Hateco, APM Terminals Study Four Vietnam Rail Freight Hubs Under $3.5bn Plan

The Analysis

Every container I have ever chased out of northern Vietnam went soft somewhere on the last forty kilometres. It leaves a plant gate in Bac Ninh or an assembly shed outside Thai Nguyen loaded with goods that will end up in someone's living room in Europe, rolls south-east onto Hanoi's ring road, comes down through Hai Duong, and then loses an afternoon. The destination is the deepwater complex at Lach Huyen off Hai Phong, and that complex can take a ship of 18,000 TEU. Whatever else is wrong with this corridor, the water side is not it.

I have run this lane. The jam is never at the crane. It is in the approach, at the gate, on the day before a mother vessel closes her books, and it is made of trucks, one box to a truck, all of them trying to arrive inside the same window because that is how a cutoff works.

That is the lens I read through when Hateco Group and APM Terminals, the Maersk-side terminal operator, signed a non-binding memorandum in Denmark on 20 October 2026 to study a rail freight terminal at Hai Phong and three further northern locations, with the stated aim of tying industrial parks into seaports by rail. Their joint Vietnam pipeline has been reported at 2.5 to 3.5 billion dollars. Do not shrug off this port. A study is precisely where a route either earns a timetable or quietly dies, and every contract signed on this corridor between now and that answer will be priced against whichever it turns out to be.

Three things in this story have steel in them, and the headline number is not one of them. The berth is real: HHIT at Lach Huyen has been working since April 2025 and handles ships up to 18,000 TEU. The drawing is real: Lien Chieu at Da Nang is planned with eight berths and a design throughput of 5.7 million TEU a year. The partnership is real: memorandum signed in Denmark on 20 October 2026, four northern locations including Hai Phong, sitting inside a joint Vietnam pipeline reported between 2.5 and 3.5 billion dollars.

Now the fact that outranks all three, exactly because it is not a fact about anybody: nothing in that paper can be bought. There is no construction start, no commissioning window, no published transit time, no dwell target, no terminal tariff, and nobody anywhere in the chain who will pay you when a studied box misses a studied connection. Four sites is a list of places worth looking at, not a network. Every field a cargo owner would need in order to price against it is still blank.

Why does this surface now, in Denmark, in the autumn? Because the constraint moved. Not so long ago the binding limit on this corridor was the ship you could bring alongside and the berth that could take her; HHIT opening in April 2025 went a long way toward removing that. Once you can work 18,000 TEU at the quay, you have built something that a road system with a practical radius of a hundred and some kilometres simply cannot keep fed. Bigger berths do not merely serve cargo that already exists, they demand new catchment, and rail is the only way to stretch a port's reach far enough to justify concrete that is already poured.

There is also the money. A pipeline of 2.5 to 3.5 billion dollars is an envelope with a list of options inside it, and envelopes tend to produce proposals. A memorandum signed in Denmark rather than in a site office in Hai Phong was signed in the language of corporate development, which is where every good project begins and where a fair share of them stay. None of that makes it unserious. All of it makes it paper.

The third push sits inside the boxes themselves. Northern Vietnam's export base is electronics assembly and components, which is the best cargo a railway ever asked for: uniform weight, uniform cube, high declared value, factories clustered into a handful of parks, output repeating week after week instead of arriving in waves. When a cargo profile looks like that, the case for a shuttle is arithmetic rather than vision, and people who sign memoranda know it.

Who carries the consequences, and how heavy are they? Start with whoever is closest to the water. Exporters inside the parks north and west of Hanoi care first, because on a corridor shaped like this one the inland leg is not a rounding error against the ocean movement. It is the volatile piece, the piece with a long tail, the piece that decides whether you make cutoff.

Importers feeding those same parks wear the mirror image of that exposure. Their problem is not getting finished boxes out, it is getting empties positioned, components delivered to a line running six days a week, and enough predictability through the Tet slowdown to leave inventory in the supplier's hands rather than their own. A rail node only helps them if it can clear cargo, not merely move it, and that single distinction will decide whether this thing is worth anything.

Truckers are where the story gets uncomfortable. Small fleets and owner-operators carry most of the swing capacity on this corridor, and they finance equipment against a three-to-five-year view of what will still be running. A credible rail prospect does not take volume off the road tomorrow; it takes confidence off the road today. Every month a fleet owner puts off replacing a tractor because the trains might be running by 2029 is a month of capacity quietly withdrawn from a market that has not gained one single train path in exchange.

Then there is land, always the earliest mover and always the most forgotten, and I say that having watched the same film play at Izmit and again at Tangier. Plots near any of the four discussed locations acquire a rail-access story long before they acquire rail access, and the story lands in rent. Nobody negotiates the rail premium against the rail timetable, because there is no timetable yet to negotiate against.

