THE 4-WEEK BILL SPIKED TO 3.73%
THE LONG END CAUGHT UP
THREE WEEKS LATER.

The chart at the top of this analysis is the 4-Week Treasury Bill Secondary Market Rate — FRED series DTB4WK — from August 2025 through August 2026. It is the least glamorous line in the fixed-income complex. It is also, empirically, the line that leads every meaningful bond-market accident of the past decade.

The trajectory across the year is straightforward. In September 2025 the print sat near 4.22%, drifting lower through the autumn as the Fed's expected path softened. By late November it was near 3.90%. December 2025 brought a step-down to the 3.55–3.60% corridor, and there it stayed — a nearly featureless range — for the entirety of the first half of 2026. Seven months of flatline.

Then, on Thursday 23 July 2026, the print jumped to 3.73%. It retraced immediately. By the following Monday the series was back to 3.63%. By the end of August it was sitting at 3.66%. If you were watching only the level, the spike disappeared into the noise. If you were watching the shape, it did not.

The 4-week bill is not a policy instrument. It is a plumbing instrument. It settles overnight-to-monthly. It is bought overwhelmingly by money-market funds, government MMFs, corporate treasuries, foreign central banks, and dealer inventory bidding on behalf of the same. It is not held for yield. It is held for cash-equivalence. And because it is held for cash-equivalence, its clearing price is a direct read on how much cash the system has spare, and how much paper the system already has warehoused.

That is why it leads. When any part of the fixed-income complex is running out of room to intermediate, the bill print sees it before the coupon curve does. Coupon auctions have monthly rhythm, dealer distribution channels, and international allocator demand as absorbers. The bill market has none of that. It has today's cash bid against today's issuance calendar. When those two lines get too close together, the bill spikes.

• 2019 — the 4-week bill printed multiple >10bp intraday moves through August 2019, in a period when SOFR was still calm. The September 2019 repo blowup arrived four weeks later. In hindsight, the bill was the earliest tape print of dealer balance-sheet exhaustion.
• March 2020 — the 4-week and 3-month bills spiked violently on 12–16 March 2020 as MMFs faced redemptions and became forced sellers of exactly the paper the Treasury needed them to buy. The Fed's Money Market Mutual Fund Liquidity Facility was announced within seven days.
• October 2025 — the 4-week bill drifted upward by ~8–12bps across the last week of September 2025, into a benign-looking corridor. The 31 October 2025 SOFR spike to 4.22% and the $50.4 billion Standing Repo Facility injection followed one month later.
• 23 July 2026 — the print now under discussion. Fifteen basis points above baseline, retraced within one session, and largely uncommented on. The next twenty-one days: 12–13 August coupon auctions clearing at multi-decade high yields, SOFR-minus-IORB drifting wider, SRF utilisation rising.

The pattern is not "bill spike causes bond accident". The pattern is that the bill spike and the bond accident are the same event, priced first at the fastest-turning-over instrument on the curve. The bill spike is a tell because the bill is the last thing to bid before the private cash market steps back and the Fed's standing facilities have to.

The plumbing that makes the 4-week bill a leading indicator has three moving parts, and they are all pointing the same direction. Read them together.

• Money-market fund AUM — the total sits above $7.3 trillion in August 2026, an all-time high. Growth is slowing. Corporate cash buckets are cash-locked into T-bills at the top of the yield curve; retail flows have been steady but marginal. The pool of dollars competing for each new bill offering has essentially stopped growing.
• Overnight Reverse Repo (ONRRP) balance — drained from over $2 trillion in mid-2023 to ~$21 billion in the last week of August 2026. This is the buffer MMFs use when there isn't enough bill supply at an acceptable yield to park incremental cash. It is now empty. There is no buffer left to draw down.
• T-bill supply mix — the Treasury's TBAC guidance for FY2026 targets 15–20% of marketable debt in bills. Actual is now at 22.4%, above the target ceiling. Every incremental deficit dollar the Treasury issues short is a dollar bidding against an already-cash-locked MMF complex.

Read the 3.66% August print in that light and it stops looking like a normalisation. It looks like a system holding its breath.

There are two clean prior episodes where a quiet upward drift in the front-end bill was the earliest surface signal of a wider funding-market break. Both show the same three-part sequence: bill drift, weeks of quiet, then a spike at quarter-end that forces official intervention. The current setup rhymes closely with both.

