When a Single Route Is No Longer Reliable: How Multimodal Shipping Levels Up from "Spare Tire" to "Primary Plan"

The Analysis

The first thing that hits you at Khorgos is not the trains. It is the cranes, and the fact that they never stop, and the other fact that never makes a press release: when one of them stops, everything behind it stops as well. I have stood in that yard watching wagons queue back past the point where I could still count them, and the thought that stayed with me was that a corridor is not a line on a map. It is a chain of yards, and a chain of yards is only ever as strong as its worst yard.

That is why the first quarter numbers deserve more than a nod. China-Europe rail ran 5,460 trains and carried 546,000 TEU between January and March 2026, up 29% and 22% year on year, with March alone at 1,959 trains, up 23%. The network now reaches 26 countries and 235 cities. Read as a demand story, that is a chart going up and to the right. Read as an operations story, it is a lot of new departures landing on transloading yards, gauge-change terminals and border crossings that were already working flat out.

I run this lane, or enough of it to have opinions. Alashankou and Khorgos on the China side, Dostyk once you are across, Malaszewicze where the broad gauge finally gives up, Duisburg at the end of the line. What follows is what the 29% means when you are the one holding the booking, where it bites first, and what I would have in place this month rather than next quarter.

Start with the arithmetic, because arithmetic is the part the coverage skips. 5,460 trains across a ninety-day quarter is roughly sixty departures a day across the entire network. 546,000 TEU spread over those trains comes to exactly one hundred TEU per train. A hundred TEU on a single departure is a full train by any loading standard I have worked with, and that one division tells you where the 29% came from. Growth of that size cannot be absorbed by spare slots on services that already exist. It gets absorbed by adding departures, and every added departure needs a crane slot, a set of wagons, a path across a border and a truck waiting at the far end.

March makes the point sharper. 1,959 trains in a single month is about sixty-three a day, a shade above the quarterly average, and it lands in the month when the post-holiday restart is still sorting itself out at the Chinese platforms. I have watched that restart from the yard side more than once. Bookings come back before equipment does. For a couple of weeks you have demand and no wagons, then wagons and a crane queue, then a crane queue and no drivers at the destination. A monthly figure running above the quarterly average in that particular month is not a demand signal. It is the network being pushed to its ceiling earlier in the year than it used to be.

What is pulling the volume is not price, whatever the slide deck says. Rail is not cheaper than ocean per box and I have never told a client that it was. What changed is the other side of the comparison. Ocean on the Asia-Europe lane, whether via Suez or around the Cape, is running 35 to 50 days, and that band is wide for a reason that has nothing to do with distance: nobody can promise you which end of it you will land on. Rail is 18 to 22 days on the good services, 13 to 25 across the range of routings actually sold. Put those two next to each other and the argument stops being about money. It becomes about variance.

A shipper with a 35-day transit and a 15-day spread is not buying transport, they are buying a guess. A shipper with a 20-day transit and a 4-day spread is buying a plan. What they pay for the second one is not a premium on speed, it is a premium on being able to name a date and have the date hold. That is why auto and auto-parts volumes moved 46% year on year. A plant does not run on average transit times. It runs on sequence, and a line that stops because one container of brackets is on day 41 of a 35-to-50-day window costs more in an afternoon than the whole freight difference on that quarter of inbound.

The other number worth pinning to the wall is the one about Malaszewicze. Roughly 85% to 90% of China-Europe rail enters the EU through that single crossing in Poland. Nine tenths of a corridor through one border point is not a design, it is a position, and it is the position every shipper on this lane is holding whether they know it or not. The reason is mechanical rather than commercial. The broad gauge that carries you across Kazakhstan does not fit the standard gauge inside the EU, so somewhere the boxes have to come off one set of wheels and go onto another, and Malaszewicze is where the equipment to do that at scale was built.

Then look at the two ends. Rail does 13 to 25 days of trunk. Road does the last 1 to 5 days, and those 1 to 5 days are where a surprising share of the disappointment lives. A train that arrives on day 19 into a yard with no chassis, no drivers and a public holiday is a delivery on day 23, and the customer will remember day 23, not day 19. The trunk leg is the part everyone prices. The drayage is the part everyone forgets, and it is the part I would fix before I argued about a single number on the quote.

Which brings me to the two failures that belong in every risk memo on this lane, because both are documented and both point the same way. A crane failure at the Khorgos transloading terminal once jammed 56 trains at once. Fifty-six. That is not a delay, that is a wall. And when the Middle Corridor's volume surged 120%, new bookings on it actually stalled, because a route that was built to be the alternative could not absorb the traffic that arrived the moment it stopped being the alternative.

