Borrow (or short) low-yielding yen to fund higher-yielding dollar assets. The engine is the US–Japan rate differential; while it stays wide, the trade is self-reinforcing — carry flows weaken the yen, which flatters the position. The risk is the unwind: a fast yen rally forces leveraged positions to close all at once (Aug 2024 was the dress rehearsal — memory reference, sized in days not months).
Macrobysunil’s point on why the 08/2026 interventions can’t kill it: intervention attacks the price, not the incentive. As long as the differential is unchanged, every intervention-driven yen rally is just a better re-entry level — so repeated intervention makes the trade bigger. The only real exits are Japan meaningfully hiking or the US side of the differential collapsing.
And Japan can’t hike cleanly — ekwufinance’s trilemma (https://x.com/ekwufinance/status/2086412621696094552): raise rates → blow up the carry trade and JGB-stuffed pension funds (Paper losses are real when you need cash); sell Treasuries to buy yen → blow up the US bond market; let the US print dollars to buy yen → import the inflation to the US. Every exit routes the pain somewhere systemic. (His conclusion — “we don’t own nearly enough hard assets” — is the Hard assets sales pitch; the trilemma stands without it.)
The rebuild is visible in the data, not just the theory: by 08/11 the dollar had “already wiped out the July 31st intervention” (Barchart, https://x.com/Barchart/status/2087100696843653187), and JPM’s positioning chart shows speculators net long ~$48bn USD — above +2 standard deviations, the most in over a decade (Barchart, https://x.com/Barchart/status/2087794940919087517). Two weeks after the joint defense, the crowd was record-long the pair’s carry side — while the “dollar-system meltdown” narrative peaked. Positioning and narrative on opposite sides; the money is in the carry.
The differential is closing from the Japanese side: the 2y JGB hit 1.75% on 08/31/2026, the highest in 31 years (Barchart, https://x.com/Barchart/status/2094245966895288713 — the chart shows the last comparable level in 1995, and negative as recently as 2021). Against a US 2y around 4.3% (memory) the gap is still ~250bp, so the carry lives — but the one exit this note names as real, “Japan meaningfully hiking,” is what a 31-year high in the 2y is pricing.