We read all 387 Treasury auctions of fiscal 2026: the primary dealers expected to bid at every sale were awarded 34% of the bills — and 11% of the bonds

Every U.S. Treasury auction leans on a standing set of bidders. The New York Federal Reserve names a small roster of broker-dealers its primary dealers — some bank-affiliated — and expects them to bid on a pro-rata basis in every sale at reasonably competitive prices. That is a participation commitment, not a purchase guarantee: dealers must show up and bid, but nothing obligates them to buy the securities or promises the auction clears. Traders watch how much of an auction those dealers are awarded — their takedown. Takedown measures how much supply the dealers absorbed at the sale, an imperfect proxy for end-investor demand, because dealers can resell what they win ^mmodeled. The convention runs that the more dealers absorb, the less appetite there was from everyone else. A low dealer takedown suggests end users cleared the shelf; a high one suggests the dealers took the balance. We pulled the complete record of what those dealers were awarded. Treasury's auction results, published security by security, cover 387 marketable auctions from the start of fiscal 2026 on Oct. 1, 2025 through the last sale on Aug. 20 — 292 bills, 71 notes and 24 bonds, $26.86 trillion in competitively awarded securities (retrieved Aug. 23). Across all of it, primary dealers were awarded 30.9% of the competitive award, indirect bidders 61.0%, and direct bidders 8.1% — the three categories partition the total exactly. The headline number hides a clean gradient. Split by tenor, the dealer award share collapses as the debt gets longer: dealers were awarded 33.8% of every bill dollar, 13.7% of notes and 10.7% of bonds. Fold notes and bonds together — the coupon-bearing debt that funds the government for years, not weeks — and end-user bidders (indirect plus direct) absorbed 86.7% of the $3.81 trillion sold; dealers took 13.4%. It is not a one-month wobble. The dealers' share of competitive awards held between 29.0% and 32.5% in every one of the 11 months. The extremes fall exactly where the gradient predicts: dealers were awarded just 2.49% of a 30-year bond on Feb. 19 and 2.10% of a 29-and-a-half-year inflation-protected bond on Aug. 20, but 71.6% of a 27-day bill on May 21 and 100% of a 2-day cash-management bill on Dec. 4.

Read the whole book of auctions and the dealer-award share sits almost entirely at the short end — the bills that roll over in weeks — and nearly vanishes at the long end, the 10-, 20- and 30-year maturities where worries about financing a growing debt actually concentrate. Take every auction of a security originally issued at a 10-, 20- or 30-year term, reopenings included: 40 sales, $962.8 billion. End-user bidders were awarded 89.6% of it — seven of every eight dollars. Fold in the shorter coupons and the whole notes-and-bonds book still leaves end users with 86.7%, near seven in eight. The part of the curve that gets the anxious headlines is the part dealers absorbed least this year. The convention that a heavy dealer takedown means weak demand is worth holding at arm's length ^sspeculative. Dealers routinely bid for bills as ordinary money-market flow, so a high bill takedown may be partly a structural feature of how the short end trades rather than a distress signal ^mmodeled. This describes what dealers were awarded at auction, not what they held afterward: an award is not balance-sheet inventory, and the takedown record cannot show whether dealers kept the bills or resold them. It is one reason the aggregate 30.9% is really the bill number wearing an all-auctions label. One label deserves care. An "indirect" bid is one routed through a dealer rather than placed directly, a category that includes foreign and international monetary authorities but also domestic funds. A high indirect share is commonly read as strong foreign or official demand, but the record labels the channel, not the buyer — it cannot, on its own, tell you a central bank in another country did the buying ^sspeculative.

The aggregate is dollar-weighted, and bills are most of the dollars — $23.05 trillion of the $26.86 trillion sold. So the all-auctions "30.9%" mostly describes bills; the honest read is by tenor, which is why the note and bond figures matter more than the blended one. Takedown share is also an allocation outcome, not a clean demand gauge. Two other tells point the same way but are not identical: bid-to-cover — Treasury's own ratio of total bids received to the amount accepted, averaged across each class and weighted by size — ran 2.91 times on bills, 2.58 on notes and 2.49 on bonds this year, so demand per dollar was thinner at the long end even as dealers were awarded less of it there. And because dealers are expected to bid pro-rata at every sale, their share reflects that standing participation as much as an appetite; a low dealer award number partly means outside bidders showed up in force ^mmodeled. Finally, the coupon figures rest on 95 auctions against 292 bills — a smaller sample, and the tenor gradient, not the exact percentages, is the durable finding.