AI infrastructure shippers that have not yet booked peak-season capacity risk pushing airfreight rates sharply higher, according to pricing data provider TAC Index. The pressure comes from shipments of servers, chips and data centre equipment landing on an already firm market.

Sources cited by TAC Index report a clear split in preparedness. Ecommerce customers and other major shippers with product launches have already secured peak capacity. Some AI hyperscalers, by contrast, have yet to book their space.

That gap matters. Peak season rewards the early and punishes the late, and the late arrivals this year include companies moving some of the most valuable cargo in the market.

What is driving the demand?

The demand side of the equation is dominated by artificial intelligence infrastructure. Servers, chips and data centre equipment are all moving by air, and all are competing for the same belly and freighter capacity that ecommerce has already reserved.

The market was firm before this demand arrived. The warning from TAC Index's sources therefore lands on a market with little slack to absorb it.

How firm is the market?

The signal comes with one counter-intuitive data point. TAC Index's global Baltic airfreight index dipped during Golden Week, the annual Asian holiday that suppresses factory output and shipments.

A Golden Week dip is seasonal, not structural. The concern raised by sources is that the underlying market remains firm enough that unbooked AI cargo could send rates sharply higher once peak-season demand resumes in earnest.

Who has booked and who has not?

According to the sources reported by TAC Index, the divide is as follows:

• Booked: Ecommerce customers and major shippers with product launches tied to the peak season.
• Not yet booked: Some AI hyperscalers, still seeking capacity for data centre equipment, servers and chips.

The pattern is unusual. Hyperscalers are typically sophisticated buyers of logistics capacity, and their absence from peak bookings suggests either late planning or an expectation that capacity will be available at prevailing rates.

If the second expectation proves wrong, the cost will land on them first and on other shippers second.

What happens next?

The arithmetic is unforgiving. Capacity in the peak season is broadly fixed, and ecommerce has already claimed its share. AI cargo arriving late to the booking window must compete for what remains, and price is the mechanism that resolves that competition.

Rates rising sharply is the outcome TAC Index's sources have flagged. For shippers of lower-value goods, that outcome would mean sharing the bill for the AI build-out whether they participate in it or not.

The question for the weeks ahead is whether the unbooked hyperscalers secure space at current levels or at whatever level the market sets once their demand hits it.