One group can stop reading here. If your volumes move through Da Nang, through central Vietnam or through a third-country gateway, nothing in this memorandum touches you this decade. Lien Chieu's eight berths and its 5.7 million TEU design capacity are a separate file, drawn separately, and the discipline is not to confuse a drawing in one region with an intention in another.

On timing, be honest that there are two clocks running at very different speeds. The physical clock: assume a four-site feasibility study takes six to twelve months, assume approvals and land clearance take another year in a case where nothing goes wrong, assume twenty-four to thirty-six months to commission the first phase of the first site, because you are building a terminal with yard, sidings and quite possibly customs infrastructure rather than laying three kilometres of track. Stack them and the earliest plausible revenue shuttle sits around 2029 or 2030 under friendly assumptions, with a wide and well-populated range to the right of that date. Every number in that sentence is my assumption, drawn from how small yard-and-siding projects actually behave; none came from the announcement, which contains no schedule at all.

The paper clock runs much faster. Land near a studied site reprices within weeks of a signature. Lease language changes inside a quarter. Most importers and exporters on this corridor close their 2027 inland agreements in the last quarter of 2026 or the first quarter of 2027, which means the negotiation you are either in or about to walk into will be conducted entirely in the shadow of something nobody can yet describe in numbers.

That mismatch is the whole event. Your decision lead time is counted in months, your benefit lead time is counted in years, and the person across the table knows both.

Here is the thing no report carried, and it is why I would not sell a gram of flexibility before 2028: a memorandum is not a service. Nothing in it can be booked, quoted, penalised or claimed against. Until somebody publishes a transit time, a dwell target and a tariff, rail on this corridor is not an alternative to your truck. It is a rumour your counterparties may charge you for.

Three bills arrive on the back of that rumour, none of them mentioned at the signing.

One comes dressed as volume commitment. A forwarder or a trucker asks for exclusivity, or for a monthly minimum, in exchange for pricing described as rail-ready against a service with no published rate. You hand over flexibility today against a discount nobody can verify tomorrow, and you keep the downside if your own volumes move.

A second bill shows up in rent. Offers for park plots and distribution sheds near the discussed locations will carry a rail-access premium from the moment the story lands, and your lease term will be five years while the first train remains nobody's problem but God's. You pay from month one and, if things go the way most studies go, you collect nothing during the term.

The third and worst bill is the one nobody puts in writing. It is paid by the fleet. Rubber, in this business, behaves like a tyre you keep meaning to replace: it holds fine right up to the trip you needed it most. Truck capacity on this corridor does not have to disappear for you to feel it, it only has to stop being renewed. That extra cost will be quoted to you as a premium for uncertainty, and it will arrive two or three years before the first wagon.

There is a structural half to this too, and it is the half the civil engineers will sort out while the rest of us argue about rates. Rail does not remove the dray, it splits one dray into two and adds two lifts. The first and last mile are still road, still priced in the same thin market, still nobody else's problem. Worse, the moment rail begins to look real, that thin market gets thinner for the reasons above, so the two short legs you are left with are priced against a road network that has just been made nervous.

Two questions then decide whether any of it is worth your attention, and both should be asked in writing. Can the node clear cargo, meaning does it have accumulated, customs-supervised status so a box is released inland rather than merely staged inland? If not, you have not shortened the chain, you have relocated the queue forty kilometres and added a lift. The gauge question is the one people skip entirely: which gauge do those four sites sit on? Vietnam's northern network is largely metre gauge with pockets of mixed track, so either containers transfer or the wagon fleet is purpose-built, and both options cost money per box and cost you frequency.

Put a tape measure on it, using numbers you can check. Suppose an exporter sixty kilometres from the Lach Huyen berths, shipping forty boxes a month, and suppose one of the discussed sites sits fifteen kilometres from their gate. Assume the present road dray, waiting time included, costs three dollars per kilometre per box. Today that is 180 dollars a box, 7,200 a month, 86,400 a year.

Now assume my own numbers for what a rail product might look like, and remember they are mine and not anybody's published tariff. Assume the long middle leg prices at 65 per cent of the equivalent road distance. Assume the fifteen-kilometre plant-to-node dray runs at the same three dollars a kilometre and so costs 45. Assume a terminal lift of 35 a box at the node. The remaining forty-five kilometres to the port would have cost 135 by road, so at 65 per cent the rail leg comes to 88. The rail door therefore costs 45 plus 88 plus 35, which is 168, against 180 today. Twelve dollars a box, 480 a month, 5,760 a year, 6.7 per cent off the inland bill.

