On Tuesday, 1 September 2026, the yield on Japan’s ten-year government bond touched 3.00 percent for the first time since September 1996. The intraday high printed at 3.005 percent. It is the kind of number that gets written up as an anniversary — thirty years, one generation, a round level — and that therefore gets misread almost every time. The news is not in the number. The news is in what the number does to a piece of arithmetic that has pointed in one direction for three decades.
That arithmetic ran as follows. If you managed money in Japan — a life insurer with yen liabilities stretching thirty years out, a regional bank, a pension fund — you earned essentially nothing at home and had to leave the country to earn anything at all. That compulsion is one of the things that financed the long end of the world’s bond markets. Japan is still the largest single foreign holder of U.S. Treasury securities: $1.12 trillion as of June 2026, already down 3.3 percent year on year, and comfortably ahead of the United Kingdom at roughly $897 billion.
At 3.00 percent on ten years and 4.18 percent on thirty, that compulsion is gone. That is the story. Not the anniversary.
What actually traded on 1 September
This was not a single point on the curve doing something odd. It was the whole curve moving. The five-year yield reached 2.265 percent — not a multi-year high but an outright record, the highest level ever traded. The two-year hit roughly 1.80 percent, a 31-year peak. The twenty-year touched 3.885 percent, a level last seen in 1996. And the thirty-year headed for 4.18 percent, also a record close.
For context: on 18 August the ten-year sat at 2.945 percent and was already being reported as a thirty-year high. Two weeks later it had added another six basis points. Measured over two years, the yield has more than tripled. There are very few post-war examples in developed bond markets of a curve everybody assumed was permanently frozen repricing this fast.
Three forces hit at once. Crude oil jumped more than five percent in a single session to top $90 a barrel after U.S. strikes on Iranian targets, which lands immediately on the import bill of a country with no energy of its own. Expectations built that the Bank of Japan will move again at its 17–18 September meeting, after Governor Ueda said the board would decide with upside price risks in mind; the policy rate has been at 1.00 percent since June. And domestic reporting flagged a record set of ministry budget requests for the fiscal year that starts next April.
The arithmetic that brings the capital home
Here is the calculation this article is about. Picture a Japanese life insurer that has to service a yen liability thirty years out. It has two options.
Option one: the thirty-year Japanese government bond. Yield 4.18 percent. The currency of the asset matches the currency of the liability. No exchange-rate risk, no hedging cost, no basis risk, no rolls to manage.
Option two: the thirty-year U.S. Treasury. Yield 5.28 percent — 110 basis points more at first glance. But it has to be hedged back into yen, or the asset does not match the liability at all. The cost of that hedge is driven by the gap between the two currencies’ short rates. The U.S. target range has stood at 3.75 percent since 10 December 2025; the Japanese policy rate is 1.00 percent. That is 2.75 percentage points of hedging cost before the cross-currency basis, which has historically added a little more for yen-based buyers.
5.28 minus 2.75 leaves 2.53 percent. Against 4.18 percent at home. The domestic bond pays that insurer 165 basis points more than the American one — in the right currency, with no hedge contract and no rollover risk. The ten-year point tells the same story: 4.80 percent in Treasuries, minus 2.75, is 2.05 percent hedged, against 3.00 percent in JGBs. Roughly 95 basis points in favour of the bond issued down the street.
That sign flip is the event. For decades, the honest answer to why do Japanese institutions buy American bonds? was not yield optimisation. It was the absence of an alternative. With a ten-year yield around zero, there was simply no domestic instrument that could generate a contractual return. So the money went abroad and either carried the currency risk or paid to remove it. Both were uncomfortable, but there was no third option. Now there is.
A budget whose central assumption was stale on the day it was filed
The second strand is fiscal, and it is sharper still. Japan’s ministries have requested roughly ¥143 trillion for fiscal 2027, a record. Debt servicing alone — interest plus redemption — is put at ¥36,638.6 billion, so a little over ¥36.6 trillion. That is ¥5,362.8 billion more than was allocated in the initial fiscal 2026 budget: up about 17 percent, and the steepest increase in twenty years.
Set that against revenue. Tax receipts in the current fiscal 2026 general account are budgeted at roughly ¥84 trillion. The requested 2027 debt service therefore equals about 44 percent of tax revenue. In the current year, at ¥31.3 trillion, the ratio is 37 percent. Out of every hundred yen the Japanese state collects, more than forty-four will go to its own creditors before a single school, pension or destroyer is funded. Measured against the total request, debt service is the single largest line item at 25.6 percent.
And now the genuinely awkward part. Those ¥36.6 trillion rest on an assumed long-term interest rate of 3.8 percent, raised from 3.0 percent for the current year — and the 3.0 percent was itself the highest assumption in 29 years. On 1 September the twenty-year traded at 3.885 percent and the thirty-year at 4.18 percent. Both sit above the assumption underpinning a budget request that had only just been filed. The calculation was overtaken by the market on the day it was submitted.
The leverage behind this is brutally simple: 0.8 percentage points of assumption cost ¥5.36 trillion, which is about ¥670 billion per ten basis points. Outstanding central and local government bonds are projected at roughly ¥1,344 trillion by the end of fiscal 2026, close to twice GDP. One percentage point of higher average yield on that stock — once fully refinanced, which takes years — is a little over ¥13 trillion a year. More than a sixth of all tax revenue, for the interest differential alone.
