Descartes Report: U.S. Container Imports Rebound 6.6% in May as China Shipments Surge 19.9%

The Analysis

When a month like May 2026 lands on my desk I do not start with the headline. I start with the map. Spread it out flat on the table, Atlantic on the right, Pacific on the left, the Gulf folded in along the bottom, and two million four hundred and thirty thousand TEU stops being a figure and becomes a distribution problem. Descartes Datamyne put US container imports for May at about 2.43 million TEU, six point six percent above April and eleven point five percent above May of last year, and my first reaction was not relief. My first reaction was to ask where it went, because growth that shows up in one month, out of one origin, through four or five gates, is a completely different animal from growth that arrives everywhere at once.

Two lines in the same release pulled me in opposite directions. The first is the geography: the East and Gulf coasts took forty four point eight percent of everything that came ashore, with Savannah up seventeen percent on the month and Houston up eighteen point two percent. The second is the speed: port transit delay at Long Beach came down from seven point three days to two point two days. One of those is a structural story measured in decades, the other is an operational story measured in weeks, and the mistake importers make is reading both with the same clock. I have spent enough years watching berths get built, channels get deepened and rail yards get graded to know which of the two will still be running long after the other one is forgotten.

What I want to do here is put the month into the right time frame, turn it into inventory and optionality, and then say what I would actually do with it in the next few weeks and the next few years. The month is real, the number is real, and I am not going to tell you it does not matter. It matters less than the shape behind it. Ten years ago this was a different map, and no single month is the reason it changed.

Start with China, because that is where the month came from. Eight hundred and sixteen thousand one hundred and ninety seven TEU originated in China in May, a jump of nineteen point nine percent over April, and that carried China's share of total US imports up to thirty three point six percent. I want to be careful with the arithmetic, so here it is in the open. If May was six point six percent above April, then April was about 2.28 million TEU. If China was up nineteen point nine percent, then April's China number was around six hundred and eighty thousand seven hundred TEU, which puts China's April share at roughly twenty nine point nine percent. The share therefore moved something like three point seven points in a single month, which in a market this size is not a rounding error, it is a lurch.

Now the number I actually care about. China added roughly one hundred and thirty five thousand five hundred TEU month on month, while the whole market added about one hundred and fifty thousand five hundred. Call it ninety percent. Nine tenths of America's import rebound in May came out of one country. That is the sentence I would write on the whiteboard, because everything else in the release is decoration next to it. When one origin supplies nine tenths of your growth, you are not looking at a broad recovery, you are looking at a concentrated one, and concentrated recoveries behave differently when something interrupts them.

India is the counterweight in the same data set, up twenty two point two percent on the month, and I want to be fair to it. Over a ten year horizon I think the subcontinent is the most interesting story in this map, and not only because of the growth rate. But the release gives me a rate and not a base, so I cannot size it, and I am not going to invent a base to make my sentence tidier. Here is how I handle it instead. Assume, purely as a placeholder, that India ran at one hundred thousand TEU in April. Then May is roughly one hundred and twenty two thousand two hundred TEU and the increment is twenty two thousand two hundred boxes, which is about one sixth of China's increment.

Push the assumption the other way and assume India ran at three hundred thousand TEU in April, a far larger base than most people would credit it with, and the increment becomes about sixty six thousand six hundred, still under half of China's. The conclusion survives both assumptions: the diversification is visible in the growth rates and not yet in the volumes. Growth rates are where a network announces its intentions. Volumes are where it keeps its promises. Watch the second one and be patient with the first.

Then there is the line in the same release that stops me from calling any of this a boom. Year to date imports remain one point nine percent below 2025. May came in eleven and a half percent above last May and the year is still behind. That is not a contradiction, it is arithmetic, and it tells you plainly that the months before May were running below last year. One strong month does not cancel a soft four months. If you sit in a planning meeting this week and hear that the market is back, ask which market and since when, because the person speaking may be describing a single print rather than a year.

I have watched too many procurement teams rebuild safety stock on the strength of one good month and then spend two quarters quietly working the inventory back down at a discount. The honest reading of May is that it looks like a catch-up month inside a year that is still slightly behind, with a pronounced tilt back toward a single origin.

Spread the map out again and look at the coastline. Forty four point eight percent of May imports came through East Coast and Gulf ports. Multiply that share by the month's total and you get about one million eighty eight thousand TEU landing on the Atlantic and Gulf side against roughly one million three hundred and forty one thousand everywhere else. Of the named gates, Savannah was up seventeen percent month on month and Houston up eighteen point two percent. Think about what that means in the time frame I work in. Port capacity is not a switch.

A berth is dredged, a channel is deepened, a rail yard is graded, an inland terminal is opened, and each of those takes years to build and then keeps paying out for decades afterwards. Savannah was not on the wall map of most importers I knew fifteen years ago the way it is now, and Houston spent most of its life being an energy and chemicals gateway that happened to move boxes as well. What you are looking at in May is not a monthly preference. It is the compounding of twenty years of concrete.