In both prior episodes the leading tell was not repo. Repo stayed benign until the day it did not. The leading tell was the bill — a front-end instrument nobody watches on a daily basis, drifting quietly upward against a policy corridor that everyone was still calling calm. In both cases the drift was ~8–15bps sustained. In 2019 the drift lasted three weeks before the break. In 2025 the drift lasted four.

The current setup is week five of an unresolved drift. The July 23 print was the peak. The retracement is real but the level has not returned to the H1 baseline. The window where the calendar mechanically forces balance-sheet scarcity — 30 September 2026 quarter-end — is now inside the same event horizon as the current drift.

The next dated event that resolves this setup is 30 September 2026. It is the first quarter-end of the second half of 2026, and it lands on top of a T-bill share of debt that is already above the TBAC target ceiling, an ONRRP buffer that is functionally empty, and a coupon calendar that has been printing tails at every long-dated auction of the past six weeks.

Three things worth watching in ascending order of consequence:

• DTB4WK holding above 3.65% into quarter-end. That would confirm the July print was a leading edge, not a one-off. Bills are pricing this in real time; there is no lag between issuance and print. If the bill is still elevated on 30 September, the July spike was the tell.
• ONRRP balance flip to structurally zero — or Fed guidance revising the floor. A permanent zero in the reverse repo facility means the last conventional MMF shock absorber is gone, and the next incremental T-bill supply either bids up the front-end further or forces the SRF to do the work the ONRRP used to do in reverse.
• The 30 September / 1 October week itself. This is the calendar collision where GSIB reporting, coupon settle, and month-end all land inside 48 hours of each other. If the SOFR-IORB spread widens further in the run-up, or if the SRF sees a single-day pull >$15B, the 2025 pattern has repeated exactly one year later.

Note that none of these watchpoints is a rate story. Every one of them is a balance-sheet story. The Fed's dot-plot on 17 September is the wrong resolution instrument for this window — because the pressure is not on the policy rate, it is on the policy floor and the standing facilities. The relevant policy dial is QT, not FFR.

The 4-week bill spike does not sit in isolation on the tape. It sits inside a six-week window where four independently sourced events have all been reads on the same underlying variable — dealer + MMF balance-sheet capacity to warehouse sovereign paper.

Reading these five events as five stories is what most desks are doing. Reading them as one story is the analysis. Bills, coupons, currency, and standing-facility utilisation are all measures of the same underlying quantity: the marginal price of parking sovereign duration on a private-sector balance sheet. That price is going up at every point on the curve, in every tenor, at every settlement window. The July 23 bill print was the earliest visible surface of it.

The 3 August intervention design is the tell. When the reserve issuer and its most important sovereign creditor coordinate to keep sovereign-scale selling off the Treasury tape, it is because the tape is already too full to absorb it. The 4-week bill spike three weeks earlier said the same thing at a smaller scale — that the incremental dollar of front-end supply already has to bid harder for a private-sector home.

The uncomfortable conclusion the 4-week bill is trying to write down is that the Fed's rate decision and its balance-sheet decision are no longer the same decision. The July 23 bill spike, the widening SOFR-IORB drift, the SRF utilisation creep, the emptied ONRRP, and a coupon curve clearing at multi-decade high yields are all balance-sheet observations, not policy-rate observations. Every one of them argues for QT to stop, or a permanent step-up in the standing facility, well before the market prices any rate cut.

• The policy rate is being set against inflation. July CPI at +3.4% y/y, core still +2.5%, shelter and services sticky — there is no cut in this data set. The rate stays.
• The balance sheet is being set against supply. Rising T-bill share, rising coupon issuance, foreign holder repatriation pressure, and a domestic MMF complex that has no more shock absorber. There is no runway to keep running QT into this.
• The market prices the conflict in the bill drift, the SOFR-IORB spread, the SRF utilisation, and the auction tails. Not in the dot-plot. The bill is telling you where the balance-sheet decision will be forced. The dot-plot will tell you nothing about it.

The plumbing tells you before the policymakers do. This time it started telling you on a Thursday afternoon in July, at 3.73%.

Because the story is a balance-sheet story that begins at the front of the curve, most of the standard rate-market dashboards are looking at the wrong prints. Below is the live set of series to watch through Q3 quarter-end and into the Nov TBAC refunding.

This is problem-side analysis, not positioning advice. The positioning conversation depends on the existing book, the mandate, and counterparty exposures — and that is a live conversation, not a chatbot answer.

— Macro Desk • Trivandrum • 29 August 2026