Eli Research · Iteration 9 · March 14 2026

The Plumbing Beneath

How the Fed Is Secretly Expanding Its Balance Sheet While Markets Fall — And Why the Liquidity Isn't Reaching You

The Paradox Nobody Is Asking About

The Federal Reserve is adding $40 billion per month to its balance sheet. The reverse repo facility — which drained $2.4 trillion out of the market at its peak — is at zero. QT ended in December 2025. By every traditional measure, liquidity is increasing.

Yet: SPY is down 4.3% in a month. QQQ down 3.2%. IWM down 6.9%. Credit spreads are widening. Even gold fell 1.3% today.

The formula that explained the last three years of market behavior — Net Liquidity = Fed Balance Sheet - TGA - RRP — with its famed 0.95 correlation to the S&P 500, has broken.

The Inversion: The Fed isn't adding liquidity to fuel speculation. It's adding liquidity to keep the plumbing from seizing up. The $40B/month isn't flowing into stocks — it's flowing into the pipes. The water pressure is going to keep the system alive, not to water the garden.

The Equation: Where the Money Lives

Net Liquidity = WALCL − TGA − RRP
$6,600B − $833B − ~$0B
= ~$5,770B
WALCL (Fed Balance Sheet)
$6.6T
As of March 4 — growing $40B/mo
TGA (Treasury Account)
$833B
As of March 6 — draining toward Tax Day
RRP (Reverse Repo)
~$0B
Drained from $2.4T peak (Dec 2022)
Bank Reserves
$2.8T
Lowest since early 2023 (-14.5% from peak)

The net liquidity number (~$5.77T) sits above the Feb 11 reading of $5.706T. By the formula, liquidity is rising. So why are stocks falling?

Because the formula is missing a variable: where the liquidity is needed.

The Three Reservoirs: Where the Water Flows

Think of the financial system as a network of pipes connecting three reservoirs. The Fed adds water. But the water doesn't flow equally to all reservoirs — it flows to where the pressure is lowest.

Fed (Source)
Reserve Management Purchases — buying T-bills & short-dated notes
+$40B/mo
Reservoir 1
Bank Reserves — the overnight funding layer. Banks need these to settle payments, meet reserve requirements, and lend to each other. Currently at $2.8T — the minimum "ample" level. Every dollar from the Fed goes here first.
$2.8T (floor)
↓ (only if Reservoir 1 is full)
Reservoir 2
Funding Markets — repo, T-bill financing, commercial paper. The pipes that connect banks to hedge funds, MMFs, and dealers. Currently under stress — repo rates rising vs. IORB.
Absorbing flow
↓ (only if Reservoir 2 is full)
Reservoir 3
Risk Assets — stocks, corporate bonds, crypto. The "garden" that gets watered last. Only receives overflow from the funding layer. Currently: no overflow.
Starving
The Key Insight: Bank reserves at $2.8T are at the Fed's self-declared floor. The $40B/month in RMPs isn't adding surplus liquidity — it's refilling the minimum. It's like pouring water into a bucket with a hole in it. The water isn't overflowing into stocks because the bucket isn't full yet.

The RRP Story: The Shock Absorber That Broke

The Rise and Fall

Dec 2022
$2,400B
Jun 2023
$2,000B
Dec 2023
$800B
Jun 2024
$400B
Dec 2024
$150B
Jun 2025
$50B
Mar 2026
~$0B

The RRP was a $2.4 trillion sponge. When T-bill yields exceeded the RRP rate, money market funds pulled cash out of the Fed's facility and bought T-bills instead. This was the great migration: $2.4 trillion flowing from the Fed's parking lot into Treasury securities — funding the government's deficit without draining bank reserves.

That sponge is wrung dry.

What This Means: For three years, the government could issue massive amounts of debt without tightening financial conditions because the RRP was absorbing the supply. Money moved from RRP → T-bills, which felt like "free money" to the system. Now? Every new dollar of government debt must come from somewhere that matters: bank reserves, money market funds, or foreign buyers. There is no free buffer left. Every Treasury auction is now a zero-sum game.

The Calendar of Pressure: What Happens Next

October 2025 — The Warning

Fed acknowledges reserves have fallen to "ample" floor. Announces QT will end. Markets celebrate but don't understand: ending QT doesn't add liquidity, it just stops draining it.

