Supply Chain Action Points
Read this first — the conclusion, and the moves to make:
- Turn your October to December forecast into kilos per week per trans-Pacific lane by 31 July, and get a written allotment with a named flight number and stated cut-off for every week of the peak.
- Split trans-Pacific volume across two carriers before 15 August, even at a 5% to 8% rate penalty, so no single allotment cut can strand a full week.
- Move the decision point for any shipment above 300 kilos from T-72 hours to T-14 days, and require written space confirmation at that mark.
- If you have any North America to Asia return flow, bundle it into the outbound negotiation at the next contract renewal; a two-way shipper trades for space instead of bidding for it.
- Write a hard rule that no trans-Pacific express shipment is promised with under 72 hours of buffer inside the door-to-door commitment, and re-quote every SKU that fails the test by 31 August.
AI-related hardware now accounts for nearly 10% of total air-cargo volume but is concentrated on the trans-Pacific, competing with parcel traffic for scarce belly and freighter space. Forwarders report tightening capacity and elevated rates on Asia–North America express lanes.
The Analysis
On a Wednesday at 18:00 Shanghai time the trans-Pacific express week is already decided, and almost nobody in the chain notices. The aircraft does not leave for another thirty hours. The cut-off is still up on the board. But the pallet positions were spoken for six weeks earlier, and by Wednesday evening the only open question is which of your cartons gets left behind.
That is the part of the AI hardware story that gets buried under the headline. AI-related goods are under 10% of total air cargo volume, a figure now repeated everywhere, but the same reporting puts 12% of global air cargo capacity out of service overnight on 28 February, first-half supply growth at 1%, demand growth at 4%, and global rates up 17% year on year. Spot rates in May ran roughly 40% above a year earlier. Global semiconductor sales more than doubled in April, up 106%.
Ten percent of the tonnage does not sound like enough to starve a lane. On the trans-Pacific it is, and the mechanism is not price. It is the calendar.
Look at the shape of the lane, not the size of it. That 10% is a global share; on the trans-Pacific it is a much bigger slice of what actually moves, because that is where the hardware is made and where it is going. At the same time the other half of the lane is shrinking: China's low-value and e-commerce exports fell 7% year on year in May, the sixth consecutive monthly decline. So the lane's total is not collapsing, demand still grew 4% in the first half, but the composition inside it has flipped. A lane that used to carry containers of light parcels now carries heavy, expensive equipment with a parcel tail attached. Every operating assumption built for that tail is now wrong.
The physics is where it bites. A loaded server rack is dense, several hundred kilos in barely more than a cubic metre. A parcel carton is the opposite: a 40 by 40 by 50 centimetre box weighing eight kilos bills as sixteen because of volume. Put them on the same aircraft and they consume two different things. The rack eats weight and main-deck pallet positions. The carton eats volume down in the belly. That distinction decides who gets squeezed first, and it is why the headline number misleads anyone shipping cartons.
The sequence runs like this. Freighter main-deck positions tighten first, because that is the only place a 900-kilo rack can go. Integrators and consolidators who block-book main deck for express products lose part of that block, and they do what anyone would do, which is push the overflow into belly space on passenger widebodies. Belly space is where parcels live. So the parcel shipper does not get squeezed by an AI shipper directly. He gets squeezed by his own forwarder five to seven weeks later, when the belly allocation comes back smaller than last quarter's. That lag is exactly why rates moved before space did. A price is a quote. Space is a schedule. They do not change on the same day.
Who hurts first, in plain terms. Below 500 kilos a week on the lane you have no allotment worth the name, you buy at T-3, and you are the first name on the roll list. Between two and eight tonnes a week you have an allotment and you will watch it get trimmed at T-14, with a polite email. At twenty tonnes, or with your own block, you are fine, more than fine, you are the reason the rest of the lane is tight, because your volume is the volume the carrier wants. Scale does not buy immunity from the market. It buys a chair in the room where the allocation is argued about.
Direction matters as much as size. The tight side is Asia to North America, outbound. The return leg is loose, because nobody is flying server racks back to Shenzhen. If you have any bidirectional flow, even a modest one, you hold a card a one-way shipper does not: bundle the return into the outbound negotiation and you are not bidding for space, you are trading for it. Most shippers never try, because the two flows sit in different budgets and two different people.
The clock, hour by hour, is the part worth memorising. Forty-two days out, the allotment is set. Twenty-one days out, it is firmed. Fourteen days out, the cuts land. Seventy-two hours out, the cargo closes. Twenty-four hours out, your last option is to split the shipment across two flights and accept that half arrives a day late. Six hours out, the truck is at the dock and there is nothing left to decide. Everything you can still influence happens inside the first of those windows. Almost nothing happens in the last.
Which brings me to the thing the coverage keeps missing. Everyone is counting tonnage, and tonnage is the wrong denominator if you ship cartons. What matters is chargeable weight per pallet position, and on a trans-Pacific freighter that number runs four to six times higher for AI hardware than for parcels. Ten percent of the kilos can be twenty-five percent of the positions in a given week, because positions are allocated by build, not by weight. Here is the part that actually costs people money: your parcel did not lose its space at the cut-off. It lost it forty days earlier, in an allocation sheet, to a shipper whose cargo weighs five times what yours does per cubic metre. By the time the cut-off arrives you are not competing. You are queuing for leftovers.
Do the arithmetic on a small shipment. Assume a thousand kilos of parcel freight, twelve pallets, a four-day door-to-door promise, and, this is my assumption and not a number from the reporting, eight hours of clearance and twelve hours of final mile. That leaves seventy-six hours in transit and about fifty-six hours of buffer inside the promise. One roll of twenty-four hours eats almost half of it. Two rolls and you have broken the promise on paper even if the cargo is only a day late in reality. On a thousand kilos at a spot level roughly 40% above last year, the second roll is not a service problem, it is a penalty clause problem.
The fix is boring and it has to happen this month, not in October. Take your October to December forecast and turn it into kilos per week per lane. Kilos, not dollars. Then ask your forwarder for a written allotment at that number, with a named flight and a stated cut-off, covering the whole peak. If he cannot give you a flight number, you do not have an allotment, you have a wish. Split the volume across two carriers even at a worse rate, because a single-carrier allotment is a single point of failure and this lane has already shown it can lose 12% of its capacity in one night.
The trap is reading the current easing as a trend. Rates drifted down a little in early July on other lanes and everybody relaxed. On this lane the thing that decides your month is not the rate, it is whether the flight you were promised is still in the schedule on the day you need it. Check the schedule, not the quote. If your Wednesday 18:00 allocation is short, you want to hear it on a Monday in August, not on a Wednesday in October with the truck already at the dock.
↑ Back to the key points