In one session it erased ¥113 trillion from the Tokyo market and knocked bitcoin from $64,000 to $49,000. Credible estimates of its size range from $500 billion to $4 trillion. That is not a rounding error. That is a factor of eight, on the largest source of hidden leverage in the financial system.
The yen carry trade is borrowing Japanese yen at a low interest rate, converting it into another currency, and buying assets that pay more. As of August 2026 the Bank of Japan's policy rate is 1.0% while the US federal funds rate is 3.50% to 3.75%, so a trader collects roughly 2.6 percentage points a year for doing nothing but holding the position. The risk is not the assets, it is the exchange rate: because the loan must eventually be repaid in yen, a rising yen turns a profitable trade into a loss, and at 2.6 points of annual carry it takes only a 2.6% move to erase a full year of it. Section 02 lets you compute that breakeven yourself, section 04 shows why the trade's total size cannot be measured, and section 05 replays what happened when it broke in August 2024.
If two countries charge different prices for the same thing, what stops you from buying in one and selling in the other?
Money has a price, and the price is the interest rate. Japan has kept that price near zero for three decades. The United States has not. So there is a trade sitting in plain sight: borrow where money is cheap, park it where money is expensive, and pocket the gap.
That gap is called the carry. It is the oldest strategy in currency markets and it requires no forecast, no insight, and no view on anything. You are simply paid for holding the position open.
The carry is just the rate difference: 2.625% a year on the money you put to work. Nothing is produced, nothing is predicted. Reset the sliders to the real August 2026 numbers with the button below.
2.6% is not exciting. Nobody builds a $4 trillion position for 2.6%. To understand why this trade matters, you have to add the thing that makes every financial crisis interesting.
If the profit is small and certain, what is the fastest way to make it large?
Borrow more. A trader who puts up $100,000 of their own money and controls $2.5 million of position turns 2.6% into 65%. Japan's financial regulator caps retail currency leverage at 25 times for exactly this reason, and offshore brokers offer far more.
But leverage is symmetric, and here is the part most explanations skip. The exchange rate move that erases a year of profit does not depend on your leverage at all. It is fixed by the rate gap alone. Leverage decides something else: how much of a move you survive.
—
Notice what happens as you drag the leverage slider. The breakeven move never budges: it is always about 2.6%, because it is set by the rate gap, not by borrowing. The wipeout move collapses toward nothing.
In July 2024 the yen moved 12.4% in three weeks. Set the leverage above 8 and read the last line.
Why do currency moves that would be ordinary in any other market turn into a stampede here?
Because the exit is the same door as the position. To close a yen-funded trade you must sell whatever you bought and buy back yen. Buying yen pushes the yen up. Pushing the yen up damages everyone else still in the trade, which forces them to close, which means buying yen.
This is the loop. It does not need panic or bad faith. It is arithmetic plus a margin clerk.
Step 4 is the one that matters for everyone else. The assets sold do not have to have anything to do with Japan. They only have to be liquid. That is why an interest rate decision in Tokyo shows up in the price of a US technology stock, a Mexican peso position, and bitcoin, within the same hour.
Which raises the obvious question. If this thing can do that, how large is it?
How do you measure a strategy that nobody has to register, report, or even name?
There is no carry trade registry. The trade is not an instrument, it is a description of a motive, and the same position can be built in at least five ways that different institutions count differently, or do not count at all.
The Bank for International Settlements, which exists to measure exactly this kind of thing, could directly observe about $250 billion of yen loans to non-banks outside Japan as of March 2024, rising to roughly $500 billion once offshore cross-border yen debt instruments are added. In the same weeks, press coverage of the unwind cited estimates as high as $4 trillion.
Both numbers are defensible. The difference between them is not data. It is a set of definitional choices, and you can make them yourself.
Flip three checkboxes and the answer moves from $500 billion to nearly $3 trillion. Push every component to the high end of its published range and it clears $4.8 trillion. Every one of those checkboxes is a genuine, arguable position held by real economists. Is a Japanese pension fund holding unhedged US bonds running a carry trade, or is it a long-term investor with a currency exposure it has thought about? The trade behaves the same way in a crisis either way.
One number is worth holding onto, because it is the one that actually is measured. The purely speculative leg, the futures positions visible in weekly CFTC data, peaked at roughly 180,000 contracts of net short yen before August 2024. At ¥12.5 million per contract that is on the order of $15 billion. The visible part of this iceberg is trivially small compared to every headline about it.
What does an unmeasurable trade look like when it closes all at once?
On July 31, 2024 the Bank of Japan raised its policy rate from a range of zero to 0.1%, up to 0.25%. In absolute terms this is one of the smallest interest rate changes a major central bank has ever made. On August 2, US payrolls came in at 114,000 against an expected 175,000, and the market decided the Federal Reserve would be cutting sooner than it had thought.
The rate gap that funded the trade was narrowing from both ends at once. The yen, which had traded past 161 to the dollar in mid-July, was near 142 by the close of the Asian session on August 5.
Below is that FX path, with the documented daily reference levels. Choose a leverage and watch a yen-funded position work through the week in real time. The calculation is done in your browser from the rates themselves.
Anything above roughly 7 times leverage was mathematically finished before the week ended, and Japan's retail cap is 25. The BIS later documented that Japanese retail margin traders had been overwhelmingly short the yen going into it.
Then the loop from section 03 did its work. Bitcoin opened August 2024 near $64,000 and wicked as low as $49,000. Ether fell from about $3,200 to $2,100. On the Tokyo exchange bitFlyer, bitcoin's yen-denominated price fell nearly 15% against 11% on Western venues.
