In-depth market analysis from a professional trader with 25 years of institutional experience across FX, precious metals and macro — not a journalist, not an algorithm.
Through the second quarter of 1998 the yen weakened steadily against the dollar, pressured by pessimism about the Japanese economy down to eight-year lows. On 17 June, US and Japanese monetary authorities intervened together for the first time in years, buying yen to arrest the slide. The New York Fed alone spent $833m that day.
It steadied the tape. Briefly. Within weeks the underlying decline resumed, because intervention had done nothing to change the reason the yen was being sold in the first place.
"Intervention can buy you a calmer afternoon. It cannot buy you a change of trend when the trend has a fundamental reason to exist. 1998 proved that. So, arguably, has every intervention since."
The yen kept falling through the rest of the summer. What eventually stopped it had nothing to do with G7 policy coordination. It started with a debt default eight time zones away.
↑ Back to topLTCM was, at the time, the most sophisticated leveraged fund on the planet — run by two Nobel laureates and a former vice-chairman of the Fed, and trusted with capital by nearly every major bank on Wall Street. Part of its book relied on the yen carry trade: borrow cheap yen, invest the proceeds elsewhere, pocket the spread.
When Russia defaulted and devalued the rouble on 17 August 1998, LTCM's Russian exposure was supposed to be hedged. It wasn't, in practice — the banks it had hedged through were themselves retreating from risk. The result was a "flight to quality" so severe it created a liquidity crisis, and LTCM's book, including its dollar/yen trades, began losing value fast even after the Fed-orchestrated bank bailout that September.
"1998 didn't need a policy decision to break the carry trade. It needed a forced seller big enough that nobody could absorb the other side. That's what turns a trend into a crash."
As LTCM and funds like it were forced to unwind, the yen appreciated with a violence nobody had intervened hard enough to produce on their own. Over eight weeks USD/JPY fell more than 15%. On 7 October alone it moved 7% — still, more than 25 years later, the largest single-day move the pair has ever recorded. That is what an external catalyst looks like when it finally shows up.
↑ Back to topThe mechanics of this intervention are genuinely unusual, and worth being precise about. The US Treasury sold euros held in its Exchange Stabilization Fund to help finance the yen purchases — not dollars. Reported reasoning: selling dollars outright would have signalled Washington actively wanted a weaker currency, complicating both inflation expectations and an already stressed Treasury market. Using euros let the US support Japan without that signal, and without forcing Japan to sell its own US Treasury holdings to raise the cash — a sale that would itself have pushed US yields higher.
Worth being precise about the mechanics, because the obvious question is: why not just print unlimited dollars and sell them for yen? The US technically can. The reason it doesn't is straightforward. Manufacturing new dollars specifically to sell into the market is monetary expansion — functionally the same act as QE. QE is inflationary by design. With core inflation already running above target on both sides of the Atlantic, deliberately adding an inflationary policy to defend a currency is the opposite of what either central bank wants. Selling euros already sitting in the ESF's reserves avoids that entirely — it's a reallocation of existing assets, not new money creation.
Layer fiscal dominance on top and the caution compounds. Both governments are effectively hostage to their own bond markets. Japan's debt-to-GDP sits near 260%; the US carries its own structural deficit. Aggressive dollar-selling would still put downward pressure on the dollar and risk unsettling a Treasury market already sitting near multi-decade-high yields — exactly the outcome a government paying interest on trillions in debt cannot afford to invite. The same logic runs in reverse for Japan: it avoided funding this month's intervention by selling its own US Treasury holdings outright, precisely because doing so would have pushed US yields — and by extension the cost of financing its own debt-heavy world — higher still. Neither government can afford to be the one that tips an already-pumped yield curve further.
The FIMA repo facility itself is a genuinely pandemic-era tool — introduced by the Fed on 31 March 2020 to stop foreign central banks dumping Treasuries for cash during the COVID liquidity crunch, and made permanent in July 2021. It is capped at roughly $60bn per institution. Japan's $59bn draw this month sits right against that ceiling — which is precisely why Bessent is now asking for it to be "upsized." Repurposing a 2020 pandemic backstop to fight a 2026 currency battle is, as more than one commentator has now put it, a fairly clear tell of how thin the options have become.
"Fiscal dominance means the bond market runs the country, not the central bank. Both governments are avoiding the two tools that would actually move the needle fastest — printing dollars, selling Treasuries outright — because both tools point straight at their own already-strained yield curves."
It's also worth noting Japanese government bond yields jumped to a multi-year high at auction this week — a stronger yen and fading confidence that the BOJ can keep rates low pushed investors to demand more to lend. That is the fiscal-dominance trap in one data point: cheap Japanese borrowing has quietly funded a huge amount of global investment for years, and any unwind of that tightens conditions far beyond Tokyo.