December 1, 2025 — QT Ends

Quantitative tightening formally ends after 3 years. Balance sheet stops shrinking. But reserves are still falling as TGA refills and RRP sits at zero.

December 12, 2025 — RMPs Begin

Reserve Management Purchases launch at $40B/month. The Fed buys T-bills and short-dated notes. "It's not QE" — but the balance sheet is growing. The distinction matters: QE targeted long-duration bonds to push down long rates. RMPs target short-duration to maintain reserves.

Year-End 2025 — Liquidity Turmoil

SRF (Standing Repo Facility) usage spikes to $75 billion. Banks scramble for year-end liquidity. The RRP is empty — there's no buffer. Fed's backstop works, but the stress is real.

January 2026 — Calm Restored

SRF drops back to zero. RMPs begin accumulating reserves. Bank reserves stabilize around $2.8T. But the structural vulnerability remains — any future stress has no RRP buffer.

February 2026 — FSB Warning

Financial Stability Board warns of "financial stability challenges in repo markets." The plumbing community is worried even as equity investors remain focused on tariffs and AI.

March 4, 2026 — Warsh Nominated

Trump officially nominates Kevin Warsh as Fed Chair. Powell's term ends May 15. Tillis blockade creates uncertainty about monetary policy continuity. The Fed is operating with a lame-duck chair — optionality constrained.

March 18, 2026 — FOMC Decision

Powell's second-to-last or last FOMC as chair. With reserves at the floor, no RRP buffer, and bank stress indicators rising, does Powell cut to ease funding pressure? Or hold because inflation is still above target?

April 15, 2026 — Tax Day

THE PRESSURE POINT. Tax payments drain ~$200-400B from bank reserves into TGA in days. The Fed is pre-building reserve buffer via RMPs specifically for this. If the buffer isn't big enough, repo rates spike. If repo rates spike, banks tighten. If banks tighten, credit dries up. The RMPs are "significantly reduced" after Tax Day — removing the one source of new liquidity.

November 2026 — Debt Ceiling

Under the House budget plan, the $40.1T debt limit is likely reached. Treasury begins extraordinary measures. TGA drawdown becomes forced — which paradoxically adds liquidity (money flows from TGA into the system). But uncertainty about the ceiling itself tightens conditions.

The Warsh Wildcard: A Lame-Duck Fed

Powell Term Ends
May 15
62 days remaining
Warsh Confirmation
Blocked
Tillis blockade over Powell probe
Senate Banking R Majority
13-11
One defection = blocked
Withdrawal Odds
5%
Polymarket: nomination stable

The liquidity story is complicated by the leadership transition. Powell is a lame duck. He has two, maybe three FOMC meetings left. His incentive to take bold action (rate cuts, expanding RMPs, changing the policy framework) is constrained by the incoming chair's known preferences.

Warsh is known as a hawk. He criticized QE during his time as a Fed Governor (2006-2011). His confirmation is blocked by Senator Tillis over a DOJ probe into Powell — a political fight that has nothing to do with monetary policy but everything to do with the Fed's ability to respond to liquidity stress.

The Forced Move: If liquidity conditions deteriorate around Tax Day (April 15), and Warsh isn't confirmed, and Powell is leaving — who decides? The FOMC still functions, but the lack of a confirmed incoming chair creates a policy vacuum. The Fed's remaining card (expand RMPs beyond $40B/month) becomes harder to play because it sets a precedent for the next chair. This is optionality consumption at the institutional level.

The Market Reality: Where Liquidity Is and Isn't Going

Where the Liquidity IS Going

DestinationEvidenceMechanism
Bank reserves (floor maintenance) $2.8T — lowest in 4 years RMPs ($40B/mo) refilling the minimum
T-bill funding Bill auctions: bid-to-cover 2.8-3.2x Government deficit financing (no more RRP subsidy)
Repo market stabilization FSB warning Feb 2026, SRF spikes year-end Dealers need overnight funding; basis trades need financing
Gold GLD +16.5% 3mo Central bank and institutional demand for non-dollar reserves

