That means that the government needs other investors to buy the securities that hedge funds no longer want to finance, potentially at higher yields.
The research lays out a different problem from the familiar story of leveraged investors dumping bonds in a crisis.
One involves managers deciding that borrowed money earns more elsewhere, while the other involves lenders or clearing arrangements demanding cash before a fund can obtain it. Both can affect Treasury prices and dollar funding, but only the second necessarily involves a scramble for liquidity that could spill into Bitcoin and other assets.
Hedge-fund balance sheets show expanding gross assets and broker financing in the second quarter of 2026, another reason not to treat a single Treasury strategy as a complete measure of funds’ borrowing.
Using the Federal Reserve’s monthly Form PF-based series, the estimated cash-futures basis position rose from $213 billion in January 2023 to $832 billion in September 2025, while the swap-spread position went from $42 billion to $305 billion.
Those strategies accounted for roughly $1.14 trillion, or 47% of the $2.4 trillion in total long Treasury exposure reported for large hedge funds that September. Even though those positions are estimates, their relative sizes show what happens when markets become difficult to finance.
How a small trade became a huge Treasury buyer
Asset managers who want exposure to government bonds can buy them outright or use Treasury futures, which allows them to obtain similar interest-rate exposure without actually holding the securities themselves.
When those futures become expensive relative to the bonds that can be delivered against them, hedge funds can buy the cheaper securities, sell futures, and try to collect the difference.
That cash-futures basis trade is designed to offset much of the exposure to interest-rate movements rather than express a view on whether US debt is worth owning.
Funds typically finance the cash bonds with repo, an arrangement that resembles a collateralized loan, and use their own capital to cover the portion a lender won’t advance, futures margin, and the other costs of keeping both legs open.
Financing a hypothetical $100 million Treasury position with a 2% repo haircut would require $2 million toward the bond purchase, excluding futures margin and other expenses.
If that haircut rose to 4% without any movement in the bonds’ value, the lender would advance $96 million instead of $98 million, leaving the fund to supply another $2 million, pledge other collateral, or sell some bonds.
Sources: OFR and the Federal Reserve’s June 2026 decomposition. These numbers cover different measures and observation dates, so you can’t add them together or treat them as four separate piles of bonds.
Selling Treasury futures is also not equivalent to betting that American bonds will fall. In a basis position, the short futures contract is paired with a long position in the securities, and the fund is trying to profit from the relationship between their prices.
That relationship can still turn against the trader, while losses on the futures leg can require immediate margin payments before gains on the bonds become usable cash.
The funds’ substitute trade uses less debt for the same risk
The Dallas Fed identified a competing strategy that also starts with a Treasury purchase but pairs it with an interest-rate swap, in which the fund pays a fixed rate and receives a floating one.
This swap-spread widener can earn money when swap rates rise relative to Treasury yields, and its recent risk-adjusted return has become comparable to the classic basis trade.
Swap-spread trades can expose a fund to more dollar risk per $1 of notional position, which means a manager allocating the same amount of risk to the new strategy needs fewer bonds to support it.
Investors can keep running relative-value trades while borrowing less against Treasuries and supplying less cash demand to the government-bond market.
Take a hypothetical manager willing to tolerate a $10 million loss under a chosen stress scenario. If the basis position would lose 1% of its bond notional in that scenario, the risk allowance supports $1 billion in Treasuries.
If the replacement swap-spread position would lose 2%, the identical allowance supports only $500 million. Those percentages explain how a pretty ordinary portfolio decision can remove hundreds of millions in bond demand without a fund closing its doors.
The original basis trade has also become less appealing to some of the asset managers buying futures. A Treasury futures seller can choose among eligible bonds to deliver, and that option has value when the cheapest security to deliver can vary substantially.
The Dallas Fed researchers find that higher yields have made the eligible bonds more similar for certain long-maturity contracts, reducing the value of that flexibility and making the futures position a less attractive way for some investors to obtain exposure.
Their paper offers another possible contributor that reaches well beyond Treasury trading. Investors seeking long-duration exposure may be getting fixed interest payments through swaps tied to AI data center financing, rather than buying Treasury futures.
The researchers describe that route as ‘plausible’ but not yet quantifiable from the available evidence, so it would be premature to calculate how much Treasury demand AI financing has displaced.
The Fed’s earlier position estimates also show that the alternative isn’t immune to its own funding problems: about $60 billion of swap-spread arbitrage unwound during the April 2025 tariff episode before positions recovered in subsequent months.
Substituting one trade for another can reduce the size of the Treasury position, but it doesn’t remove dependence on leverage, dealers, and counterparties when markets become hard to finance.
The monthly figures also make it hard to treat every hedge-fund Treasury position as one enormous trade that must unwind together.
In April 2025, as the tariff announcements disrupted bond markets, the Fed’s estimated swap-spread exposure dropped from $295 billion to $237 billion, a 20% contraction.
Estimated basis exposure moved in the opposite direction, increasing from $725 billion to $829 billion during the same month, before easing to $798 billion in May. These are month-end estimates and can miss violent intra-month moves, but they don’t support the idea that both strategies liquidated in parallel throughout that episode.
Original comparison of the Federal Reserve’s accessible monthly position data, in nominal dollars. Estimated exposures do not measure realized profit, trading flow, or daily forced sales.
Historical comparisons also make this look significantly less alarming. The basis estimate fell from $429 billion in February 2020 to $312 billion by the end of March, a 27% contraction during the pandemic-era Treasury seizure.
In contrast, the April 2025 month-end series shows a larger basis position alongside a shrinking swap trade, and by September 2025 the basis estimate had recovered to $832 billion, below its June peak of $894 billion.
A shrinking position in one period can reflect forced selling, a less attractive trade, or a later portfolio rotation, and monthly position data alone can’t show which.
Washington still gets buyers, but perhaps at a different yield
If hedge funds gradually reduce their holdings, pension funds, overseas buyers, banks and households can purchase the Treasuries instead, provided the price offers enough return to persuade them.
Bond prices and yields move in opposite directions, so a market that needs a more patient buyer can reach one through somewhat cheaper bonds and higher yields, rather than through a disorderly liquidation.
Stablecoin issuers, including Tether and Circle, also buy government securities to back dollar tokens, including Circle’s USDC, but their demand is concentrated in short-term Treasury bills and can’t automatically replace demand for longer maturities.
The problem starts when funds need dollars at the same time. Lenders demanding more collateral, futures positions generating a margin payment, or a repo contract that isn’t renewed can all force an investor to raise cash quickly, potentially by selling bonds into a market where dealers are reluctant to expand their inventories.
The IMF’s October discussion of hedge-fund leverage explains why the same firms that improve trading liquidity in normal conditions can amplify price moves when crowded positions must be reduced together.
That distinction also limits what can (responsibly) be said about Bitcoin. Funds under immediate cash demands will sell whichever liquid assets they can, while tighter dealer financing can affect other leveraged investors, but a decline in Bitcoin alongside a Treasury selloff wouldn’t establish that one caused the other.
Bank-reserve data and dollar funding encounter the same attribution problem: evidence would have to connect repo rates and terms, futures margin activity, Treasury market depth and selling during the same episode.
The Fed’s Oct. 27-28 meeting could alter expectations for the short-term rates that feed into repo financing, although private lenders can tighten terms even without a policy-rate move.
But in the much quieter scenario Dallas Fed researchers describe, the more revealing numbers may be hedge funds’ cash Treasury holdings, futures positioning, and swap exposure over successive reporting periods, especially if bond demand falls while funding markets continue operating normally.