Why the yen is not responding
The textbook reaction would be: rates up, currency up. It is not happening. The yen traded around 160.27 to the dollar on 2 September, at a level the authorities have been treating as critical for months. Over twelve months the currency has lost roughly eight percent. The joint Japan–U.S. intervention of late July has given back more than half of its effect.
The reason is that the market here is no longer trading the rate differential. It is trading fiscal credibility. In a normal cycle, rising yields are an argument for a currency. In a country with ¥1,344 trillion of debt and a debt-service line consuming 44 percent of tax revenue, they are simultaneously an argument against it — because they weaken the entity paying the interest. Ryutaro Kimura of BNP Asset Management put it on Tuesday as the bond market having, through the rise in yields, already sounded a warning against fiscal expansion. That is precisely the mechanism: the hike does not support the yen because it strains the budget, and the strained budget weighs on the yen.
Why the long end everywhere came along
On the same day Tokyo printed 3 percent, the thirty-year gilt yielded 5.89 percent — the highest since March 1998 — and the ten-year 5.2501 percent, the highest since June 2008. The thirty-year Bund rose above 3.84 percent, its highest since 2011, and the ten-year Bund above 3.25 percent. France’s ten-year OAT reached 4.215 percent, the highest since November 2008, putting it above the Italian BTP at 4.188 percent. In the United States the ten-year sat at 4.79 to 4.81 percent and the thirty-year at 5.28 percent.
Each of these has a local explanation — the euro area reported August inflation of 3.3 percent that morning with energy up 14.3 percent, Britain has its own fiscal problem, France its own risk premium. But it is worth seeing the common component. When the largest cross-border buyer of duration in the system acquires a domestic alternative that pays 165 basis points more on a hedged basis, some of the bid for foreign thirty-year paper simply stops showing up. Not all at once, but at the margin — and prices are made at the margin.
What it means for a portfolio
The most obvious consequence is the dullest one: the risk-free domestic rate is back, and not only in Japan. A thirty-year Treasury at 5.28 percent is a different instrument from a thirty-year Treasury at 1.5 percent, and it competes directly with everything that has been bought as a bond substitute since 2012.
The losers are the same on both sides of the Pacific: anything whose valuation depends on a low discount rate. Real estate investment trusts live on the gap between property yields and financing costs, and every point at the long end works directly against net asset value and against refinancing — Realty Income, Prologis and American Tower all carry that exposure explicitly. Regulated utilities such as NextEra face a version of the same problem, softened by the fact that allowed returns eventually reset upward. And a growth company whose value sits mostly in cash flows beyond 2035 loses roughly half its present value when the thirty-year discount rate moves from 1 percent to 3.8 percent.
The winners are net-interest-margin businesses. Japanese banks — Mitsubishi UFJ and Mizuho — rose on Tuesday, and the logic extends to JPMorgan and Bank of America, though the latter also carries one of the largest unrealised losses on held-to-maturity securities in the industry, which is the exact mirror image of the same move. For life insurers such as MetLife and Prudential Financial the picture cuts both ways: higher reinvestment yields are a long-run relief for anyone with guaranteed liabilities, but the existing bond book is sitting on substantial paper losses, and rotating out of it realises them.
The case against this thesis
Three objections deserve weight. First, the hedge ratio. Japanese life insurers are estimated to hedge only 50 to 60 percent of their foreign assets. For the unhedged remainder, the arithmetic above simply does not apply — and a yen at 160 makes American assets look good in yen terms, which delays repatriation. The capital flow home is a story measured in years, not weeks.
Second, the ¥143 trillion figure is a request from the ministries, not an enacted budget. The Ministry of Finance routinely cuts it hard. The ¥36.6 trillion of debt service, however, is the line least amenable to cutting — it falls out of the stock of debt and the level of rates, not out of political will.
Third, a large part of this rate move is an oil event. If crude falls back from above $90 toward the seventies because the Strait of Hormuz calms down, much of the inflation argument dissolves — in Tokyo as in London and Frankfurt. The structural component, a Japanese debt stock meeting normalised interest rates, would be untouched by that.
What to watch
The next three weeks are unusually crowded. The European Central Bank decides on 10 September; the Federal Reserve on 15–16 September, where futures markets have priced between 60 and 66 percent odds of a hike since Kevin Warsh’s Jackson Hole speech; and the Bank of Japan on 17–18 September. Three major central banks that could tighten in the same direction inside eight days is not something that has happened since the early 1980s.
A Japanese hike would, interestingly, cut both ways for the arithmetic described here. It raises the yen short rate and therefore lowers hedging costs for yen-based investors, which would make the hedged American yield look somewhat better again. At the same time it pushes the domestic curve up further and worsens the budget maths. What matters is the difference between the two effects, and that turns on whether the Federal Reserve moves on 16 September.
The uncomfortable lesson of this week is broader than Japan. For thirty years one of the most reliable constants in global finance was that Japanese capital had to go abroad, because it earned nothing at home. That constant is gone. Anyone drawing an investment conclusion from it should not start with Japanese bank shares but with the question of which positions in their own portfolio have quietly been betting that thirty-year discount rates stay near one percent. On the tax side nothing changes: U.S. investors still face ordinary income treatment on holdings under a year against long-term capital gains rates above it, plus the 3.8 percent net investment income tax at higher incomes, and the wash-sale rule still applies if losses are harvested on bond funds. Non-U.S. holders of American securities should keep a valid W-8BEN on file to claim treaty withholding rates on the coupons that are, at long last, worth collecting again.
Try TradingView Free for 30 Days
Plus get a $15 discount on your first subscription through this link.
