Why
The identity is short enough to hold in your head, which is what makes it useful. Outside money X, collateral ratio c. Deposit X, borrow cX against it, redeposit, repeat. Total deposits converge to X/(1−c) and total borrows to cX/(1−c). Subtract: net TVL = X. Net TVL is exactly the external capital. Everything above it is the same money wearing more coats.
The consequence that matters is that gross divided by net is a leverage multiple, and it is published without being labelled. A market showing deposits of $164.5M against borrows of $122.5M has $42M actually in it and a multiple of 3.9×. Nothing is hidden — both numbers are on the page — but only the larger one gets quoted, and it gets quoted under a word, TVL, that most readers hear as money present.
A second consequence explains a pattern that otherwise looks strange: fee revenue moves superlinearly with the multiple. Fees accrue on total borrows, which scale as c/(1−c) — so a modest fall in the collateral ratio collapses both together. When borrows fall 87% and fees fall 80% in the same window, that is not two facts. It is one fact, and it says the fees were a leverage gauge rather than a demand gauge.
And the general form is worth stating plainly, because it is not confined to lending: everyone reports gross and calls it net. A chain reports TVL that is partly looped. An exchange reports volume that is partly wash or partly the same collateral re-pledged overnight. A payments network reports throughput that counts both legs of a round trip. The discipline is a single division, and the reason nobody performs it is that the larger number is the one being sold.
The honest limits belong here too. The identity is exact only for single-asset recursion; cross-asset loops double-count and need a per-asset decomposition. And a high multiple is not fraud — leverage is a legitimate product, and a lending market exists precisely to provide it. The claim is narrower and harder to argue with: the multiple should be published, and it never is.
How it works
The identity
| Quantity | Formula | Meaning |
|---|---|---|
| Total deposits | D = X / (1 − c) | The headline |
| Total borrows | B = cX / (1 − c) | What the incentive pays on |
| Net TVL | D − B = X | The external capital, exactly |
| Recycling multiple | D / (D − B) | The number nobody publishes |
Worked example
| Reported | Value |
|---|---|
| Total deposits | $164.5M |
| Total borrows | $122.5M |
| Net | $42M |
| Implied collateral ratio | c ≈ 0.745 |
| Recycling multiple | 3.9× |
$42M wearing four coats. Note the tell that comes free: when borrows exceed deposits in a headline, the smaller figure is already a net number and the reporting has mixed units.
Why fees move superlinearly
Fees accrue on B, and B/X = c/(1−c):
| c | Multiple D/X | Fee base B/X |
|---|---|---|
| 0.50 | 2.0× | 1.0 |
| 0.70 | 3.3× | 2.3 |
| 0.745 | 3.9× | 2.9 |
| 0.80 | 5.0× | 4.0 |
A small move in c swings the fee base hard in both directions — which is why an incentive programme ending can take fees down 80% without a single user leaving.
The general form
| Domain | Gross number reported | What net would be |
|---|---|---|
| Lending | TVL | Deposits − borrows |
| Chains | TVL | Same, summed across markets |
| Exchanges | Volume | Net of wash and of re-pledged collateral |
| Repo | Daily turnover | The same collateral re-counted every night |
| Payments | Throughput | One leg, not both |
Everyone reports gross and calls it net. The division is the whole method.
Limits, stated up front
- Exact only for single-asset recursion. Cross-asset loops need decomposition or they double-count.
- A high multiple is not fraud. Leverage is the product. The objection is to the label, not the leverage.
- The multiple is a snapshot; it moves with c, which moves with price.