The pattern is identical in both cases and it is the thing I want a client to hear before they hear a number. Redundancy has a capacity limit. Once enough people treat a backup route as the backup route, the backup becomes the primary bottleneck, and congestion moves off the water and onto the land, off the vessel and onto the crane, off the sailing schedule and onto wagon supply. You cannot fix that by booking earlier. You fix it by never letting one corridor carry more than you can afford to have stopped.

Policy is tilting the same direction, and it is worth being precise about what changed and what did not. The EU multimodal aid rules took effect on 30 March 2026 and lifted the mode-shift subsidy cap from 50% to 90%. Notice what that is. A cap on support, not a rate, and not a promise. It does not make your box cheaper by itself. What it changes is the arithmetic for whoever funds the shift, which in practice is a terminal, a combined-transport operator or a service provider, and the effect reaches your quote through them, on their timetable rather than yours. Meanwhile Austria raised tolls 7.7% and Flanders 20%. Those two numbers are the ones that turn up in a drayage invoice with nobody announcing anything at all.

Put the two halves together and you get the shape of the next two years on this lane. The trunk is being supported and the last mile is getting more expensive. That combination pushes volume toward rail plus drayage and pushes cost toward the short leg, which means the winners are not the shippers with the cheapest line-haul. They are the ones who already know who runs their trucks out of Malaszewicze and Duisburg on the Tuesday after a long weekend.

At 200 FEU a month and up, the conversation changes completely. You are not choosing a mode, you are choosing an exposure. With 85% to 90% of the corridor going through one Polish crossing, a book that is fully on rail is functionally a book fully exposed to that crossing, and a two-week stoppage there is a two-week stoppage on your inbound. The answer is not to move everything back to ocean, because ocean at 35 to 50 days is not a fix for anyone with a line to feed. The answer is to write a cap into your own plan: no single gateway above a set share of monthly volume, no single origin platform above another set share, and a named second routing that has been priced and tested rather than merely discussed in a meeting.

I would also separate the contract from the season. A line-haul rate agreed in a quiet quarter is a rate agreed against a network that will not exist in the fourth quarter, and the difference gets settled at spot when you least want to be at spot. Large shippers get this wrong in one specific way. They negotiate hard on the per-box number and then discover the number was never the constraint. The constraint was wagons in week 41 and trucks in week 42.

Below that, at 20 to 50 FEU a month or a few thousand parcels a week, the mechanics differ and the advice has to differ with them. You will not get a dedicated block train and you should stop asking for one. The lever you actually have is origin. Shenzhen gives you sea out of Yantian and Shekou and air out of Bao'an, but the platforms that matter for Europe are inland: Yiwu, Chengdu, Chongqing, Xi'an, Zhengzhou. Getting cargo from a Shenzhen factory to an inland platform is a multi-day move plus a customs step, and I would plan it as one rather than treat it as free. Which means the number you should be comparing is not rail against ocean. It is inland-platform rail, door to door, against Yantian ocean, door to door.

Most small shippers never run that comparison. They take a platform quote, take an ocean quote from their own port, and conclude that rail is slower than the timetable claims. They are right. The timetable was never door to door, and nobody at the platform was lying to them, they were just answering a narrower question than the one being asked. The second lever at that size is consolidation. At 20 FEU you are buying LCL or a slot on somebody else's block train, and in both cases you inherit their cut-off, not yours. Get the cut-off in writing, get the latest cargo acceptance time in writing, and build your factory pick-up backwards from that hour rather than from the vessel sailing.

Not everything on this lane belongs on rail, and the 46% in auto and parts is a good filter for thinking about what does. Parts ride the train because the cost of a day is lopsided: a day early costs a bit of warehouse space, a day late costs a stopped line. Electronics behave the same way with a shorter fuse and a higher value density, which is why the same cargo sometimes flies instead. Cross-border e-commerce sits in a third bucket. The parcel does not care about a container, it cares about the promise made at checkout, and a 4-day spread is survivable while a 15-day spread is not.

What should stay on the water is the cargo where nobody is watching the calendar. Heavy, low-value, non-seasonal, going into a distribution centre that already holds eight weeks of cover. For that cargo a certainty premium buys insurance nobody needs, and the correct answer is to leave it on the long route and treat the 35-to-50-day window as free storage. Knowing which bucket each SKU sits in is worth more than any single negotiation, because it stops you from pushing the wrong 30% onto rail and then wondering why cost per unit went up.

