Global air freight index BAI00 +0.2% to 2,444, up 20.2% year-on-year to Sep 14

The Analysis

An index is an average speed on a motorway. It tells you traffic is moving. It does not tell you the bridge at kilometre forty is down to one lane. BAI00, the Baltic Air Freight Index, is 2,444 for the week to 14 September, up 0.2 percent on the week and 20.2 percent above a year ago, and after enough years on these corridors I can tell you the number that decides whether your cargo makes its date is never the average. It is the one leg that jams.

Shanghai outbound, BAI80, came in 0.7 percent lower on the week and still 18.5 percent above last year. Frankfurt, BAI20, rose 3.2 percent to 1,139. Chicago, BAI50, jumped 15.1 percent to 1,337. Four numbers, four different roads, and only one of them is the road your box is on.

Fuel is why this is happening now. Jet has more than doubled year on year on Platts, crude is above 100 dollars a barrel, and forwarders are already saying there is another leg up coming before peak. What they do not say, because it is not in the data, is that the alternative corridors fill up before the index admits there is a problem. That is the part I want to walk you through, because it is the part you can still act on.

Start with what an average is. BAI00 blends every lane in the panel into one number, so a 0.2 percent weekly move can hide a 15 percent move in one direction and a fall in another, and that is exactly what this week looks like. Reading the headline and stopping there is how people end up booking the wrong gateway three weeks late.

Shanghai is the pool everything else is measured against. A dozen carriers put capacity there because that is where the tonnage is, so when fuel doubles, Shanghai is the last place to show it. The first carriers to cut frequency cut it where the yield is thinnest, not where the volume is. That is why BAI80 can print minus 0.7 percent in the same week Chicago prints plus 15.1. One is a pool with depth. The other is a gateway that just lost a freighter rotation.

Frankfurt at 1,139 and up 3.2 percent is the European end of the same road, and it behaves like a European end always does. FRA is not a destination, it is a junction. The cargo lands, clears, and goes by truck to Poland, Czechia, Hungary, the Rhine, sometimes further. The airport rate is the visible half. The truck that leaves the cargo terminal at six in the morning is the invisible half, and it is the half that decides whether your three days of air speed survive the last six hundred kilometres.

Chicago is where the week actually breaks. Fifteen percent in seven days is not a demand story. Somebody moved metal. A freighter rotation is not a tap you turn on, so when a carrier decides the Atlantic and the US interior are paying better than the lane it is on, that lane loses a frequency and the rate on it jumps while the average barely moves. If you ship into the US interior, BAI50 is your early warning and it just went off.

The fuel is the reason underneath all of it. Jet up more than 100 percent year on year is a cost that scales with tonne-hours, not with pieces, so a long sector carries a much bigger fuel bill per kilo than a short one. On a short intra-Asia hop the fuel share is manageable. On Shanghai to Chicago it is the dominant line in the cost stack. That is the arithmetic behind the 15.1 percent, and it is why the spike landed where the sectors are longest.

Now the first thing the coverage did not say. The weekly number is a lagged number, not a live one. Fuel surcharges reset on a published schedule, and between resets the rate you are quoted is carrying a fuel assumption that is already old. So a flat week during a doubling of jet is not a plateau. It is a queue of resets waiting their turn. The bill for this week's fuel arrives over the next four to eight weeks, in tranches, on whatever dates the mechanism happens to use. If you budget off the flat print, you are budgeting off a photograph of a moving lorry.

What costs real money is the order in which things run out. The release valves on this trade are the rail corridor through Khorgos and Alashankou and the truck corridor over the same passes, and both of them sell their peak capacity early. Rail planners take their December slots in September because the terminals at the Polish border and the gauge-change yards at Brest are the bottleneck and everybody knows it. So the sequence is: index stays quiet, everybody stays on air, the rail slots go, the index finally prints the peak in October, and at that point the valve is shut and the only option left is the expensive one. You do not buy the alternative when the index tells you to. You buy it before, or you do not get it.

