The Basis Trade Has Shrunk 28% This Year—and the 30-Year Auction Just Discovered What Demand Looks Like Without the Carry Crowd

Between the January 2026 peak and the September reading, Morgan Stanley's basis-trade tracker fell from $1.26 trillion to about $900 billion—a roughly 28% drawdown in nine months. The Federal Reserve's June 2026 decomposition already showed the building blocks: at end-September 2025, the basis trade accounted for about $830 billion of hedge fund long-Treasury exposure, equivalent to 35% of the $2.4 trillion total, with the rest split across swap-spread arbitrage, maturity-matched trades, and steepener positions. Layered on top of that was $1.4 trillion of hedge fund short futures, per the Office of Financial Research—the mirror image that made the arbitrage work.

The most striking feature of the September cut is where it isn't. Morgan Stanley's data shows the decline is \"largely contained to the two- and five-year areas of the Treasury curve,\" while \"for the futures contract tied to Treasuries maturing in 25 to 30 years from now, basis positions have increased.\" The hedge fund crowd, in other words, has rotated into the long end while shrinking its overall book—buying more 30-year cash against30-year futures even as it pulls back the front-end carry. That rotation coincides with the worst part of the curve: 30-year yields climbed from 4.73% a year ago to 5.42% on September 24, with the 30-year–5-year steepness near 109 basis points, the widest since 2021.

The implication is uncomfortable: the marginal leveraged buyer at the front end is gone, but at the long end—the part of the curve where the Treasury now issues most of its net supply—the basis book is being expanded into a market where yields are already 5.4%. Hedge funds are not capitating on duration; they are concentrating it.

Morgan Stanley's Carter cites \"less-attractive returns… due to a smaller arbitrage opportunity\" and Citi's Williams blames \"dislocations and volatility [that] have fallen materially over the past few years.\" Both are right, but neither is the whole story. The arithmetic of the basis trade closed from three directions at once, and none of them look like a forced unwind.

The test of the new equilibrium came in the August and September 30-year auctions, and the prints tell two different stories. The August 13 auction cleared at5.216%—the highest yield since 2001—with a bid-to-cover of 2.39, near the low end of the trailing 12-month range of 2.29–2.66, and with primary dealers forced to absorb 11.5% of the takedown. The September 10 auction went better: 5.308% yield, bid-to-cover of 2.61, indirect bidders taking 79.5% of the award—comfortably above the recent average. Foreign central banks and pension funds picked up the slack the basis crowd left behind.

The September print looks reassuring, but it happened with 30-year yields already at 5.3%—the auction priced into a market that had already absorbed the term-premium shock. The August auction was the real signal: when the leveraged carry crowd wasn't there, the Treasury had to pay the highest real yield on a 30-year bond since the dot-com era, and dealers had to absorb nearly one-eighth of the deal.

The basis trade is also a prime brokerage business. Hedge funds running it pay for repo financing, futures clearing, securities lending, and balance sheet—all revenue lines on Wall Street. Morgan Stanley, Goldman Sachs, JPMorgan, and Citigroup all run the largest prime brokerage franchises, and the Q2 2026 numbers showed what the trade had been doing for the top of the franchise: Morgan Stanley booked $21.35 billion in total revenue with equities trading up 69% year-over-year to $6.3 billion and prime brokerage client balances at record levels; Goldman Sachs printed record quarterly revenue of $20.34 billion (+39% year-over-year) with FICC up 32% to $4.59 billion. The same quarter hedge funds went from $1.26 trillion to roughly $900 billion in basis trade notional.

With the basis crowd concentrated in the top 50 funds—which account for ~90% of gross Treasury exposure per the Fed—the prime brokerage revenue at stake is sitting at a small number of very large clients. If those clients shrink repo balances by even 10–15%, the Q4 revenue read-through to the largest prime brokers is real: average client balances drive financing revenue, financing revenue is the largest contributor to prime brokerage economics, and lower balances compound with smaller trade notional.

Three dates matter more than the rest. First, the October 30-year auction—the first test of the new equilibrium after the September 10 print, and the first auction where the Treasury's doubled buyback program ($4 billion per operation in the 20–30 year sector) is fully in effect as a backstop. Second, the October 28–29 FOMC meeting, where traders are now pricing a >75% probability of another rate hike versus roughly 49% a week ago, per CNBC; New York Fed President John Williams has already called another hike by year-end \"reasonable,\" and Fed Governor Michael Barr has signaled further policy adjustments. Third, the November 4 Quarterly Refunding, when Treasury will reset buyback sizes and signal whether the August increase is permanent or a temporary response to the August auction weakness.