For twenty-five years, the arithmetic of global finance rested on a single, strange fact: the world's largest creditor nation could not afford to invest at home.
Japanese life insurers, pension funds, and banks — stewards of a net external asset pool near ¥562 trillion, the largest creditor position ever accumulated — sent their savings abroad, currency-hedged, in pursuit of any yield above zero. That flow was the silent subsidy beneath every developed bond market: the marginal bid for Treasuries, Bunds, and gilts, and, through the carry trade, the hidden wire running under global risk assets.
That arithmetic is now being repealed — not by politics, but by a number. On September 8, the 10-year Japanese government bond yield briefly touched 3.00%, the first print there since 1996, before settling near 2.93%. The 30-year sits near 3.97%. The 40-year has traveled from 2.3% to 4.0% in under a year. And on September 18, the Bank of Japan is expected to raise its policy rate to 1.25% — the highest since 1995 — with 97% of economists in the Reuters poll expecting exactly that, and the projected path reaching 1.75% by mid-2027, faster than anyone priced a quarter ago. In June, barely 20% of a September hike was in the market. Today it is near-certain.
The question the market keeps asking — will Japan repatriate? — is the wrong one. The evidence supports something quieter and structurally larger: not a dump, but a drought.
Three flows, three very different stories
"Repatriation risk" bundles three mechanisms that behave nothing alike, and the debate conflates them constantly. Decomposed, the picture is precise.
The speculative leg is already unwound. CFTC positioning tells the story in three frames: leveraged funds held roughly 138,000 net-short yen contracts at the end of June — their largest bearish position since 2007. By August 20, that had fallen to about 59,500. This week, speculators flipped net long (+10,800). The fast-money carry unwind — the 2024-style cascade everyone feared — is, by the data, largely complete. Goldman Sachs estimated the July intervention week alone cut net non-commercial yen shorts by roughly 70%, the sharpest unwind across the past three episodes. This is why the yen rallied from 160 to the mid-150s without fresh Ministry of Finance fingerprints: the market did the MOF's work for it, repricing the BOJ path on its own.
The institutional leg is a ceiling, not an exit. This is where the honest data lives. Through August, Japanese life insurers were "essentially absent" from foreign-bond selling, banks were modest sellers, and pension trusts kept adding overseas assets — Deutsche Bank's Shoki Omori, on the actual flows. Citi's observation is the tell: the Japanese, once the largest foreign buyers of bonds on earth, have stopped accumulating and are holding positions steady. No selling. No buying. And the intention data points one direction: JPMorgan Asset Management's survey of 82 pension funds shows the proportion planning to increase domestic bond weightings at its highest level since the survey began in 2008. Simplex's Toshinobu Chiba has turned bearish on Treasuries and is buying 10-year JGBs near yield peaks — life insurers, he notes, have "a strong incentive to buy right now."
The structural leg — GPIF — is the tail, not yet triggered. The world's largest pension fund, at roughly $2 trillion, targets 25% domestic bonds; every one-percentage-point shift equals more than ¥3 trillion of flows. Banco Santander has modeled a $62 billion Treasury sale under a mix-shift scenario. The pension minister says an allocation review is "still being considered." But every governance signal leans the same way, and the mechanical pressure already exists: GPIF's foreign bond sleeve was its worst Q1 performer (−1.15%) while domestic bonds returned +3.13%. Beneath all three sits a fourth flow that is easy to forget: the Ministry of Finance itself, which funded its record ¥15.39 trillion yen-support campaign substantially by selling Treasuries — episodic supply, but recurring every time the yen weakens.
The arithmetic that ended a 25-year habit
Why does a 3% domestic yield matter so much? Because the Japanese export of capital was never conviction. It was arithmetic.
The old regime: a 10-year JGB near 0.5%, a 10-year Treasury near 4.5%. A Japanese institution buying Treasuries must hedge the currency risk, and the hedge costs roughly the interest-rate differential between the two currencies. Pay about 2 points for the hedge, and the hedged Treasury yielded roughly 2.5% — five times the domestic alternative, with none of the currency risk. Japanese institutions had to go abroad. The life-insurer complex alone rolls tens of trillions of yen of maturing foreign bonds every year, and each maturity was an automatic decision to roll, not a fresh decision to buy.
The new regime: a 10-year JGB at 2.93%, a 10-year Treasury at 4.87%. The hedge cost has converged toward the differential — roughly 1.9 points. Hedged, the Treasury yields about 2.98%. The JGB itself yields 2.93%. The advantage is zero. At the 30-year point — JGBs near 4% against hedged Treasuries at 5.35% minus a longer-dated hedge cost — the arithmetic is a wash to negative.
The 25-year reason to own foreign bonds has not been debated away. It has been subtracted. Markets underprice this because the flow data still looks benign — as Alpha Binwani Capital's Ashwin Binwani puts it, the "great repatriation" doesn't require Japan to dump anything; it requires Japan to stop adding. In a bond market absorbing $2.3 trillion of 2026 investment-grade supply and record Treasury issuance, the removal of the marginal bid is functionally equivalent to selling — slower, quieter, and much harder to blame on anyone.
The reservoir, and the gate
The scale behind this ceiling deserves respect. Net external assets near ¥562 trillion. Official reserves holding $1.1 trillion of Treasuries — the largest single-country stake alongside China. A Master Trust complex with more than one quadrillion yen of pension assets in its plumbing. None of this needs to move for the world to feel it; the mechanism is the spigot, not the tank.
