The numbers look almost wrong. In 2025, AI-related goods made up just 7% of global air cargo volume — yet they accounted for 53.5% of its total value. Korean Air's second-quarter cargo revenue jumped 46% year over year, and its executives say the growth engine is no longer e-commerce. It is AI chips, server racks and data-center gear.
June freight demand on the Asia-North America lane, the busiest international corridor, grew 14.7% year over year — the fifth straight month of growth, according to IATA. This is not a blip. Air freight has become the fastest-visible window into how quickly the AI infrastructure buildout is actually moving.
Seven Percent of Volume, Half of All Value
Value density is the headline. IATA estimates that in 2025, AI goods contributed 53.5% of air freight value while occupying only 7% of volume. A single AI server rack can carry hundreds of thousands of dollars of equipment. When that much capital is in transit, weeks at sea are not just slow — they are expensive.
That is why air dependency is so extreme: roughly 56% of servers ship by air, 68% of data-storage units, and memory chips are close to 100%. E-commerce once filled the bellies of freighters; now compute hardware does.
Why AI Hardware Has to Fly
Three forces push AI hardware into the sky:
Value density. High-value, time-sensitive, physically small cargo is what air freight exists for. Sea shipping ties up tens of millions in working capital for weeks and adds physical risk.
Deployment deadlines. For hyperscalers and large tech firms, a delayed data-center launch carries real operational and financial damage. Air cuts a multi-week ocean crossing to days, which matters when equipment must be on-site by a specific go-live date.
Short technology cycles. AI chips turn over fast. Companies air-freight the latest silicon to keep their edge, and Korean Air says hardware orders already extend two to three years out — a sign of how deeply the deployment pipeline is committed.
Air Freight Is the Physical Layer of the AI Economy
Container ships carry the consumer-goods economy; widebody freighters now carry the compute economy. The switch from e-commerce to AI hardware as the growth engine is structural, not cyclical — and that makes logistics data a leading indicator that lands before quarterly earnings do.
The geography tells the same story. Spot rates from Northeast and Southeast Asia to North America rose 33% between late February and late July, while Europe-North America rates fell 27%. The buildout is concentrated in a handful of Asian hubs: Taiwan for advanced chips and AI servers, Korea for high-bandwidth memory, with Malaysia, Thailand and Singapore also tight. Japan Airlines is already restructuring its freighter network around semiconductor hubs, expanding services linking Taipei, Bangkok and Hanoi with Tokyo Narita.
The trend connects to the broader supply-chain story we covered in China's indium grip on the CPO supply chain — the same buildout is showing up in both materials and logistics. And as data centers themselves get denser and more automated, infrastructure is becoming a self-optimizing system, as we explored in AI optimizing AI infrastructure.
How to Read the Barometer
Air cargo is now a real-time gauge of AI capex. Watch three signals, and treat them as a set rather than individually:
Capacity tightness. If demand for high-end freighter capacity on Asian lanes stays strong and spot rates hold, the buildout is still accelerating.
Lead times. If semiconductor orders, wafer demand and chip packaging lead times start shortening, the production surge may be cooling.
Inventory dwell. If parts linger longer in hub warehouses — inventory accumulating faster than it ships — that is an early warning.
For operators, the practical takeaway is simple: the airplane is now part of the critical path. Build air freight cost and lead time into deployment schedules, and watch the cargo data — it will tell you where the AI cycle is heading before most earnings calls do.