Where the Liquidity IS NOT Going

Starving AssetPerformanceWhy
Large cap equities (SPY) -4.3% 1mo Risk appetite requires surplus liquidity, not survival liquidity
Tech (QQQ) -3.2% 1mo Duration-sensitive; rising long rates compress multiples
Small caps (IWM) -6.9% 1mo Floating-rate debt burden; tightest credit conditions hit first
High yield credit (HYG) -2.0% 1mo Spread widening as funding costs rise
Investment grade (LQD) -2.3% 1mo Duration + credit double hit
Long bonds (TLT) -1.7% 1mo Term premium rising; RMPs buy short-end only
Bitcoin (BTC) +6.6% 1mo Anomaly: alternative store-of-value bid like gold?
The Critical Distinction: In 2020-2021, the Fed added liquidity via long-duration QE. That money went to Reservoir 3 (risk assets) because Reservoir 1 (reserves) was overflowing. Today, the Fed is adding liquidity via short-duration RMPs. That money stays in Reservoir 1 because the reservoir is at its minimum. The transmission mechanism from Fed balance sheet to stock prices is broken — not because of amount, but because of destination.

The Duration Ladder: Who Gets the Liquidity

RMPs buy T-bills and notes under 3 years. This directly supports short-duration yields. Watch how the yield ladder reflects this:

InstrumentYield / PriceDailyWhat It Tells Us
3-month T-bill (^IRX) 3.603% -0.06% Anchored by RMP buying + Fed funds rate
SHY (1-3yr ETF) $82.55 +0.06% Supported by RMP demand
SGOV (0-3mo T-bills) $100.52 +0.03% Rock solid — direct RMP target
BIL (1-3mo T-bills) $91.51 +0.03% Rock solid — direct RMP target
5-year note (^FVX) 3.874% -0.26% Transition zone — some RMP support
IEI (3-7yr ETF) $118.75 +0.03% Mild support
10-year note (^TNX) 4.285% +0.28% NO RMP support. Yield rising. Market on its own.
IEF (7-10yr ETF) $95.59 -0.10% Falling as yields rise
30-year bond (^TYX) 4.908% +0.47% NO RMP support. Term premium bid only. Approaching 5%.
TLT (20+yr ETF) $86.54 -0.49% Falling hardest. Long end left to fend for itself.
The Two-Tier Market: The Fed's RMPs create a wall at 3 years. Below that line: yields stable, prices firm, stress contained. Above that line: yields rising, prices falling, term premium exploding. The 30-year at 4.91% is approaching the 5% "gravity well" identified in our yield curve report. The Fed is choosing not to support the long end — that's QE territory, and calling it "reserve management" won't fly.

The Auction Signal: Plumbing Under Pressure

Recent T-Bill Auctions (Short End = RMP Territory)

TermDateBid-to-CoverIndirect %
8-WeekMar 123.1061.2%
4-WeekMar 122.7759.2%
17-WeekMar 113.1957.5%
26-WeekMar 93.0961.6%
13-WeekMar 92.9251.2%

Bid-to-cover ratios of 2.8-3.2x are healthy. Indirect bidders (foreign buyers + institutional) at 51-62% show steady demand. The short end is working because the Fed is a backstop buyer.

Recent Note Auctions (Long End = On Its Own)

TermDateHigh YieldBid-to-CoverIndirect %
10-Year NoteMar 114.217%2.4574.3%
3-Year NoteMar 103.579%2.5559.6%

The 10-year note tells a different story: lower bid-to-cover (2.45 vs 3.0+ for bills), but 74.3% indirect bidding is remarkably high. Foreign buyers are showing up for the long end — but demanding higher yields to do so. The 4.217% auction yield is where the market clears without Fed support.

The Inversion Signal: Short-end auctions are oversubscribed (Fed backstop). Long-end auctions attract demand but at higher yields (no backstop). The spread between 3-month (3.60%) and 30-year (4.91%) is 131 basis points — and widening. The Fed is choosing to let the long end price freely while controlling the short end. This is financial repression with plausible deniability.

The Tax Day Cliff: April 15, 2026

This is the event the Fed has been preparing for since December.

Estimated Tax Day Drain
$200-400B
Flows from bank reserves to TGA
Current Reserves
$2.8T
Already at "ample" floor
RMP Cushion Built
~$160B
4 months x $40B since Dec 12
Post Tax Day
RMPs Reduced
"Significantly" per NY Fed

The math is tight: the Fed has built ~$160B of cushion through RMPs. Tax Day could drain $200-400B. If tax receipts come in at the high end (plausible given capital gains from 2025's market rally), the cushion isn't enough. The SRF becomes the pressure release valve — exactly like year-end 2025 when it spiked to $75B.