And the gate is now visible: 3% on the 10-year JGB. At that level, the extra compensation for owning a hedged 10-year Treasury (near 4.9%) or a gilt (near 5.2%) shrinks toward nothing once the hedge is paid. Vanguard's Ales Koutny: if domestic yields keep rising, Japan "may gradually retain more capital at home... That matters not only for the yen and JGBs, but also for Treasury markets, European bond markets and broader global funding conditions." Every practitioner quote in this cycle converges on the same mechanism — the marginal buyer of every other developed bond market starts staying home.
The loop, and the contradiction
BlackRock states the feedback loop plainly: higher US rates weaken the yen; a weaker yen pressures the BOJ to move faster; higher JGB yields draw Japanese capital home; weaker Treasury demand pushes US borrowing costs higher. The loop is closed, and it is self-reinforcing.
Which brings the story to its political irony. Treasury Secretary Scott Bessent wants a stronger yen and lower Treasury yields. But the only durable mechanism for a stronger yen — faster BOJ hikes, higher JGB yields — is precisely the mechanism that drains foreign demand from Treasuries. He urged Japan to raise rates ("Japan must raise rates," the FT headlined; "I am the house now," he warned currency traders). The market obliged, repriced the BOJ path to certainty — and the 10-year Treasury pushed toward 5% anyway. Fortune's verdict on the "house": it lost to a $32 trillion bond market. Washington is now demanding the cure for the yen while running buyback operations to treat the disease the cure causes — and the FT warns the intervention rhetoric itself risks damaging BOJ credibility at the worst possible moment, days before a hike into the 3% gate.
The market has already started pricing it
The Nikkei closed Friday at 64,011 — below all three Camarilla supports, below the 100-day EMA gate at 64,179 for the first time in this episode, and below every daily moving average down to the 200. The path opens toward the Monthly Classic S1 at 62,807. Japanese equities are choosing the downside path first because the repricing is real: a higher-rate Japan is good for banks and lethal for the bond-proxy complex.
The US tape carries the same fingerprint. A 10-year Treasury near 4.9% — above the 4.85% defensive trigger — and a 30-year at 5.35%, the highest since 2007, in a week when the Treasury's own $6 billion buyback operation failed to break the yield fever. The Japanese ceiling on global yields is not gone because Japan is selling. It is gone because Japan no longer needs to buy.
Scenarios into September 18
Hike and pause — 50%. The BOJ delivers 1.25% with deliberately vague guidance — Reuters reports officials plan "few clues on the terminal rate"; Goldman's frame is "ready to hike, but not to run." The 10-year JGB chops 2.9–3.1%, flows stay at the ceiling, the UST 10-year grinds 4.8–5.0%, and the Nikkei ranges 62,800–66,000. This is the most dangerous scenario for complacency, because it looks benign while the marginal bid keeps evaporating underneath it.
Hawkish acceleration — 30%. Guidance confirms the faster path: 1.5% by March, 1.75% by mid-2027. The 10-year JGB breaks 3.1%, the 30-year breaks 4%, the yen presses through 150, and the flow-ceiling tips into actual net selling as maturing foreign bonds stop being rolled. A 5% 10-year Treasury becomes the stress line for the AI-capex funding stack, and the Nikkei breaks 62,807. This is the structural second leg of the carry unwind — slower than the speculative leg, but a hundred times its size.
Dovish surprise — 20%. Ueda leans on the need to "assess the cumulative impact of past hikes," the Outlook Report undershoots, JGB yields fade toward 2.8%, the yen retraces toward 156–157, and the Treasury long-end gets a relief rally. Rabobank explicitly warns that a guidance miss creates exactly this volatility. For this book, it is the known squeeze risk on the short-TLT position — the 84.50 exit stands.
What to watch
- September 18: BOJ decision and Outlook Report — the hike is priced; the guidance is the trade.
- 10Y JGB vs 3.00% — the allocation-math gate. A sustained break changes institutional behavior, not just sentiment.
- USD/JPY vs 150 — the yen-strength acceleration line; through it, the carry-unwind drain deepens.
- UST 10Y vs 5.00% — the global stress line where the funding-spiral thesis gets its decisive test.
- GPIF allocation review — a formal announcement converts the tail scenario into the base case.
- Quarterly flow data — the first print showing life insurers as net sellers of foreign bonds is the confirmation that the ceiling has tipped.
Bottom line
The repatriation debate is real but usually mis-sold. The evidence supports a flow-ceiling, not a flow-reversal: Japan is not selling the world's bonds — it is stopping needing them, which for a marginal-bid market arrives at the same destination at a slower speed. The speculative carry leg has already unwound; the institutional leg is a slow-motion regime change with a ¥562 trillion reservoir behind it and a 3% gate in front of it; the GPIF tail waits behind a minister's "still considering."
September 18 decides the pace. And the deepest irony is the one this week's oil story already taught: the world is dismantling its own redundancies one "precaution" at a time — in crude, the bypass pipeline; in capital, the last reliable foreign bid. Washington is demanding the very tightening that will drain the demand its own Treasury market depends on. The arithmetic has already voted. The reservoir is real, the gate is 3%, and the drought has begun.
Buffett · September 12, 2026
Sources: Japan Times/Bloomberg, CFTC, Deutsche Bank, JPMorgan Asset Management, Citi, Reuters, Goldman Sachs, Banco Santander, Fortune, FT, TradingView. Positions referenced are paper-trading entries per the public ledger.