Timing matters as much as the plan, because a good plan executed in the wrong month is just an expensive plan. The first wave is the one still open to you: bookings four to six weeks out, where there is room to test a routing without betting a quarter on it. Put a slice of volume on a second corridor now, thirty to sixty days before you need the answer, and you will know what the transit actually is instead of what the timetable says it is.

The second wave is repricing, which on this lane runs on a quarterly rhythm and lands 60 to 90 days ahead of the volume it governs. If your renewal falls in the next quarter, the number you agree will be struck against a network whose rail share just grew 29% and whose Polish crossing carries nine tenths of the traffic. That is an argument for capping exposure in the contract language, not only in the rate.

The third wave is the peak structure, and it is the one people underrate. Peak used to mean ocean congestion. It now means a rail peak layered on top of an ocean peak, in the same weeks, competing for the same trucks in Europe and the same drivers. The shippers who come through it are the ones who booked the drayage before they booked the train, which sounds obvious and is done by almost nobody.

Here is the small-shipper arithmetic with the assumptions stated, because they are mine and not the market's. Assume you move 20 FEU a month, assume 30% of that is time-sensitive enough to consider rail, and assume one day of cover for that SKU set is worth USD 8,000 in working capital. Ocean runs 35 to 50 days, so you are holding cover against a 15-day spread. Rail at 18 to 22 days carries a 4-day spread. Move those six boxes and you can cut 11 days of safety stock on that portion, which releases roughly USD 88,000 of capital, and you have not touched the other 14 FEU at all.

The number is small on purpose. I am not selling anybody a corridor with a working-capital story. The point is that at 20 FEU a month the win is not in the rate, it is in the variance, and variance is measurable in days of cover before it is measurable in money. Run it on your own figures before you accept mine.

Now the large shipper, same discipline. Assume 200 FEU a month split 140 ocean and 60 rail, and assume both routings are priced and running. That 60 looks like diversification until you apply the crossing figure: at 85% to 90% through Malaszewicze, 51 to 54 of those 60 boxes stand behind the same gate. Your real exposure is not 30% of the book. It is 27% of the book sitting on one crane and one gauge change. If a stoppage costs you a week, you are missing 54 FEU for a week, and no amount of ocean capacity on the other 140 makes that week shorter.

Capping rail at 50% through any one gateway means a second crossing or a different corridor entirely, and it means paying for the privilege of having tested it before you needed it. Assume that test costs a modest premium on 15 FEU a month for one quarter. Compare it against one week of a stopped line. The comparison is not close, and yet it is the comparison almost nobody runs, because the premium is visible every single month and the stoppage is hypothetical right up until it is not.

Here is the part I have not seen written anywhere, and I think it matters more than the headline. The 5,460 counts trains. It does not tell you how many of those were full-timetable departures, and the full-timetable product is the one that saves over 30% of transit time against a regular service. Those two populations behave nothing alike when a border backs up. A scheduled departure with a committed path gets recovered first. A spot departure added to chase demand gets held, and held again, and then it is a train sitting in a yard while a customer is on the phone asking where their cargo is.

So the question to put to your platform is not how many trains ran last quarter. It is what share of the departures on my specific origin-destination pair are full-timetable, and what happened to that share in March when the network was running sixty-three a day. If the answer is that the added trains were mostly spot, then the 29% is a demand number wearing an operations costume, and your transit plan has to be built on the scheduled subset rather than the total.

Two ways this flips, and both belong on paper. If you do nothing and keep a single-corridor book, the cost will not arrive as a rate increase. It will arrive as one week in week 41 when the crossing backs up, the drayage you never booked is not available at any price, and the only remaining option is air freight on cargo that was never meant to fly. That is the expensive version of doing nothing, and it is the default.

The other direction matters just as much. If ocean tightens its band and holds it for two consecutive quarters, the certainty premium rail is charging loses its justification, and the correct move is to shift the time-sensitive slice back to water and keep rail for the cargo that genuinely cannot wait. I would set that trigger explicitly rather than feel my way to it. Two quarters of a stable ocean band, and the mix gets re-run from scratch. A corridor plan that cannot be unwound is not a plan, it is a habit.

I have been on this lane long enough to be suspicious of any number that only moves one way. The 29% is real, and the direction is real, but what is actually being sold on this corridor in 2026 is not speed and it is not price. It is the ability to name a date and be right. Buy that where it is cheap, keep a second road open, and check the tires before the weather turns rather than after.

↑ Back to the key points

— By Ömer Kaya