Walk the US road properly. Shanghai to Chicago has three shapes. Nonstop PVG to ORD, which is the fastest and the first to be repriced. PVG to Anchorage to ORD, where Anchorage is a fuel stop and a crew change and, on some routings, a gateway in its own right, and do not underestimate that airport, because a tech stop that saves fuel uplift on a long sector is worth a lot when jet has doubled. And PVG to the West Coast, then rail or truck into the interior, which trades a day or two for a domestic leg you can actually price.

The West Coast shape is the one people forget in a spike. A box that lands in Los Angeles and rides the double-stack to Chicago is not as fast as a through air waybill, and it is not meant to be. It is meant to be the thing you can still book when ORD is full and the through rate has gone somewhere unreasonable. The mistake is treating it as a backup you will arrange on the day. Domestic capacity into the interior tightens exactly when air into ORD tightens, because both are being pulled by the same demand, so the backup has to be booked while it is still cheap and boring.

Europe is a different road with a different failure mode. Shanghai to Frankfurt, Amsterdam, Liege, Paris, Luxembourg, all of them work, and the airport choice matters less than what happens after wheels down. Frankfurt gives you the best trucking web into Central Europe and the worst congestion at the cargo terminal on a Monday. Liege gives you a quieter terminal and a longer road. Amsterdam sits in the middle. Pick by where the consignee actually is, not by which airport you have always used, because the last six hundred kilometres is where a three-day air transit turns into five.

Istanbul deserves its own paragraph, because it is the hinge nobody puts on the slide. IST takes the Central Asian and Middle Eastern tonnage and feeds it into Europe by truck through Kapikule, and it takes European cargo the other way. When the direct Europe lanes get expensive, Istanbul becomes a transfer point rather than a destination, and the truck leg from there into the Balkans and Central Europe is a road I have run more times than I can count. It is not faster than a direct lane. It is a pressure valve, and pressure valves matter in exactly the weeks we are in.

The rail corridor is the one I know best, so let me be precise about where it hurts. The box goes by rail from the Shanghai area to the border at Alashankou or Khorgos, changes to the Kazakh network at Dostyk, runs north through Almaty and Ekibastuz and Yekaterinburg, crosses into Belarus, and then hits the real bottleneck at Brest, where the gauge changes from the wide Russian standard to the European one and every container is physically handled. After that it is Malaszewice and on to Duisburg or Hamburg or Prague. Two gauges, one handover, and the handover is where the schedule goes to die.

Think of the gauge change as a tyre change in the rain. Everybody can do it. Nobody does it fast, and if the yard is full the car sits there. That is why a rail plan that looks fine on paper in July is a different plan in November, when the same yard is taking everybody's peak volume and the weather is against the crane. When I quote a rail transit to a customer I quote the yard, not the timetable.

The truck corridor over the same passes is the flexible one and the fragile one. Khorgos into Kazakhstan, north through Aktobe and Kazan, into Belarus, and then the same Brest crossing, or south around the Caspian. Truck is faster than rail and much more exposed to a single bad border, a single permit problem, a single driver who cannot be replaced at peak. In the last quarter of the year the good drivers are already committed, so the rate you are quoted in November is for the driver who was left over.

The Caspian route is the one that looks elegant on the map and teaches patience in practice. Kuryk or Aktau on the Kazakh side, ferry across to Baku, then rail or truck through Georgia and on to Kars and Istanbul. Two ferries if you are going further, two borders, two handovers. In summer it is a genuine alternative and the arithmetic works. In winter the crossing is weather and the schedule is a suggestion. If your cargo has a hard date in December, this is not the corridor you bet the date on.

Count handovers, not kilometres. That is the only rule I trust on these routes. Every handover costs a day, a handling risk, and a document you did not have before. Air has one handover, at the destination airport. Rail has two or three. The Caspian route has four. When someone shows you a route that is cheaper per kilo, ask how many times the box is touched, and price the touches before you believe the number.

