One Selloff, Four Countries: The Long Bond's Anchor Buyers Have Gone Missing

On Tuesday the 30-year US Treasury yield pushed above 5.3%, its highest since 2007, according to CNBC¹. The same day, Japan's 10-year yield touched a three-decade high just under 3%, Germany's Bund yields hit levels last seen in 2011, and Britain's 30-year borrowing costs neared the 1998-era peaks reached in May, Reuters reported². Oil climbed back above $90 a barrel, up roughly 50% this year, as hopes for a US-Iran deal faded. A recent US 30-year auction cleared at 5.216%, a 25-year high. Four sovereign bond markets with four different inflation rates, four different central banks, and four different fiscal stories are selling off in unison, and the question that matters is what single force can do that.

The candidates on offer are oil, artificial intelligence, and deficits. I think the honest answer is that these are not competing explanations but inputs to one mechanism: a historic glut of long-dated paper, public and private at once, is arriving in a market where the buyers who never asked about price have left, and the buyers who remain are demanding to be paid for risk they spent a decade holding for free. The oil shock is the accelerant, not the engine. And the scariest version of the story, a cascade of forced sellers, is real but conditional, with a visible tripwire rather than an inevitability.

Start with the meter that measures this: the term premium, the extra yield investors demand for locking money up in a long bond instead of rolling short-term bills. It can be estimated from models like the New York Fed's ACM decomposition, and its recent behavior is telling. When the Fed cut rates by a full percentage point starting in September 2024, the 10-year yield rose anyway, and the St. Louis Fed's own analysis³ showed the term premium, not expected policy rates, doing the work, climbing from near zero to its highest since 2011. By mid-August 2026 the ACM 10-year term premium stood around 0.8%⁴, a full regime change from the negative readings that prevailed for most of the 2014-2021 era. The bond market has not lost faith in central banks' rate paths. It has stopped subsidizing duration.

Why now? Look at what is being asked of it. Governments are borrowing heavily everywhere; the US July deficit was the largest monthly total since March 2021, per CNBC¹, swollen by tariff refunds after the Supreme Court struck down last year's emergency tariffs. But the genuinely new supply is corporate. Amazon, Alphabet, Meta, and Oracle issued about $194 billion of bonds this year through July 7, up 79% from all of 2025, and Goldman Sachs expects the five big hyperscalers to issue roughly $250 billion in 2026 and $400 billion in 2027 to fund about $750 billion of AI capital spending, Reuters found⁵. The Dallas Fed calculates⁶ that AI-related investment-grade issuance, plus the interest-rate swaps data-center borrowers use to fix their floating loans, adds duration supply equal to roughly an eighth of the Treasury's own. And demand is visibly straining: cover ratios on hyperscaler deals fell from nearly five times in February to below two in July, while the extra yield needed to place new deals jumped from about 2 to 12 basis points. Whether the borrower is the US Treasury or Alphabet, it is the same pool of long-term savings being tapped, at the same moment central banks that once absorbed trillions of this paper are shrinking their holdings instead.

There is a serious rival reading, and it deserves to be stated plainly: this is reflation, not indigestion. Higher yields, on this view, are the rational price of a hotter, more capital-hungry world economy. The Bank of Japan is set to hike as soon as September⁹ and weighing a faster pace, with surveys showing household and corporate inflation expectations near or above 2% and wholesale inflation at three-year highs. The 30-year TIPS yield, which strips out inflation compensation, hit 3.09%, its highest since 2008, per Reuters⁷, suggesting a genuine real repricing. And with the Strait of Hormuz still effectively closed, some 8.3 million barrels a day of Gulf output shut in, and global supply forecast to fall 4.3 mb/d this year according to the IEA's August report⁸, inflation risk is not paranoia.

Each piece is true. The synthesis fails on Japan, which is the cleanest test case. Japan's actual core inflation printed 1.6% in July, below the BOJ's target, as CNBC noted¹⁰, even as its long yields hit thirty-year highs. A pure reflation story cannot explain record borrowing costs in an economy whose measured inflation is under target and whose central bank has hiked to all of 1%. What explains it is that the BOJ, the buyer that absorbed half the JGB market, is tapering its purchases into rising issuance under a spending-minded government, and private buyers want compensation for uncertainty. Rising real yields cut the same way: a real yield is exactly where you would expect a duration glut to show up, because investors demand real compensation for absorbing real paper. The oil shock matters mainly through variance rather than level. Nobody can price 2036 inflation with Hormuz closed and spare capacity thin, so the premium for bearing that uncertainty over thirty years goes up, which is a term-premium effect wearing an energy costume.

That leaves the question of who breaks if this keeps going. The popular answer is Japan's life insurers and Britain's pension funds, and the balance-sheet damage is genuine: unrealized losses on domestic bonds at Japan's major life insurers reached ¥30.86 trillion, about $194 billion, by end-June, up 60% in a year and now exceeding their unrealized equity gains, Nikkei reported¹¹, with Nippon Life booking ¥44 billion of impairments last quarter. Yet the same insurers were net buyers of ¥630.5 billion of super-long JGBs in June, the most since July 2023, per Bloomberg¹². Wounded institutions buying at 4% coupons are stabilizers, not forced sellers. The fragility is a threshold, not a trend: BNP Paribas's Ryutaro Kimura has warned that a 30-year JGB yield above roughly 4.5% would put insurers at significant risk of impairment-driven sales¹³, because accounting rules force realized losses once bonds fall 50% below cost. The UK's 2022 crisis shows how nonlinear that can be; Bank of England research¹⁴ found forced sales by liability-driven investment funds, the leveraged structures pension schemes use to hedge their obligations, accounted for about half the gilt price collapse. Below the tripwire, the vulnerable players are the leveraged funds still positioned for rate cuts that keep not coming. Above it, the selling becomes mechanical.

So the verdict: this is one market event, a repricing of duration itself, driven by the largest combined public-and-private long-bond supply wave of the modern era meeting a buyer base that no longer includes central banks at any price. To read it instead as healthy reflation, you would have to believe that Japan's sub-target inflation justifies thirty-year-high yields, that record auction tails and collapsing cover ratios signal confidence, and that it is coincidence that the term premium began climbing the moment central banks stopped buying. I find that a much harder set of beliefs to hold than mine. The 30-year JGB sits a little above 4%; the distance between there and 4.5% is now the most important half-point in global finance.