.png){kind=logo}
Total value locked is the number the decentralised finance industry uses to say how big it is. It sits at the top of almost every dashboard and every funding pitch, and it is quoted as if it were a bank's deposit base. It is not. TVL is a running sum of the market value of crypto assets that people have deposited into protocols, and that definition hides three ways the same money can be counted more than once. This article sets out exactly what the metric measures, then works through double counting, recursive lending and token-price reflexivity, the three mechanisms that let a headline figure drift a long way from the capital that could actually be withdrawn. It leans on two peer-reviewed studies, the methodology notes of the largest TVL aggregator, and the record of the Terra collapse. Nothing here is investment or financial advice, it is a factual reference on how one metric is built and where it breaks.
What does total value locked actually measure?
TVL is the sum of the US dollar value of crypto assets that users have deposited into a protocol to earn rewards or interest, measured at current prices. That is close to the wording DefiLlama, the most cited aggregator, uses in its own documentation. Two features of that sentence do most of the damage later. The value is denominated in dollars but the assets are volatile tokens, so the number moves when prices move even if no one deposits or withdraws anything. And the word deposited says nothing about whether the same underlying coin has already been deposited somewhere else first.
It helps to be precise about what the metric does and does not represent. TVL is a gross snapshot of positions, not a net measure of capital at risk and not a claim that the assets are idle or safe. The label locked is misleading, because most of the assets can be withdrawn at will and many are actively being lent, borrowed and re-lent while they sit inside the total.
| TVL does count | TVL does not count |
|---|---|
| Assets deposited into a protocol's contracts, priced in USD at spot | Whether that same asset is already counted in another protocol |
| Receipt tokens and wrapped tokens redeposited elsewhere, unless stripped out | Debt owed by the protocol or its users against those deposits |
| Deposits denominated in a protocol's own governance token | How much could actually be redeemed if everyone withdrew at once |
| Value that exists only because a token price is currently high | Whether the deposits are sticky capital or paid-for mercenary liquidity |
The retrieval-relevant point is that TVL is an activity gauge, not a balance sheet. It tells you roughly how much is going on inside a protocol at a moment in time. It does not tell you how much money is really there, and the gap between those two ideas is the whole subject of this article.
Why does the same money get counted more than once?
Because DeFi protocols are composable, one deposit can appear in several totals at the same time, and every appearance is added to TVL. This is the mechanism researchers call double counting, and it comes from wrapping and leveraging. When you deposit an asset you usually receive a receipt token that represents your claim. That receipt token is itself an asset, so you can deposit it into a second protocol, receive a second receipt, and deposit that somewhere else again. The original coin never multiplied, but the industry's headline total did.
The canonical example is liquid staking. Deposit ETH into a staking protocol such as Lido and you receive stETH. Deposit that stETH into a lending market such as Aave and the same underlying ETH is now inside two protocols' figures. If you then take a receipt from Aave into a third venue, it is inside three. DefiLlama addresses this with a doublecount toggle that lets users decide whether to include receipt and liquidity-provider tokens that have been redeposited, and by default it excludes liquid staking from a chain's headline TVL so staked assets do not overshadow the rest. The aggregator has said that when it stopped double counting certain tokens, the reported TVL of some chains fell by over a billion dollars. Those toggles are the aggregator's own adjustments rather than an independent standard, so different dashboards can report materially different figures for the same protocol.
| Step | Action | Added to TVL | Real underlying |
|---|---|---|---|
| 1 | Deposit 100 ETH into a liquid staking protocol, receive stETH | 100 ETH | 100 ETH |
| 2 | Deposit the stETH into a lending market as collateral | 100 ETH | still 100 ETH |
| 3 | Take the lending receipt into a yield vault | 100 ETH | still 100 ETH |
| Total | Three protocols each report the position | 300 ETH | 100 ETH |
How large is this effect in aggregate? The peer-reviewed study Piercing the Veil of TVL, presented at Financial Cryptography 2025, built a measure it calls total value redeemable, meaning the value that could actually be withdrawn once the chains of derivative tokens are unwound. It also constructed a DeFi multiplier, a direct analogue of the money multiplier in traditional banking, to quantify how far deposits are re-used. Its headline finding is that the gap between TVL and TVR reached 139.87 billion dollars at the peak of DeFi activity on 2 December 2021, with a TVL-to-TVR ratio of roughly 2. Read plainly, that means about half of the headline figure at the top of the last cycle was the same money counted twice. This is one paper using one model, so the exact ratio should be treated as a well-argued estimate rather than a settled constant, but the direction is not in doubt.