Now the impact by shipper, because a 20.2 percent year-on-year index does not mean the same thing to everybody. Consumer electronics is the cleanest case. High value per kilo, short product cycles, a customer who counts days, so air is not a choice, it is the business model. For that shipper this is a straight cost increase with no mode escape, and the only levers are density, consolidation and which gateway. Density is the one people skip. A pallet built properly carries more value in the same cubic, and at these rates cubic is what you are paying for.

Auto parts and aircraft-on-ground spares are the ones that cannot negotiate at all. When a line is down, the freight is not a cost, it is an insurance premium, and everybody in that chain knows it. What can be managed is the pipeline ahead of the emergency: keeping a small buffer of the parts that stop a line, moved by rail while rail is cheap, so the air spend is reserved for the genuine break. That buffer is the cheapest air freight you will ever buy, because you bought it as rail.

Apparel and fast fashion sits in the middle and has the most room to move. A season that slips a week is expensive but survivable, which means a portion of that tonnage can go by rail when air goes where it went this week. Not all of it. The replenishment that chases a sold-out size has to fly. The bulk that was ordered in June and only shipped late can ride.

Machinery and project cargo mostly should not be on air at all, and the interesting question is why some of it is. Usually because a factory date slipped and nobody told the freight desk until the ship was full. That is not a rate problem, it is a planning problem, and it shows up as air freight in the fourth quarter every year.

For parcels and small e-commerce volumes the answer is consolidation, and it is the same answer it has always been. Loose pieces pay the worst rate on the lane and get the worst handling. A consolidated pallet pays less and gets touched less. What changes in a week like this is that the consolidation cut-off gets earlier, because the consolidator is filling a given pallet and once it is full it is full, no matter how good your relationship is.

By corridor the picture splits cleanly. Europe is the market where rail is a real alternative, because the corridor ends in Poland and Germany and the consignee is often within a day of the terminal. The US interior is the market where air is close to unavoidable and the only question is which gateway and which domestic leg. Central Asia and Turkey are the markets where truck and the Caspian route are live options, with the caveat about winter.

On timing, three clocks are running. The first is the surcharge clock, which is days to a couple of weeks, and which nobody controls. The second is the capacity clock, which is weeks, and which is when the forwarders commit their blocks for October to December and when the rail terminals close their peak allocation. The third is the calendar clock, which is the pre-holiday build in January, when everybody ships before the factories stop, and which is the reason the fourth quarter never gets cheaper on its own.

The surcharge clock is already running and you cannot do anything about it. The capacity clock is the one with a decision in it, and the decision is due now, not in October. The third clock is the one that decides whether next year starts with a cost problem or a service problem.

Let me put numbers on this, and I will be explicit that all of the unit rates below are my own working assumptions for illustration, not quotes from any carrier or index. I am using them to show the size of the exposure, nothing else.

Assume a shipper moving 30 tonnes a month out of Shanghai, which is 30,000 kilos. Assume a blended Shanghai outbound rate of 3.50 dollars a kilo, which is consistent with a lane running 18.5 percent above a year ago from roughly 2.95. At 3.50 the monthly air bill is 105,000 dollars. A year ago, at 2.95, the same 30 tonnes was 88,500. The year-on-year move alone added 16,500 dollars a month, and that is before any surcharge reset that has not landed yet.

Now the bigger one. Assume 150 tonnes a month, split 90 tonnes to Europe and 60 tonnes into the US interior. At 3.50 for Europe that is 315,000 dollars. At an assumed 3.90 into the interior, because that is the lane with the 15.1 percent weekly move behind it, the 60 tonnes is 234,000. Call it 549,000 dollars a month for the programme. If Chicago repeats anything like this week, another 15 percent on that 3.90 is 0.59 a kilo, which is 35,100 dollars a month on those 60 tonnes alone.

