China's magnet shipments to Japan ran 44 tonnes short of last August

Two months of customs data cannot separate a squeeze from a base effect, and the newer month is the milder one. August's 17.1 per cent annual fall in magnet volume ^[2] sits 35.1 points above July's decline ^[2]. One reading has July as the outlier and August as the drift back toward normal. The other has both months under last year, with all six of the product lines in the data down ^{[2][4][5][6][7]}.

In tonnes, August's gap is about 44 ^[1]. Hold that rate for a year and it is roughly 528 tonnes of permanent magnets that Japanese assemblers source elsewhere or do not build ^[3]; the flow that did arrive annualises to about 2,544 tonnes ^[6].

Unwrought tungsten shipments fell 94.7 per cent ^[7], which leaves about 5.3 per cent of last year's volume, near a nineteenth of it ^[4]. The customs data as reported gives a tonnage only for magnets, so there is no tungsten tonnage to set against that percentage ^[10]. Rare earth compounds fell 48 per cent against 11.2 per cent for metals ^[5][4], about four times as fast ^[5].

That spread is what argues against reading this as a single blanket restriction. A uniform political freeze would not produce 11.2 per cent on metals and 94.7 per cent on tungsten in the same month ^[4][7]; category-by-category handling would. The South China Morning Post links the extended decline to worsening Beijing-Tokyo relations since late last year, after Prime Minister Sanae Takaichi's remarks on Taiwan ^[8], and describes the cause as China's trade limits ^[9].

For a buyer, the planning figure is the worse month. A purchasing department that sizes safety stock for a 17 per cent shortfall and receives a 52 per cent one covers the difference from inventory or idles a line ^[2][3]. So the cash goes into stock. Magnet inventory crowds out the work of qualifying a non-Chinese supplier, and the inventory decision gets made monthly while a qualification programme takes quarters.

In my view the sensible assumption into next year is constrained and lumpy supply. What would break that: September magnet volume back near the 256 tonnes of a year earlier ^[1] and tungsten off its floor, which together would make July and August a two-month administrative gap. A third month under 212 tonnes with tungsten still near a twentieth of last year would settle it the other way ^[1][7].