The New Buyer of America's Debt is Borrowing 90% of the Money

I keep thinking about my grandfather.

He bought U.S. Savings Bonds during the war. He held them until they matured. He thought of it as patriotic duty and as a safe place to put his money.

He didn't borrow money to buy them. They sat in a drawer for thirty years.

The Dallas Fed published a paper two weeks ago about who buys U.S. Treasuries today. None of the buyers look like my grandfather.

The paper asks a question that sounds simple. America issues debt. Someone has to buy it. Who?

The traditional buyers are stepping back. Foreigners now hold 32% of marketable Treasuries, down from nearly 50% in 2014. The Fed holds 13%, and it's still shrinking its balance sheet. Pension funds and insurance companies — the patient, long-term holders — have not grown nearly fast enough to absorb new issuance.

So who closed the gap?

Hedge funds. A small group running two strategies called the cash-futures basis trade and the swap spread trade. Both bet on tiny gaps between a Treasury bond and a derivative tied to it. The gaps are measured in basis points. To earn anything worthwhile from a gap that small, the funds borrow enormous amounts of money against the bonds.

The Dallas Fed put a number on it. Hedge fund long Treasury holdings rose from $600 billion in 2014 to $2.4 trillion at the end of 2025. Their net overnight borrowing reached roughly $1.8 trillion. That borrowing number more than doubled in two years.

In other words: the new marginal buyer of U.S. debt is borrowing most of the money to buy it.

The new marginal buyer of U.S. debt is borrowing most of the money to buy it.

I have no idea when the next shock arrives. But I know the shape of one. The overnight funding market tightens. Prime brokers demand more collateral. Hedge funds sell Treasuries to meet the calls. The selling moves yields. Higher yields widen the calls. More selling.

We've seen this before. March 2020. April 2025. Both times the Fed had to step in to break the loop. It worked. But the leverage in the system was smaller then than it is now.

It's the same setup, larger. That's what worries me.

The U.S. Treasury market is the deepest, most liquid market in the world. Everyone repeats that line. But beneath it sits an overnight loan market called repo, where Treasury bonds are pledged as collateral against borrowed cash. Trillions of dollars change hands there every business day.

Who lends that cash matters.

For years the assumption was that the big Wall Street dealers — Goldman, JP Morgan, Citi — were financing the trades. They weren't. A Fed paper from 2025 looked at dealer balance sheets and found that for every dollar a dealer lent in repo, eighty-five cents was matched by a corresponding borrowing on the other side. The dealers were passing the cash along, not providing it.

The actual cash providers are money market funds. They operate under strict rules: short average maturity, daily liquidity, constant redemption risk. When markets get nervous, money market funds get nervous first. They pull back from repo lending before they do anything else.

This is the chokepoint. If the funds step back, hedge funds have to sell collateral to cover the gap. In Treasury markets, the collateral is more Treasuries. Selling pressure lifts yields. Higher yields move more positions against the funds. More selling.

That's how a margin event becomes a market event.

And here's where it spreads. Across the Pacific.

Japan has been the largest single foreign holder of U.S. Treasuries for years. Japanese institutions love Treasuries. Their domestic bond yields were near zero for most of two decades, and lending money to the U.S. government paid better. Life insurers, pension funds, and banks became reliable, patient buyers of long-dated American debt.

That's changing.

The Bank of Japan ended its negative interest rate policy. Japanese government bond yields rose. The 30-year JGB now yields around 4%, the highest level since the bond was introduced in 1999. Suddenly, a Japanese insurer can buy domestic bonds at competitive yields without the currency risk that comes with owning dollar assets.

In April, Japan's major life insurers laid out their plans for the new fiscal year. None plan to increase their U.S. dollar bond holdings. Several plan to cut them. In the first quarter of 2026 alone, Japanese investors sold $29.6 billion of U.S. government, agency, and municipal debt.

Twenty-nine billion in one quarter isn't enough to break the Treasury market. But it's the direction that matters. The largest, most reliable foreign buyer is becoming a seller. The next-largest, China, has been a net seller for years.