That is the number I would put in front of a board, because it is the one that turns a freight line into a forecast problem. Fifty-four thousand a month is a rounding issue for a company that size. Six hundred thousand a year is not.

Now the rail substitution, with the same warning that the rate is mine. Suppose 30 of those 150 tonnes can move by rail, and suppose the all-in rail equivalent is around 1.10 dollars a kilo. Thirty thousand kilos at 3.50 is 105,000 by air and 33,000 by rail, so the monthly saving on that block is 72,000 dollars, and across a four month peak it is 288,000. The price you pay for it is transit. Air door to door I plan at five to seven days, rail at twenty to twenty-five, and those are my planning numbers, not published figures. If your stock cannot sit an extra two and a half weeks, the saving is theoretical and the air bill is real.

There is a version of this that is not about rates at all. If you know the peak is coming and your product has a predictable sell-through, the cheapest air cargo is the cargo you moved by rail in September and put in a warehouse. Warehouse rent is a known number per pallet per month. Air is a number that can move 15 percent in a week. Trading a known number for an unknown one is the whole job in a week like this one.

The traps are all at the handovers. A rail booking is not confirmed when the forwarder says it is; it is confirmed when the terminal gives you a loading date, and at peak those dates move. Ask for the loading date, not the booking. At the gauge change, damage is real and it is usually the cheap pallet that gets crushed, so build the pallet for the handling it will actually receive rather than the handling you specified.

On the truck side the trap is permits and drivers, in that order. A permit that is fine in October can be a queue in December, and a driver who knows the route is worth more than the truck. If you are going to lean on the road this quarter, name the driver, not just the carrier.

On the Caspian the trap is the weather and the ferry schedule, which are the same trap. A crossing that takes a day in August takes three in January and the schedule does not hold. If the consignee has a hard date, this corridor is a bet, and it should be priced as a bet rather than as a plan.

Counterfactual one. If crude falls back below 100 and jet eases, the surcharge resets come down, air softens, and every tonne you moved to rail at 1.10 looks like an overreaction. That is a real risk and I will not pretend otherwise. The mitigation is to commit rail for the peak window only, October to December, rather than for a year, so the mistake has an expiry date.

Counterfactual two. If Chicago normalises next week, and 15.1 percent weekly moves usually do not repeat, then the US interior premium collapses and the pre-positioned stock you paid warehouse rent on was insurance you did not need. That is the cost of buying protection in a spike, and it is smaller than the cost of not having it, which is why I still buy it.

Counterfactual three, and this is the one that would really change my advice. If the border at Brest clears and the rail terminals open extra peak capacity, then rail stops being a scarce slot and becomes a commodity, and the whole argument about booking the valve early falls apart. I do not expect that. I have never seen a fourth quarter where the yard was empty.

So what do you actually do this week. Rank the moves by what they save, the way you would rank anything. Fixing the gateway on the US lane is worth more than any negotiation, because a 15 percent weekly move is bigger than any discount a forwarder will give you. Moving tonnage to rail is worth 72,000 a month on 30 tonnes, if the lead time allows. Locking a block on the lane that spiked is worth whatever the next reset would have cost. Renegotiating the rate on a lane that printed flat is worth almost nothing, and people spend their week there because it feels like work.

The rule I would write on the wall is this. When one hub moves 15 percent and another moves minus 0.7 in the same week, the average is lying to you about both. Watch the two ends, not the middle, and buy the alternative corridor while it is still boring. That is not a forecast. It is what the map has been telling us every September for as long as I have been running these roads.

Close it out on the road I know. The box does not care about the index. It cares whether the truck was there at six in the morning, whether the yard had a crane free, whether the ferry sailed, whether the driver had the right permit. All of that is decided weeks before the number that everybody quotes on the news. The index tells you the weather. The corridor tells you whether you get there.

↑ Back to the key points

— By Ömer Kaya