hundred and forty-four new projects is a real number, but it is a number about coverage, not about speed. Check whether your port pair is among them, ask what the single document actually covers, and file one shipment under it before you rely on it. The document is what makes a corridor usable in practice, and the corridor with the best timetable and the wrong paperwork is just an expensive way to park a container in Kazakhstan.

The line still walks. It has walked through every winter I have loaded a train for, through gauge changes, through sanctions, through grain seasons and coal seasons, and it will walk through this one. What has changed is how much of it you can take for granted. Fifty-nine tonnes a wagon, half the network on coal, grain accelerating into the fourth quarter, a European corridor running hot at 15% and a Central Asian one flat at 2%, and a road leg growing faster than everything it connects to. Read the nodes, book against the constraint rather than the headline, and the nine-month release stops being a set of achievements and becomes what it really is: a map of where your cargo will wait.

↑ Back to the key points

— By Ömer Kaya