Someone has to absorb what they're selling. Increasingly, that someone is borrowing the money to do it.

Meanwhile, at the Federal Reserve.

The Fed runs a backstop facility called the Standing Repo Facility, or SRF. If overnight funding rates in the private market jump above the Fed's policy target, banks and primary dealers can come straight to the Fed and borrow against Treasury collateral. It exists to put a ceiling on funding stress.

It existed for years and almost nobody used it. From launch in July 2021 through the end of 2024, daily borrowing was usually zero, occasionally a few hundred million dollars.

December 31, 2025: $74.6 billion.

That was a record. The Fed handled so much demand on the year-end day that it spent the next two weeks adjusting the rules. In December the FOMC removed the aggregate daily limit. Whatever eligible counterparties want to borrow against safe collateral, they can have.

Plumbing, not crisis. A central bank facility doing what it was designed to do, on a scale the designers didn't expect. The Fed has become a permanent counterparty to the Treasury market's overnight funding loop. That wasn't supposed to be the steady state.

Three more things worth keeping track of.

Tether disclosed $122 billion in U.S. Treasury bills at the end of 2025. That puts it among the twenty largest holders of T-bills in the world, ahead of many sovereign states. Five years ago Tether barely registered in this market. Standard Chartered projects a trillion dollars of fresh T-bill demand from the broader stablecoin sector by 2028. A new buyer for the same paper, all at very short maturities. The category shifts the front end of the curve in ways nobody has fully modeled.

Retail investors put $19 billion into CLO exchange-traded funds across 2025 and the first six weeks of 2026. Total assets in those vehicles crossed $35 billion. The funds buy slices of leveraged corporate loan pools — pools that hold debt from the same middle-market borrowers that private credit funds finance. Retail thinks it bought a safer way to play credit. It bought the same exposure in an ETF wrapper.

Germany's manufacturing PMI fell from 51.4 in April to 49.9 in May — the first contraction reading since the end of last year. New orders dropped. Input costs accelerated. Europe's industrial engine has been struggling to find traction for three years and the latest print says it's sputtering again. The Iran war has pushed shipping insurance and energy costs higher across the continent. We'll see.

In 1994 a group of bond traders left Salomon Brothers and started a hedge fund. They were the smartest people in the room. Two of them won the Nobel Prize in economics three years later.

They had an idea. Treasury bonds and Treasury futures sometimes drift apart by a few basis points. The gap always closes. If you bet on it closing, you make a small return. If you borrow enough money to make the same bet a thousand times at once, the small return becomes a large one.

The fund was called Long-Term Capital Management. In its first three years it returned 21%, 43%, and 41% after fees. Money poured in. By the start of 1998, LTCM had $4.8 billion in equity and $125 billion in assets. Twenty-five dollars of bonds for every dollar of capital. And derivative contracts on top of that with a notional value above a trillion dollars.

They worried they were getting too big. So in late 1997 they returned $2.7 billion to outside investors. The trades stayed on. The capital shrank. The leverage rose to nearly thirty to one.

On August 17, 1998, Russia defaulted on its ruble debt. LTCM wasn't directly exposed to Russia. But every spread in the world widened. The basis between Treasury bonds and futures, the spread between mortgages and Treasuries, the gap between on-the-run and off-the-run bonds — every gap LTCM had bet would close, opened.

Every gap LTCM had bet would close, opened.

By the end of August LTCM had lost $1.8 billion. The prime brokers asked for more collateral. There wasn't enough.

On September 23 the New York Fed locked fourteen Wall Street banks in a room and walked out with a $3.65 billion bailout. Nobody wanted to put the money in. They all knew that if LTCM had to sell its book at once, every one of them held similar positions on their own balance sheets. The prices would move against all of them simultaneously.

LTCM had $125 billion in assets at its peak.

Today the U.S. hedge fund industry collectively holds $2.4 trillion in Treasuries, funded mostly through overnight repo loans against that same collateral. The strategy is the same one LTCM ran. The funding plumbing is the same. The scale is roughly twenty times larger.

We'll see.