Supply Chain Action Points
Read this first — the conclusion, and the moves to make:
- By 10 September, compare September bookings against the fourth-quarter plan; if September already exceeds 40% of planned fourth-quarter air volume, treat the peak as pulled forward and move budget this month.
- Lock positions for the first three weeks of September covering at least 120% of last year's same-period tonnage, with the number of kilos stated per flight rather than per month.
- Confirm buildup readiness in writing for your top five air SKUs: pallet supply, screening slot and cutoff time at the origin terminal, checked before 20 September.
- Split the air book by yield, with AI-adjacent and high-value SKUs on contracted positions and general merchandise on a named backup lane, so one displacement event cannot take the whole programme.
- Set a written trigger: a quoted rate up more than 10% week on week for two consecutive weeks moves the designated 10% of volume to ocean with an air tail instead of chasing space.
The Middle East conflict removed about 12% of global air cargo capacity overnight in late February and lifted first-half rates 17% year on year; AI hardware now drives the transpacific, with April semiconductor sales up 106%. Flexport warns renewed disruption and ocean-to-air shifts could pull the peak to early September rather than Q4, and Xeneta has revised its 2026 air-rate view to +5-15%.
The Analysis
Late February, air cargo lost roughly 12% of global capacity more or less overnight when Middle East airspace shut. First-half rates ran 17% above the year before. The transpacific is now being pulled by AI hardware, with April semiconductor sales up 106%. The warning going round is that renewed disruption plus an ocean-to-air shift could pull the peak into early September instead of the fourth quarter, and the 2026 air-rate view has been revised to plus 5 to 15%.
An early peak sounds like a scheduling curiosity. It is not. Capacity is added on a calendar, through charter aircraft, extra sections and winter belly space, and a peak that arrives early arrives before any of that is in place.
If you fly anything out of Asia in the next eight weeks, the question worth asking is not whether rates go up. It is whether a position exists for your freight on the day you need it.
Twelve percent of global capacity removed overnight usually gets read as a flight-count number. It is not only that. When airspace closes, the flights that keep operating get longer. A longer routing burns more fuel, fuel is weight, and weight comes out of payload. The capacity that disappeared is therefore larger than the departures that stopped, because every surviving flight is lifting less than it did before. Treat 12% as the floor of the loss, not the total.
Then add the cargo mix. AI hardware is dense, urgent and pays well per kilo, which is the profile an airline wants when weight is the binding constraint. What gets displaced when that freight takes the weight is the low-yield general merchandise most importers actually ship. This is allocation, not conspiracy. The operator fills the aircraft with freight that defends its own space, and general cargo does not defend anything. A 106% jump in semiconductor sales is therefore not only a story about chips. It is a statement about what your cartons are bidding against for the same position.
The ocean-to-air shift adds a second problem that is purely physical. Cargo built for a container is not built for an aircraft. It is palletised for a forty-foot box, sometimes floor-loaded, often over-height for a lower deck. At the terminal it has to be rebuilt into unit load devices, screened, and staged on a buildup floor with a fixed number of square metres and a fixed number of cans. In a normal week that floor has slack. In the week when ocean reliability fails and every importer calls a forwarder on the same Monday, the floor and the cans run out before the aircraft does. The constraint at that moment is not lift. It is the build.
Who gets hurt in what order. A shipper holding a block space agreement is protected on positions and exposed at the point where the carrier asks for a surcharge or asks for part of the block back. A mid-size shipper has no block and no leverage; its forwarder will find space at a number, quoted with a validity measured in hours. A small shipper does not get a number at all, it gets a call saying the booking did not go. On direction, transpacific eastbound is where the AI hardware sits, so a general merchandise importer on that lane is the one being displaced. Perishables and anything with a shelf life pay whatever is quoted, because the alternative is a total loss, which is why the top of the distribution moves first and hardest.
The early-peak claim is where the calendar does its damage. Extra capacity arrives with schedules. Charter freighters are contracted ahead of the season, extra sections have to be filed, and belly space grows when the winter timetable begins. A peak that lands in early September lands on the summer schedule, before any of that. The market is asking for charter aircraft in a month when they are already committed, and the answer is no. So the squeeze is not simply bigger. It is shorter and steeper, and it will feel like a two-week wall rather than a three-month slope. A wall is harder to plan inventory around than a slope, because there is no week where you can catch up.
The number that gives the game away is inside the forecast. The 2026 view sits at plus 5 to 15%, while the first half already printed plus 17%. A range whose top is below the realised first half is a forecast that assumes the second half calms down. If the peak pulls into September, that assumption is what breaks, and the run rate does not decelerate, it extends. The second-half air budget most shippers built, which will have used the low end of that range, is built on a number the market has already beaten.
There is a second consequence that planning models tend to miss. The same capacity loss that lifted rates also removed the slack that used to absorb ocean overflow, so ocean and air are now correlated within days rather than weeks. A model that treats the two as substitutes with a lag will under-forecast air demand in precisely the week the switch happens, and it will do so in the same week the buildup floor runs out of cans.
Run the account. Assume an annual air budget of one million dollars with half of it in the second half, and that split is my assumption, not from the release. If the second half decelerates to the bottom of the range at plus 5%, the increase on that half is twenty-five thousand dollars. If the peak pulls forward and the second half holds the first-half run rate of plus 17%, it is eighty-five thousand. The gap between those two planning cases is sixty thousand dollars, and the only difference between them is an assumption about which month the peak arrives in.
Waiting for confirmation is the expensive option, because confirmation arrives as a rate. Pre-buying too much is the other error, since unused block space is a straight loss with no salvage. The trap in the middle is assuming an early peak means an early end. It usually means the same tonnage compressed into fewer weeks, so the weekly requirement rises even when the annual total does not, and a plan built on annual tonnage runs short every week of the wall.
The useful move today is not a forecast. It is a call about positions for the first three weeks of September, and an answer in writing about how many kilos you can have on which flight.
↑ Back to the key points