For an importer the significance is not that one coast is better than the other, because that argument has been going since the first container ship cleared the Panama Canal and it has never had a stable answer. The significance is that the population and the inland network have been moving for a generation and the boxes follow the people with a long lag. The map I actually look at is not the coastline, it is the set of inland points you can reach in two days of rail from each gate, and that map has been redrawn far more in the last fifteen years by yard projects and inland rail investment than by anything that happened on the water.

When a gate absorbs seventeen percent more boxes in a month, it is spending capacity that somebody decided to build a decade earlier. Your routing decision is downstream of that decision, and the correct response is not to chase it but to position against it. You want a footprint that can take volume through more than one coast, because the coast that wins the next decade is not knowable from one month of data and you should not have to be right about it.

That said, the Pacific side is not finished, and anyone writing it off on the strength of one month is misreading the same map I am reading. The West Coast gates sit closest to the factories that fill the boxes, they have the deepest water and the longest-established Asian service networks, and a seven point three day delay falling to two point two days at Long Beach is a reminder of how much capacity is sitting there when the queue clears. Geography gave that coast its role and geography has not changed. What changed is that it no longer has the role to itself. Two coasts with different strengths, both capable of taking real volume, is a better position for an importer than one coast with everything, and May is a fair picture of how that balance currently stands.

I should also say something about how I read these monthly releases, because the method matters more than the month. Port statistics are the tail end of decisions made a long way upstream. A box counted in Savannah in May was ordered months earlier, built before that, and routed by someone choosing between two or three gates. By the time the number is published the thing it describes is finished and cannot be changed, which means the only useful question is what the number tells you about the next decision you still get to make.

That is why I keep two clocks running in my head. The operational clock runs in days and weeks and covers dwell, appointments, yard density and the queue. The structural clock runs in years and decades and covers berths, channels, rail, inland points and which countries can make things at scale. Every number in this release belongs to one clock or the other, and almost every expensive mistake I have watched importers make came from using the wrong one.

Take the Long Beach figure, which most people will read as pure good news and which I read with one hand on the brake. Port transit delay fell from seven point three days to two point two days. Five point one days of sitting time came out of the system at that gate. If you move boxes through there, that is money, and I want to show you the money rather than wave at it. Assume you run two hundred TEU a month through Long Beach. Assume each TEU carries thirty thousand dollars of goods, a placeholder you should replace with your own average.

Assume your cost of capital is eight percent a year. Two hundred TEU at thirty thousand dollars is six million dollars of monthly flow, or about two hundred thousand dollars a day on a thirty day month. Five point one days of that flow is one million twenty thousand dollars of inventory that is no longer sitting in the terminal, and at eight percent that is roughly eighty one thousand six hundred dollars a year of carrying cost released. That is real money and you should book it.

Here is the part that gets missed, and it is the part that costs people. For several years congestion was functioning, without anyone's permission and without anyone's invoice, as free warehousing. A box that sat seven point three days in the terminal was a box you did not have to pay to store somewhere else, and a supply chain running on that accidental buffer quietly arranged itself around it: reorder points were set against it, promise dates were quoted against it, and the buffer never appeared in anybody's budget because nobody was paying for it.

Take it away and you do not simply save money. You also lose a shock absorber, and if you do not replace it deliberately you will discover the loss the first time a vessel slips or a booking rolls. Assume your daily demand is two hundred divided by thirty, call it six point seven TEU a day. Five point one days of pipeline cover is about thirty four TEU that used to exist for free. If you decide you still want cover, you now have to buy it.

Assume US side storage runs at twelve dollars per TEU per day, a figure you should check against your own warehouse contract. Holding thirty four TEU for a year at that rate is about one hundred and forty eight thousand nine hundred dollars, which is more than the eighty one thousand six hundred you released. Read that twice. Rebuilding the whole buffer costs more than the capital you freed up.

So do not rebuild the whole buffer. The arithmetic points at a specific answer and I will give it to you. Rebuild one day of it. One day of cover is six point seven TEU, and at twelve dollars a day that is about twenty nine thousand two hundred dollars a year. That leaves four point one days of released inventory, which is roughly twenty point seven TEU or about six hundred and twenty thousand dollars of goods, and at eight percent the capital release on that is about forty nine thousand six hundred dollars a year.

Against twenty nine thousand two hundred dollars of deliberate storage you come out somewhere around thirty six thousand dollars a year ahead, you still keep a one day cushion against a slipped sailing, and your cash cycle is four days shorter than it was. That is a decision you can defend in a budget meeting with the sums attached, and it is a decision you can only reach by doing the arithmetic instead of celebrating the delay number.

