Editor’s note: Washington is quietly thinning the liquidity buffers that hold the financial system together, at the Treasury and inside the banks. When buffers held on the balance sheet are replaced by liquidity promised on demand, investors start to value assets whose liquidity is transparent and verifiable in real time. That is one reason tokenized Treasuries and digital securities are moving from experiment to infrastructure.
From the Fed put to the Treasury’s turn
After 2008, and especially until 2022, the Fed and other major central banks built a powerful policy backstop under markets, and with it, serious moral hazard. The Fed’s balance sheet grew nearly tenfold, from about $900 billion before 2008 to roughly $9 trillion by early 2022.
One side effect was the “good news is bad news” dynamic. Strong economic data could push markets lower, because investors feared the Fed would raise rates or stay tight for longer. The May jobs report, released on June 5, 2026, was a recent example: hiring came in above expectations, the Nasdaq fell sharply, and market pricing for a Fed hike jumped.
One of Kevin Warsh’s goals as Fed Chair, and not an easy one, is to stop the Fed from being the market’s constant source of trading signals. Today the Fed sits somewhat further back, and the Treasury has the ball. With federal debt above $40 trillion, persistent deficits and rising long-term yields, fiscal policy and Treasury debt management under Secretary Scott Bessent have moved to center stage.
The Bessent Put
First came what markets quickly called the “Bessent Put.” On August 19, the Treasury announced it would raise the maximum size of its liquidity-support buybacks in the 10–20 year and 20–30 year sectors from $2 billion to at least $4 billion per operation, running from September 9 to November 4. The official rationale was to support liquidity at the long end of the curve. Markets also read it as a Treasury growing more willing to intervene against rising long-term yields.
Markets did not, however, treat it as QE. The Treasury cannot create money the way the Fed can. To buy back long bonds, it normally has to issue new debt at other maturities. Buying long bonds and selling short bills does not reduce total government debt; it mainly changes the duration of the debt held by the private sector.
The TGA card: quasi-QE, not QE
Next, the Treasury General Account (TGA) came into play. The TGA is the Treasury’s bank account at the Fed, and its balance currently stands at roughly $950 billion. It handles day-to-day cash management and serves as a liquidity buffer in case the Treasury temporarily loses access to markets. Tax receipts, borrowing and debt repayments make it swing sharply. In 2023, when the Treasury hit the debt ceiling, the TGA fell from about $447 billion at the start of the year to just $23 billion on June 1, then was rebuilt after the debt-ceiling deal.
On August 24, CNBC reported, citing two senior Treasury officials, that using TGA cash to fund long-end buybacks is among the options. If that happens, a “quasi-QE” comparison is fair. When the Treasury uses TGA cash to buy long bonds from the private sector, the TGA liability on the Fed’s balance sheet shrinks, bank reserves rise, and the duration held by the private sector falls. Those two effects resemble QE.
But it is not real QE. Every dollar spent from the TGA today has to be replenished later. If the Treasury refills it through new borrowing, part of the liquidity added now is pulled back out of the market later. The approach does not address the root causes of high long-term yields: the debt stock, the deficit and the term premium. And because only part of the TGA can realistically be used, it is temporary support rather than a structural fix.
Warsh, Jackson Hole and the September hike
The Fed has now confirmed its hawkish turn. At Jackson Hole in late August, Warsh stressed that inflation was still too high, and market odds of a September hike jumped from about 35% to 56%. On September 16, the FOMC delivered: a unanimous 25 basis point increase to a 3.75%–4.00% range, the Fed’s first hike since 2023, with the median projection signaling one more increase in 2026.
That sits awkwardly with a President who openly wants lower rates. Some market participants who struggled to explain why a historically hawkish Warsh was chosen, Citrini among them, read the policy options now on the table as part of an understanding between Warsh and Bessent: the Fed fights inflation, while the Treasury and bank regulation work to contain long-term yields.