↑ Back to topLeopold Aschenbrenner left OpenAI in 2024, aged 25 and with no prior trading track record, to launch Situational Awareness LP on the thesis that AI compute and power would be the trade of the decade. He was, directionally, right — the fund grew past $20bn. He ran it at roughly 4x leverage into a narrow book: SK Hynix, CoreWeave, Nebius, Micron, Bloom Energy.
When the AI infrastructure trade turned in July, those names fell 35 to 47% as the Philadelphia Semiconductor Index dropped 28.6% from its 22 June peak. At that leverage, the fund had no room to absorb the move. Goldman Sachs, JPMorgan and Bank of America, as prime brokers, issued margin calls. A capital raise, and talks with Millennium Management and Jane Street, both failed. Forced liquidation followed — the exact mechanical shape of an LTCM-style blow-up: leverage, concentration, a forced seller with no natural buyer.
"For about a week this had the shape of the shock that finally breaks the carry trade. Then Citadel showed up and bought the book. I want to be precise about what that does and doesn't tell us."
Here's what's actually confirmed: Citadel bought the bulk of Situational Awareness's public equity portfolio, and the Philadelphia Semiconductor Index rallied roughly 7.5% in the days that followed. That's one forced seller, dealt with. What isn't confirmed, and what I can't claim to know, is whether that single rescue was the whole story. The broader AI-infrastructure rout eased for reasons bigger than just one buyout — other capital came back into the sector, other positioning unwound quietly, without a single name attached to it. Aschenbrenner's fund is simply the one blow-up that became public. It tells us a forced seller existed. It doesn't tell us whether a forced seller big enough to do what LTCM did has shown up yet, or is still building somewhere we can't see.
My own theory, not anyone's confirmed statement: BOJ and Fed officials, watching an AI-driven deleveraging event unfold in real time, may well have wondered whether this was their externally-supplied off-ramp — the 2026 equivalent of Russia and LTCM. Whether it was, or wasn't, is genuinely an open question right now, not something either I or anyone outside those institutions can answer with confidence.
↑ Back to topThe yen carry trade has a genuine fundamental reason to exist right now, in a way that makes this feel less like a speculative squeeze and more like 1998's repeat. Japan's fiscal position — debt-to-GDP near 260%, a central bank that cannot hike aggressively without making its own government's borrowing costs unmanageable — gives the market a real, structural reason to sell yen and fund elsewhere. That is a very different animal to a trade running purely on momentum, and it is exactly the kind of trade that historically has not been stopped by verbal or even funded intervention alone.
What stopped it in 1998 was external and largely unrelated to Japan: a Russian default that forced a systemically important, highly leveraged fund to unwind its book in full. This month the market produced a candidate of its own — the AI-infrastructure deleveraging. One visible piece of that got dealt with by a single well-capitalised buyer inside days. Whether that was the entire shock, or just the part of it that happened to become public, isn't something I can say with confidence either way.
"Their words are about to be tested again. Not because anyone doubts they'll act — Japan and the US have shown this month they will. My view is that this becomes a repeated test, not a single event: the market keeps probing the line until either it holds cleanly, or something forces a real answer."
None of this is a call on where USD/JPY goes next, and I want to be clear about that. It's an observation about mechanism: this is the same structural fight as 1998, fought with a different toolkit — euros instead of dollars, a repurposed pandemic facility instead of open-market Treasury sales — and so far without confirmation that the one thing which actually settled it last time has fully arrived. Whether it has, in some form we haven't identified yet, or hasn't at all, is the question I expect the coming months to keep testing.
↑ Back to topGMD Trader spent 25 years as a professional trader at ABN-Amro, ANZ and JPMorgan, covering FX derivatives and precious metals markets across London, Amsterdam, Singapore, Melbourne and Asia. Retiring as Managing Director in 2016, the last decade has been spent managing a personal portfolio across equities, fixed income, funds and alternative assets.
This article represents the author's personal views and is published for informational purposes only. Nothing here constitutes financial advice or a recommendation to buy or sell any financial instrument. Global Macro Drivers is not authorised or regulated by the Financial Conduct Authority or any other financial regulatory authority. All trading involves risk, including the risk of loss beyond initial capital on leveraged instruments.
Every article is written by a professional trader with 25 years of institutional experience. No fluff, no filler — just direct market analysis from someone who has managed real money.
Published when the market warrants it, not on a schedule.
Subscribe — £29/month · Less than £1/dayDisclaimer: All content published on Global Macro Drivers is for informational and educational purposes only. Nothing on this website constitutes financial advice, investment advice, trading advice, or any other form of regulated financial guidance. Global Macro Drivers does not recommend the purchase or sale of any financial instrument, security, or asset. Past market commentary does not guarantee future results. Markets are volatile and all trading and investment activity carries risk, including the risk of total loss of capital. You should always conduct your own research and consult a qualified independent financial adviser before making any investment or trading decisions. Global Macro Drivers accepts no liability for any losses or damages arising from reliance on the content published on this site.