How it works
Asset managers who want interest-rate exposure often prefer futures to actual bonds, because a future costs a margin deposit rather than the full price of the bond. That persistent demand pushes the futures price slightly above the price of the underlying cash Treasury. The difference is the basis.
A hedge fund takes the other side. It buys the cash bond, sells the future, and waits. At delivery the two prices must converge, because the contract is settled by handing over an actual bond. The fund keeps the gap. It is not a bet on yields rising or falling — both legs move together — it is a bet that two prices for nearly the same thing end up equal, which they nearly always do.
The problem is that the gap is a few basis points. As illustration, on $100 of bonds a two-basis-point basis is two cents. Nobody runs a fund on two cents. So the fund takes the bond it just bought to the and borrows against it, at close to the overnight rate, then buys more bonds with the proceeds and repeats. The lender protects itself with a haircut — again as illustration, a 2% haircut means the fund puts up $2 of its own money per $100 of bonds and borrows the other $98, which is fifty-to-one leverage. Two cents on $2 of equity is a perfectly respectable return, and it repeats every quarter.
So the trade has two dependencies, neither of which is the direction of interest rates. The first is that the cash and futures prices stay close. The second is that repo funding stays available at something like the overnight rate — SOFR was 3.62% at the time we wrote about the trade in August — and that the haircut does not move. Raise the haircut from 2% to 4% and the fund must halve the position, whatever it thinks about value.
There is a further wrinkle traders are currently hunting. Treasury futures can be settled with any of several eligible bonds, and a sharp move in yields changes which one is cheapest to deliver. That optionality sits with the short — the basis trader — and it is worth something when yields move a lot.
Why it matters to this crash
The basis trade is how a absorbing close to $2tn of annual deficits keeps clearing. The marginal buyer of US is increasingly not a pension fund or a foreign central bank but a relative-value hedge fund financed overnight. That is a different kind of holder. A pension fund holds through a drawdown. A fifty-times-levered fund holds until its financing costs change, and then it sells.
This matters more now because the long end is already strained. We covered the 30-year Treasury reaching 5.33%, its highest since 2007, alongside every other developed long end at multi-decade highs, and the AI issuance wave in which the biggest tech companies alone borrowed roughly $200bn — about 25% of the Treasury's net note and bond issuance to private investors, five times their 2025 share. Someone has to buy all of that duration. Increasingly the answer is leverage.
We wrote up the size on the same day: roughly $830bn as of September 2025, about twice what it reached in early 2020, when the trade unwound over about a week and bought $1tn of Treasuries in a fortnight to stop it. Two honest caveats. That figure reaches us as a secondary summary of Fed research rather than a Fed press release. And the growth of has genuinely made the funding cheaper and more stable than it was in 2020. The trade is also useful: it is a large part of why cash and futures prices track each other at all.
What unsettles us is how invisible it is. When USC economists surveyed bond investors, voters and finance graduates about what actually supports the Treasury market, nobody named it. All three groups put the odds of a US debt crisis within ten years near 50%, and none of them mentioned the mechanism most likely to be the transmission channel.
What would make this dangerous
Watch funding first, because funding is what breaks. Repo rates printing persistently well above the Fed's administered rates, rather than in a single quarter-end spike, would mean dealer balance sheets are full and the marginal lender is scarce. Rising on Treasury repo, or a futures exchange raising initial margin after a volatile session, would force position reduction on a schedule the funds do not control.
Then watch the basis itself. Convergence is the whole trade; a basis that widens on a day when yields are moving sharply is the signature of forced sellers, not of value. Add auctions that tail badly, and you have the 2020 sequence.
The compounding risk is who else is standing there. If the non-bank firms that provide liquidity in a stress are simultaneously nursing their own losses — 's $15bn July is the case we have covered — then the buyer of the bonds a deleveraging basis fund is dumping may not show up. Nobody capitalises that layer, and nobody sees it until the month it matters.