Air cargo rates drift lower but AI boom keeps them aloft

The Analysis

WorldACD's numbers for the week ending July 5 are in, and the headline says air cargo is cooling: the global average rate sat at $3.13 per kg, spot at $3.62, both drifting down as Gulf capacity comes back and jet fuel eases. But before you rush to renegotiate every contract, look at the second number in the same release: AI-related semiconductor shipments saw global sales up 106% year on year in April. That is the engine now, not the e-commerce boxes that filled bellies through the last cycle.

What this means for a shipper moving goods by air is uncomfortable. The rate you see printed is an average. The capacity behind it is being pulled toward a lane that pays a premium and tolerates delay far less than your deferred freight does. Yesterday a client asked me if now is the time to lock a cheaper rate, and my answer ran opposite to what they expected.

Start with the fact slice. Three numbers matter more than the trend line. One: $3.13 per kg average, $3.62 per kg spot, week ending July 5. Two: that average is still well above pre-crisis levels even after the drop. Three: semiconductor shipments are up 106% year on year in April, and they are now the lead growth driver replacing e-commerce. Everything else in the release is commentary around those three.

Why this is happening now. Two forces are pushing in opposite directions at the same time, and that is why the market feels frozen instead of falling. The downward push is real: Gulf capacity that disappeared is returning, and jet fuel has eased, so the cost floor under each kilo is lower than it was. The upward push is also real and structurally stickier: AI chip volumes are growing fast enough that they absorb belly and freighter space which used to be available to general cargo. When one shipper's product is both time-critical and high-value, it wins the allocation every time. So rates do not crash; they hover.

Who feels it and by how much. The layered impact is not even. Large electronics and semiconductor shippers with volume commitments get priority boarding and often pre-negotiated blocks of space, so their effective cost is shielded even if the published rate stays high. Mid-size forwarders and brands shipping deferred air get squeezed twice: they pay the elevated average and they lose space to the chip lane at peak windows. Small shippers on spot are the most exposed. They see the $3.62 spot, they do not get the allocation, and they wait.

Transmission timing. The softness in the printed rate takes roughly two to four weeks to reach a contract you sign today, because carriers reprice on published indices with a lag and your quarterly deal was struck against a higher baseline. The jet fuel component is already moving: surcharges track the Gulf Coast kerosene price with about a two-week lag, so the easing you read about this week shows in your invoice in mid-July. Plan cash flow against the invoice date, not the news date.

The overlooked point. Most coverage frames this as a price story. It is actually an allocation story. When AI semiconductor volume takes a freighter belly, your regular consolidation slot does not get cheaper, it gets cancelled or pushed. The killer for an importer or exporter is not the $0.20 per kg drift down; it is the Tuesday your 800 kg of samples does not fly because a charter took the whole aircraft. Capacity, not cost, is the constraint you should be managing. Build the backup lane before you celebrate the lower average.

Counterfactual and where it breaks. If AI demand cools or Gulf capacity fully restores, the average could fall 15 to 20% from here and spot pressure eases across the board. If the Middle East disruption flares again and removes capacity the way it did earlier this year, spot snaps back toward the highs and allocation gets brutal. The number to watch is the gap between total available tonne-kilometres and AI-driven demand. When that gap closes, your service level, not your rate, is what moves.

Worked example with assumptions stated. Assume a shipper moves 8 tonnes per month of consumer electronics by air at the $3.13 per kg average. Against a pre-crisis baseline of roughly $2.40 per kg, the premium is $0.73 per kg, so the monthly overpayment is 8,000 kg times $0.73, about $5,840, or roughly $70,000 a year, before any space-loss penalty. Now assume 30% of a key lane's weekly belly is diverted to chip charters during peak windows. On a 20-tonne weekly consolidation that is 6 tonnes bumped, and at a typical three-day rebooking delay the carrying cost on $40,000 of goods at 1.2% monthly financing is about $48 per day of slip, roughly $144 per affected week.

The rate drift saves you almost nothing next to the allocation risk. The move is to secure space, not chase the discount. One more operational note that gets lost in the weekly numbers. The printed rate resets on a schedule you can see, but the allocation decision happens in the hours before each flight closes. If you only manage the rate, you are managing the symptom. Put a named owner on capacity, not price, for your top lanes, and review it weekly against the chip-volume signal instead of waiting for the invoice to tell you something broke. The clients who sailed through the last crunch were the ones with a backup lane already warmed up, not the ones who negotiated the best penny per kilo.

Think about inventory posture while you are at it. When the constraint is space rather than cost, the cheap hedge is buffer stock positioned nearer the destination, so a missed flight does not become a missed shelf. A shipper holding two extra weeks of cover at the import end absorbs a three-day slip without touching the customer. That cover costs money, but it is predictable money, unlike a spot rate that doubles on a Tuesday. The math favours the buffer the moment allocation risk is the binding constraint, which is the market the AI chip lane has created.

↑ Back to the key points

— By Nadia Pryce