The Shock Absorber That Made the Last Treasury Cash Rebuild Painless Is Empty. The Fed Built a New One, and It Works Differently Than People Think.

Every so often the U.S. Treasury does something mundane that quietly moves hundreds of billions of dollars through the financial system's plumbing: it refills its bank account.

That account is the Treasury General Account, or TGA, the federal government's checking account, held not at a commercial bank but at the Federal Reserve. When it runs low, the Treasury sells a wave of short-term IOUs called Treasury billsTreasury billsShort-term IOUs the government sells to raise cash quickly, usually maturing in a year or less. to rebuild it. Investors hand over cash, the cash lands in the TGA, and here is the part that matters: that cash has to come out of somewhere else in the system.

This quarter, the Treasury is doing exactly that. Its plan for July through September, laid out in the 4 May 2026 borrowing estimate, is about $671 billion of net new marketable borrowing, aiming to lift the TGA from roughly $771 billion in early July toward a $950 billion balance by the end of September.

So the drain to watch is modest by the standards of recent years. The suspense is not the size of it but who pays for it, and, this time, whether the tool that is supposed to cushion it actually works over a full quarter.

The last big rebuild was the closest thing to a stress test. In early June 2023, Congress resolved the debt-limit standoff, and the Treasury, which had been running its account down to almost nothing to avoid breaching the ceiling, rushed to refill it, adding roughly $600 billion over the following months.

In theory, pulling $600 billion of cash out of the system should have drained bank reservesbank reservesThe cash that commercial banks keep parked at the Federal Reserve; it is the system's core working liquidity, so draining it too far can stress money markets. and tightened money markets. It didn't. The reason was a giant pool of idle cash sitting in the Fed's overnight reverse-repo facility, or RRP, a parking lot where money-market fundsmoney-market fundsInvestment funds that hold very safe short-term assets like Treasury bills and repo; they are the main users of the RRP parking lot. stash spare cash overnight at a guaranteed rate. When the Treasury flooded the market with new bills yielding a touch more than the RRP, the funds simply rotated: they pulled cash out of the parking lot and bought bills. The Treasury got its cash, and bank reserves were barely touched.

You can see the mechanism cleanly in a three-week snapshot. Between 31 May and 21 June 2023, the TGA rose $243.6 billion while the RRP fell $217.8 billion to fund most of it. Bank reserves ended the window at about $3,204 billion, essentially where they started.

The RRP did not just soften the drain in 2023. It absorbed nearly all of it, automatically and dollar-for-dollar, so reserves never had to move.

That is the problem now. The parking lot is empty. The RRP peaked at about $2.55 trillion at the very end of 2022. Through 2023 and 2024 it drained steadily as money funds chased higher yields in bills and private repo, and by mid-2026 it holds close to zero on an ordinary day, rising to only a few tens of billions at quarter-end, when banks and funds briefly shuffle cash for reporting.

A cushion that is empty cannot absorb anything. So this quarter's ~$180 billion drain has nowhere obvious to go except bank reserves, the core working liquidity of the banking system. If reserves fall too far, the plumbing that keeps overnight funding markets calm starts to creak. That is the scenario the 2023 episode never had to face.

Here is where the story turns, and where the common framing goes slightly wrong. The Fed has, in a sense, built a replacement cushion, but it is not another pile of parked cash. It is a different mechanism altogether.

On 1 December 2025 the Fed ended 'quantitative tightening,' the multi-year process of shrinking its balance sheet by letting bonds mature without replacing them. Two weeks later it began doing the opposite: buying Treasury bills at a pace of roughly $40 billion a month. It calls these Reserve Management Purchases, or RMPs, and their stated job is to keep bank reserves 'ample' by topping them back up as other forces, like a TGA rebuild, drain them away.

The distinction is not academic. The RRP was a passive buffer: a stock of idle cash that got used up so that reserves were never touched at all. RMPs are an active offset: they let reserves fall and then buy them back. Same target, ample reservesample reservesThe Fed's target regime where reserves are kept plentiful enough that the plumbing runs smoothly; the debate is over exactly how low reserves can go before 'ample' stops being true., but very different plumbing. A passive buffer works automatically, dollar-for-dollar, whether or not anyone is paying attention. An active offset works only as fast and as large as the Fed's trading desk chooses to make it. Its failure mode is not 'runs out' but 'too slow or too small to keep up.'

The RRP kept reserves from ever falling. RMPs let them fall and then buy them back. That is the whole difference, and it is why 'the Fed built a new one' is only half true.

So does the new tool hold? Start with the arithmetic. Bank reserves stood at about $2.97 trillion in early July 2026. Subtract the full ~$180 billion net TGA rise with no offset at all, and reserves land near $2.79 trillion.

That figure sits just above the level Fed Vice Chair Christopher Waller has called 'roughly ample', about $2.7 trillion, and roughly $500 billion above the zone where analysts such as Joseph Wang and Fed staff start to worry about genuine scarcity, somewhere around $2.2 to $2.4 trillion (about 7% of a ~$32 trillion economy). And that is the no-help case. In practice the Fed's ~$40 billion-a-month bill-buying is leaning against the drain the whole time, so the realistic landing is a little higher, in the mid-$2.8 trillions.

There was even a dress rehearsal. In April 2026, tax payments pulled a large slug of cash into the TGA, the kind of steep, temporary drain that has historically run to a few hundred billion. The Fed had pre-positioned its purchases for exactly that, reserves stayed in the ample range, and the drain reversed within weeks. April was a sharp, short shock. This quarter is a slower, longer one, the first true test of whether an active flow can lean against a drain that lasts a full quarter, not just a few weeks.

Because the risk here is a policy miss rather than a mechanical cliff, the warning signs are the ones traders already watch. The clearest is the repo marketrepo marketThe market where banks and funds borrow cash short-term by pledging securities as collateral; it is where liquidity stress shows up first as a rate spike., where banks and funds borrow cash overnight against Treasury collateral, the place liquidity stress shows up first as a rate spike. If SOFRSOFRThe Secured Overnight Financing Rate, a benchmark that tracks the cost of borrowing cash overnight against Treasury collateral in the repo market; if it drifts above the Fed's own paid rate, liquidity is tightening., the benchmark overnight repo rate, starts printing persistently above IORBIORBInterest on Reserve Balances, the rate the Fed pays banks on their reserves; SOFR printing persistently above IORB is a classic early sign that cash is getting scarce., the rate the Fed pays banks on their reserves, that is the classic early signal that cash is getting scarce. Quarter-end days, when balance-sheet pressures peak, are the moments to watch most closely.

Two backstops make a 2019-style blowup unlikely even if the desk misjudges the pace. The Standing Repo Facility, created after the September 2019 repo spike, will lend cash against Treasuries at a set rate and caps how high overnight rates can go. And the RMPs themselves can be dialed up if reserves fall faster than expected. A wobble would most likely show as a few days of firm repo rates and a nudge to the Fed's purchase pace, uncomfortable for leveraged trades like the Treasury basis tradeTreasury basis tradeA hedge-fund strategy that profits from the small gap between Treasury futures and the underlying bonds, financed with cheap overnight repo; it is highly sensitive to repo rates spiking., which lives on cheap overnight financing, but a long way from a funding crisis.

The honest conclusion, then, is undramatic, which is itself the point. The cushion that made 2023 painless is gone, and the replacement genuinely works differently, actively, at the Fed's chosen pace, rather than automatically. But the drain is small, the starting reserve level is comfortable, and the new tool is already running. This is a test of the Fed's judgment about how fast to buy, not a test of whether the system survives. Watch the repo rate, not the headline issuance number.