Japan has now raised rates four times. Why is the trade still here?
Every account of August 2024 ended with the same forecast: normalization in Japan would close the rate gap and retire the carry trade for good. Japan normalized. The Bank of Japan raised its policy rate to 1.0% in June 2026, the highest since September 1995, and held it there on July 31, 2026 by an 8-1 vote, with one member arguing for 1.25%.
The yen is weaker now than it was during the 2024 crisis. In late July 2026 USD/JPY approached 164, the weakest level since 1986. On July 31, 2026 the United States joined Japan in coordinated currency intervention, the first joint action of its kind since 1998, and the pair snapped back to 156.5 in a session.
| Condition | Aug 2024 | Aug 2026 | Status |
|---|---|---|---|
| BOJ policy rate | 0.25% | 1.00% | tighter |
| Fed funds rate | 5.25–5.50% | 3.50–3.75% | lower |
| Rate gap | ≈5.1 pts | ≈2.6 pts | halved |
| USD/JPY | 161 → 142 | ≈158 | weaker yen |
| Official intervention | Japan alone | Japan + US | escalated |
The gap has halved and the trade has not gone away, because 2.6 points levered 20 times is still 52% a year. The carry trade does not need a large gap. It needs a gap and a currency that does not move, and Japan spent 2026 supplying both, right up until the moment its central bank and the US Treasury had to intervene together to stop the yen falling further.
Goldman Sachs moved its twelve-month USD/JPY forecast to 165 from 155 on July 6, 2026. That is a forecast for the funding currency of the world's largest leveraged trade to keep weakening, which is another way of saying the position keeps paying until it does not.
Bitcoin fell 23% in the same 48 hours. Does that prove the yen did it?
This is where most coverage overreaches, so here is the careful version. There are two claims and only one of them is well supported.
The supported claim is mechanical. BIS Bulletin No. 90 attributes crypto losses of up to 20% during the August 2024 episode to margin pressure: a volatility spike tightened collateral requirements, and traders sold what they could sell. Bitcoin is liquid, it trades on weekends, and it sits in the same brokerage accounts as everything else. It was not sold because of a view about bitcoin. It was sold because it was sellable.
You can compute that yourself. This is the same margin call from section 02, viewed from the perspective of the asset that had nothing to do with it.
—
The unsupported claim is directional. The popular version says a weak yen funds bitcoin rallies and a strong yen kills them, so watch USD/JPY to trade bitcoin. The data has stopped cooperating. By June 30, 2026, bitcoin's 52-week correlation with USD/JPY had reached about -0.90, meaning bitcoin has been falling when the yen falls, the exact opposite of what the funding story predicts. The likelier driver is broad dollar strength affecting both at once.
So: the carry trade is a real transmission channel for forced selling during a deleveraging event, and a poor predictor of bitcoin's price the rest of the time. Both things are true. Anyone selling you a USD/JPY signal for bitcoin is working from the 2024 episode and ignoring the two years since.
Each one corrects something commonly written about this trade.
Ask what is being counted before you react to the number. $500 billion and $4 trillion are both honest answers to different questions. Any article that quotes one without saying which choices produced it is repeating a number it has not examined.
The trade is paid by the differential and killed by the volatility. A yen at 158 tells you nothing on its own. A gap of 2.6 points with a quiet currency is a live carry trade. The same gap with a currency moving 2% a week is not.
The move that erases a year of carry is fixed at roughly the size of the rate gap. What borrowing changes is how much of a move you live through, and that number falls fast. At 25 times, a 3.8% currency move ends the position regardless of how right the underlying idea was.
What gets sold is what can be sold. That is why bitcoin fell 23% in 48 hours in August 2024 while nothing about bitcoin changed. It is also why holding an asset outright, rather than as collateral in a leveraged account, is a materially different position during exactly these weeks.
The yen carry trade is borrowing Japanese yen at a low interest rate, converting it into another currency, and buying assets that pay more. As of August 2026 the Bank of Japan's policy rate is 1.0% while the US federal funds rate is 3.50% to 3.75%, so the gap is roughly 2.6 percentage points a year. The trader keeps the difference for as long as the exchange rate does not move against them.
Nobody knows, and the disagreement is enormous. The Bank for International Settlements could directly measure about $250 billion of yen loans to non-banks outside Japan as of March 2024, rising to roughly $500 billion once offshore cross-border yen debt instruments are included. Estimates cited in the press after the August 2024 unwind ran as high as $4 trillion. The gap is definitional, not statistical.
The Bank of Japan raised its policy rate to 0.25% on July 31, 2024, and a weak US jobs report followed on August 2. The yen rose sharply and leveraged positions were forced to close. On August 5, 2024 the Nikkei 225 fell 12.4%, its worst session since 1987, erasing about ¥113 trillion of market value. The S&P 500 fell 3% and the VIX spiked above 65.
It has, but the link is weaker than commonly claimed. Bitcoin fell from about $64,000 to roughly $49,000 within 48 hours of August 5, 2024, and BIS Bulletin No. 90 attributes crypto losses of up to 20% during that episode to margin calls forcing traders to close positions in unrelated assets. But bitcoin's 52-week correlation with USD/JPY reached about -0.90 by June 2026, which points at broad dollar strength rather than yen funding as the dominant driver.
No. The Bank of Japan raised rates to 1.0% in June 2026, the highest since September 1995, and held there on July 31, 2026 by an 8-1 vote. The yen is nevertheless weaker than it was in 2024, trading near 158 per dollar in August 2026 after touching almost 164 in late July, its weakest since 1986. A rate gap of roughly 2.6 points still pays traders to borrow yen.