The same discipline belongs on your gate mix. If forty four point eight percent of the national book comes through the East and Gulf coasts and your own book is still seventy or eighty percent through a single West Coast gate, you are not diversified, you are inherited. I do not think the answer is to swing everything east, because the May gains at Savannah and Houston, seventeen percent and eighteen point two percent, are exactly the kind of numbers that show up later as yard density and appointment scarcity at those same gates.

Capacity that absorbs a surge in a month is capacity that feels the surge. The answer is a split you can defend and a trigger you have written down before you need it. My own trigger is uncomplicated: when one gateway carries more than sixty percent of my volume I start standing up a second one, and I give that project a quarter rather than a fortnight, because drayage, rail and a place to deliver from do not appear on command.

Standing up a second gate has a price and I would rather name it now than let you discover it in month three. You need a drayage or rail arrangement at the new gate. You need visibility of your containers there, which means someone or some system that can see them. You probably need at least one distribution point on that side of the country, because a gate you cannot deliver outbound from is a theory rather than an option. Assume ninety to one hundred and twenty days to stand it up, and assume it eats a real slice of a planner's week over that period.

There is also a floor below which a second gate is a distraction rather than a hedge. If you move fewer than about fifty TEU a month, splitting that across two coasts buys optionality you cannot exercise and costs you the attention you would have got by concentrating. Below that level, buy the optionality a different way: make sure your inland provider can move a box either direction even if you normally ship only one.

Now the Strait of Hormuz, which the release flags as an active disruption, and here I want to be geographically honest because this is where commentary usually overreaches. Hormuz is not on the container lane into Houston. It is on the lane that sets the price of the fuel your ships burn and the insurance your cargo sails under, and it sits on the same map as a Gulf coast that just absorbed eighteen point two percent more boxes in a month. Those are two different exposures and they behave on two different clocks. The energy and insurance exposure shows up in your cost lines with a lag of weeks and it does not care which port you use.

The routing exposure only shows up if a disruption runs long and hard enough to change how operators deploy tonnage, and that is a quarterly question rather than a monthly one. If you are building a Gulf heavy network, and the May numbers say plenty of people are, you are taking on exposure to a geography whose energy routes are contested. That is not a reason to stay out of the Gulf. It is a reason to have made the choice knowingly and to keep a second coast genuinely live rather than theoretically available.

US trade policy is the other flagged risk, and I am going to stay out of the detail deliberately, because detail is not where I can be useful. What I can speak to is the network level version. Policy rarely changes your volume on the day it changes. It changes where the next factory gets built, which takes two to five years, and then it changes which gates those factories ship through, which takes a few more years to surface in port statistics. That is precisely why a nineteen point nine percent monthly surge out of China reads to me as positioning rather than as a verdict.

Cargo owners move ahead of policy when they think they can see it coming, and the May number has that shape: a pull forward concentrated in one origin, arriving in a month when the year to date is still one point nine percent down. If that reading is right, the second half of the year hands some of it back, and the importers who treated May as a new baseline will be the ones left holding the inventory.

There is a version of this where I am wrong and I should say what it would look like. If June and July come in at similar levels and the year to date gap closes, then May was not a pull forward, it was a turn, and the right response changes: you build inventory into a rising market rather than hedging against a false dawn. The test is two consecutive months of year to date improvement, not two consecutive months of month on month growth, because month on month growth in this series is noisy and year to date is not. I would also be wrong about the concentration if the next two prints show China's share coming back down toward thirty percent while other origins hold their absolute gains. That outcome would be the healthier one, and it would argue for patience rather than for a restructuring of your sourcing map.

So what I would take from this month is not a forecast. It is three positions to hold, and none of them depends on being right about the next print. Hold a gate mix that does not require you to be right about which coast wins the next decade. Hold a deliberate buffer that you pay for on purpose, sized by arithmetic, instead of an accidental one you used to receive free from congestion. Hold an origin mix that is measured in TEU rather than in growth rates, and give that one a horizon of years, because that is how long it takes a country to become a genuine alternative source rather than a promising line in a monthly report. The month tells you where the network currently flexes. It does not tell you where it settles, and it is settling that you are being paid to anticipate.

I have been looking at maps like this one long enough to have watched a succession of ports become inevitable in hindsight, and the thing they all had in common was that somebody started building before the volumes justified it and finished long after the volumes arrived. What changed in May is not that America bought more from abroad, though it plainly did. What changed is visible only if you put two months side by side: the boxes concentrated into one origin, the coasts rebalanced toward the Atlantic and the Gulf, and the queue at Long Beach drained. The queue will fill again, because queues always do. The coasts will not move back, because concrete does not move back. If you spend your energy watching the queue you will be working on the wrong clock, and the clock is the whole game.

↑ Back to the key points

— By Tomás Ferreira