Now price the dwell, which is the part nobody quotes. Assume that box carries goods worth 120,000 dollars and that carrying that inventory costs you 10 per cent a year. One day of hold is 120,000 times 0.10 divided by 365, which is 32 dollars 90 a box per day. Assume the new routing adds dwell, because a start-up shuttle with no published slot release will, and call it one extra day door-to-gate. Save 12, lose 33. Break-even is 12 divided by 32.9, which is 0.36 of a day, call it nine hours. Anything slower than nine hours and cheap rail loses money on this cargo.

Vivian Zhao would take that twelve dollars and ask the follow-up I would miss: whether the terminal lift and the rail leg come to you as separate lines on the quote or folded into one all-in figure, because only a split line lets you see whether the saving exists at all or has quietly been absorbed somewhere else on the page.

Run the same test from the berth side, using the one hard capacity figure in the story. Assume a ship of 18,000 TEU calls at Lach Huyen and, conservatively, that a quarter of her nominal capacity — 4,500 boxes — is worked locally and has to move inland. Assume one truck per box, which is how it goes today. Assume the books close forty-eight hours before sailing. Four and a half thousand truck moves inside two days is roughly 94 an hour through the approach road, sustained. Assume a shuttle takes 50 boxes. Clearing that same pile by rail takes 90 shuttles; at six paths a day that runs fifteen days. No ship waits fifteen days.

Which leaves the sentence I would nail above your desk: rail takes the base flow, rubber takes the peak. Nothing in this memorandum changes anything for the box that must move in the last forty-eight hours. Everything might change, eventually, for the forty boxes that leave the same plant in the same direction every week.

Now turn it around, as any decent operator should. Upside case: all four locations advance, one gets customs capability and a decent gauge solution, and somebody publishes a real timetable with a dwell commitment. Then the northern corridor genuinely becomes two modes instead of one somewhere late in this decade, and location stops being a rent question and becomes a cost-structure question. A plant sitting inside twenty kilometres of a working node gains something structural; the same company sitting sixty kilometres away gains nothing and may lose ground to a competitor who does.

Downside case, which is the one my gut backs: the study finds the base volumes are there and the last-mile economics are not, the four locations become one, or the file simply slides for a couple of seasons with nothing broken. Then nothing happens. No crisis, no embargo, no disruption headline, just continuation, and the approach road outside Lach Huyen behaves exactly as it does now. If you kept your flexibility, fine. If you signed three years of minimum volume against a service that never once quoted you a price, expensive.

The disagreement worth naming is between people reading the 2.5 to 3.5 billion as intent and people, myself among them, reading it as a ceiling on a list. Only one item on that list has anything in the water, and it is the one at Lach Huyen that started working in April 2025. A pipeline is not a plan, and even a plan is not a timetable.

The condition that flips the whole argument is easy to name and impossible to shortcut: customs status at the node, alongside gauge. Give me a node that clears cargo and moves standard boxes without a transfer, and I will happily redraw this corridor around it. Give me a rail ramp with no clearance and a metre-gauge transfer, and I have simply added a handling point to a road trip.

So what does a cargo owner do with a study? You treat it as paper until it becomes a product, and you set the test yourself: three published items, being transit time, dwell target and terminal tariff. Until all three exist in public print, nothing on this corridor has changed except the weather in your negotiations.

Do not sell flexibility in that weather. If a quote asks for exclusivity or monthly minimums in exchange for future rail pricing, counter with a term that expires unless the service publishes its own numbers, and keep the calendar rights to walk. Paper given away in 2026 comes due in 2029, and it always comes due at the wrong hour.

Get the geography settled for your own plants, because that is the part with no ambiguity. Ask whoever quotes you to name the nearest discussed location and give you the road distance to it in writing. Anything beyond about twenty-five kilometres of dray is not rail-shaped on any version of the arithmetic above, because two drays and two lifts will eat whatever the middle leg saves.

Then go after road capacity while you still own the leverage, which is right now, before the fleets decide anything. A two-year agreement on the inland legs, with a rate collar and a fuel adjustment written in, signed while the driver still has more trucks than work, will beat whatever spot market survives two years of everyone waiting for trains. This is where the money actually moves, not in the memorandum.

Set one date and keep it. If neither of the two nearest locations has announced a construction start within eighteen months of the 20 October 2026 signature, take rail out of your network plan entirely, reopen any rent premium paid for rail access and spend the effort on trucking instead. A study that has not broken ground in eighteen months is not a project, it is a habit.

Call it what it is: this is a road problem wearing a rail hat. Until a train runs on a timetable you can buy a slot on, this line still goes through the approach at Lach Huyen, forty kilometres of it, and the answer to whether you can use it this season has not moved one inch. Yes, you can use it, with a buffer day, booked early, through somebody whose phone gets answered at seven in the morning when the queue starts to form.

↑ Back to the key points

— By Ömer Kaya