The LCR proposal: counting liquidity that doesn’t exist yet
On the Fed side, the most interesting piece of the Warsh–Bessent thesis is a possible change to the Liquidity Coverage Ratio (LCR). The idea is examined in the Fed’s March 2026 working paper FEDS 2026-019 as one of several options that could let the Fed shrink its balance sheet.
The LCR measures a bank’s high-quality liquid assets (HQLA: reserves at the Fed, cash, US Treasuries and certain other liquid assets) against its net cash outflows in a 30-day stress scenario. Its purpose is to ensure a bank can survive a severe short-term liquidity shock.
LCR = HQLA ÷ Net cash outflows over 30 days
Consider a bank holding $114 billion of HQLA, including $30 billion of reserves at the Fed, plus $500 billion of mortgages, corporate loans and other credit. If its regulatory 30-day net cash outflow is $100 billion:
LCR = 114 ÷ 100 = 1.14
The $500 billion loan book has real economic value, but most of it does not count as HQLA, because it cannot be turned into cash quickly without losses. The bank may, however, have pre-pledged part of those loans as collateral at the Fed’s Discount Window. After a haircut, the Fed grants it a borrowing capacity it can draw on quickly in a crisis. Today, that unused capacity does not count toward the LCR unless the bank actually borrows.
The proposal would let a portion of this pre-pledged Discount Window capacity count in the LCR. Industry discussions center on a cap of around 20% of HQLA, though nothing is final. In our example, the bank could hold $95 billion of actual HQLA instead of $114 billion:
95 × 20% = 19 → 95 + 19 = 114 → LCR = 1.14
The bank shows the same 1.14 ratio, and the $19 billion it no longer needs to park in HQLA can go into loans, mortgages or other higher-yielding assets. Across the system, this lowers banks’ structural demand for reserves, so the Fed can shrink its own assets without creating a reserve shortage.
The scale is significant. According to the Fed paper, banks subject to the LCR hold about $4.1 trillion of HQLA, roughly a third of it in Fed reserves. At a 20% cap, the theoretical maximum regulatory relief is about $820 billion, though not all of it would come out of reserves. The paper estimates the actual drop in reserve demand from the LCR change at $50–450 billion, and the combined effect of all options studied at a $1.2–2.1 trillion reduction in the Fed’s balance sheet.
The Treasury has openly backed the idea. In a March 3, 2026 statement, it supported “appropriate capped recognition” of pre-positioned Discount Window capacity in the LCR and other liquidity rules, arguing that the current framework forces banks into excessive self-insurance and deepens the stigma around using the Discount Window.
Will the Fed change how it measures inflation?
No, at least for now. Warsh set up two review teams, one on data sources and methodology and one on the inflation framework, which made some investors worry the new leadership might change how inflation is measured. At Jackson Hole, Warsh was unambiguous: the Fed’s 2% target, measured by the PCE price index, is a “firm, fixed target.” That largely closed the debate over replacing PCE.
Liquidity is changing form, not disappearing
The TGA and LCR options suggest US policymakers are reaching for clever engineering to get through the day rather than solving underlying problems. On the Treasury side, the debate is about running a thinner pre-funded cash buffer; on the banking side, a thinner pre-funded liquidity buffer. In both cases, the system leans less on liquidity it already holds and more on liquidity created when needed: the Treasury relies on markets, and banks rely on the Fed’s Discount Window.
Part of the liquidity that sits ready on balance sheets is being replaced with contingent liquidity. In normal times this is efficient: balance sheets work harder, the Fed can operate with a smaller footprint, and banks can lend more. In a crisis, however, the system’s resilience depends increasingly on the Fed’s role as lender of last resort and on the Treasury’s uninterrupted access to markets.
What these policies share is that they do not remove risk. They change where and when liquidity risk is carried. Less liquidity is held on the balance sheet, and more is assumed to be available when it is needed.