The Sequence of Forced Responses

If Tax Day drain is manageable ($200B):

  • Reserves dip to $2.6T, then recover
  • RMPs reduced to ~$20B/month post-April
  • Short-term stress but no systemic issue
  • Equities: neutral to mildly positive

If Tax Day drain is severe ($400B):

  • Reserves crash to $2.4T — below "ample"
  • Repo rates spike — SRF usage explodes
  • Fed FORCED to increase RMPs or reintroduce emergency tools
  • This IS the forced response that changes the game
The Inversion Theory: The extreme tax drain forces the Fed to add more liquidity. But adding more liquidity in a crisis context is expansionary — it's the forced response that reverses the tightening regime. The worse Tax Day is, the more the Fed is forced to respond, and the more liquidity enters the system. The drain creates the flood. This is the self-correcting loop hiding inside the plumbing.

The Broken Correlation: What Comes After

The 0.95 correlation between net liquidity and SPY worked from 2020-2024 because:

  1. QE pushed liquidity into long-duration assets (stocks are long-duration claims on earnings)
  2. RRP drawdown released $2.4T that flowed into T-bills → freeing up cash for risk assets
  3. Bank reserves were abundant — well above the floor — so surplus flowed into speculation

The correlation broke because those three conditions no longer hold:

  1. RMPs target short-duration, not long — liquidity stays in the plumbing
  2. RRP is at zero — no more "free" liquidity release
  3. Reserves are at the floor — every new dollar goes to maintenance, not speculation

When Does the Correlation Restore?

The correlation will re-engage when reserves return to "abundant" (above $3.2-3.5T). At that point, surplus liquidity overflows from Reservoir 1 into Reservoir 3 (risk assets). The math:

Months to Abundant = (Target − Current) ÷ RMP Pace
($3,300B − $2,800B) ÷ $40B/mo
= 12.5 months (if RMPs continue at current pace)

But RMPs are "significantly reduced" after Tax Day. If the pace drops to $20B/month:

$500B ÷ $20B/mo
= 25 months (Q2 2028)
The Implication: Unless the Fed accelerates RMPs beyond $40B/month or restarts long-duration QE, the traditional "liquidity drives stocks" regime doesn't return for 1-2 years. Stocks must rise on fundamentals, not flows. In a tariff/recession uncertainty regime, that's a hard ask.

The Prediction Layer: What Markets See vs. What Plumbing Shows

QuestionPrediction MarketPlumbing SignalDisagree?
Will the Fed cut in 2026? ~22% by Dec (previous iteration) Reserves at floor + rising repo stress = cuts more likely than priced YES — plumbing says cuts needed sooner
Recession by end 2026? 34% (Polymarket) Credit tightening from reserve scarcity hits small business first PARTIAL — plumbing stress usually leads recession by 6-12mo
S&P negative Q1? 71% No surplus liquidity for stocks until reserves refilled ALIGNED — plumbing confirms no near-term fuel
Warsh confirmation? ~80-90% (per individual senator votes) Policy uncertainty keeps RMP expansion off the table
The Signal: The biggest disagreement is on Fed cuts. Prediction markets see ~22% chance of cuts by December. But the plumbing is screaming: reserves at the floor, RRP exhausted, repo rates rising, FSB warnings. If Tax Day stress forces the Fed's hand, cuts come faster than anyone expects. The plumbing doesn't care about inflation targets — it cares about whether the system functions.

Bottom Line: The Invisible Architecture

Everyone is watching tariffs, AI earnings, and Powell's press conferences. Almost nobody is watching the plumbing. But the plumbing is what determines whether markets can rally even if the narrative improves.

Right now, the plumbing says: no.

Bank reserves are at the minimum. The RRP buffer is gone. The Fed is adding $40B/month just to keep the lights on. Every dollar of new Treasury issuance competes directly with the financial system for funding. The long end of the yield curve is unsupported and rising. And the biggest liquidity drain of the year — Tax Day — is 32 days away.

But here's the inversion: the plumbing that's starving stocks today is building pressure for the forced response that feeds them tomorrow. If Tax Day stress forces the Fed to expand RMPs, or if reserve depletion forces an emergency cut, or if the debt ceiling forces Treasury to drawdown TGA (releasing hundreds of billions into the system) — each of these "crises" triggers a liquidity injection that was unavailable before the crisis happened.

The system must break slightly before it can be fixed. The fix IS the liquidity event. And the calendar is pointing to April.