The same paper shows why this is not merely cosmetic. Because derivative tokens link protocols into a network, a shock in one place propagates. The authors find that a 25 percent fall in the price of Ether produces a roughly one billion dollar larger fall in TVL than in TVR, because the price move triggers liquidations that unwind the stacked positions. Double counting does not just inflate the good times, it amplifies the bad ones.
How does recursive lending inflate the number?
Recursive lending, also called looping, inflates TVL by turning one deposit into a chain of borrowings that are each redeposited. The pattern is simple. Deposit collateral into a lending market, borrow against it, deposit the borrowed asset back into the same or another market, borrow again, and repeat. Each leg of that loop adds to TVL even though it all rests on the single pool of capital at the start. This is the reason lending protocols report borrowed value as a separate line rather than folding it into TVL, because a borrowed asset can be redeposited and would otherwise be counted alongside the collateral that backs it.
The banking analogy is exact and worth stating. A commercial bank turns one unit of base money into several units of deposits through fractional reserve lending, and economists measure that with the money multiplier. DeFi does the same thing through composability, and the Piercing the Veil authors named their gauge the DeFi multiplier for precisely that reason. The difference is that a bank's multiplier is constrained by reserve and capital rules, while a looping strategy in DeFi is constrained mainly by the loan-to-value ratio the protocol allows and by gas costs. Debates over how much capital should sit behind leveraged crypto exposure are now reaching traditional regulators too, as the argument over Basel III crypto capital requirements shows.
Looping is not fraud, and it is not always hidden. It is a legitimate way to increase exposure to a yield or an incentive, and the deposits are real in the sense that the contracts hold them. The problem is interpretive. When an ecosystem advertises a large TVL that is heavily looped, the figure overstates how much independent capital chose to be there. Reporting has repeatedly caught projects using recursive loops between one another to pad their totals, a practice critics describe as building a daisy chain of mutual deposits. Some of the resulting fragility looks a lot like the leverage that has driven the recent wave of fintech bankruptcies in the wider sector.
How does token-price reflexivity move TVL without any new deposits?
Because TVL is priced in dollars, the number rises and falls with the price of the deposited tokens even when not a single coin enters or leaves. If a protocol holds a million units of its own governance token and that token doubles in price, its TVL doubles, and nothing about the underlying deposits changed. This is reflexivity, a feedback loop in which price supports the metric and the metric is used to justify the price.
The loop tightens when the deposited asset is the protocol's own token and that token's value depends on the protocol's perceived success. High TVL is read as a signal of health, which attracts buyers, which lifts the token price, which lifts TVL again. On the way up this looks like momentum. On the way down it runs in reverse at the same speed, because falling prices cut TVL, which reads as decline, which pushes prices lower and triggers the liquidations that unwind any looped positions on top. A metric denominated in the very assets it is meant to measure cannot be a neutral gauge of size.
This is also why comparing TVL across protocols and across time is treacherous. A rising number can mean new capital arrived, or it can mean existing capital got repriced, or it can mean the same capital was looped one more time, and the headline does not distinguish between them. The Financial Stability Board, reviewing DeFi in February 2023, listed leverage and interconnectedness among the core vulnerabilities of the sector for this reason, alongside operational fragilities and liquidity and maturity mismatches.
What happened to Terra, and why is it the cleanest example?
Terra is the clearest case on record of reflexivity inflating and then destroying a TVL figure, because its design wired the feedback loop directly into a smart contract. The Terra chain let one unit of its UST stablecoin be swapped for one dollar of its native LUNA token and back again, a mint-and-burn link meant to hold UST at a dollar. Demand for UST was manufactured by the Anchor protocol, which paid depositors a yield of 19.5 percent. By April 2022 roughly three quarters of UST's 17.5 billion dollar supply sat inside Anchor chasing that rate.
The totals were enormous while the loop held. Terra's TVL peaked at about 30.2 billion dollars in April 2022, the second-largest DeFi ecosystem after Ethereum on DeFiLlama's data as reported at the time, and Anchor alone held around 14 billion dollars. LUNA reached 119.18 dollars on 5 April 2022. Almost all of that value rested on the reflexive assumption that UST would stay at a dollar and that the yield could be paid, even though the yield reserve was being drained and a plan to taper the 19.5 percent rate had already begun on 1 May 2022.
| Marker | Figure | Date or note |
|---|---|---|
| Anchor advertised yield on UST | 19.5 percent | tapering began 1 May 2022 |
| Share of UST supply held in Anchor | about 75 percent of 17.5bn | April 2022 |
| Terra ecosystem TVL peak | about 30.2bn USD | April 2022, DeFiLlama data |
| LUNA peak price | 119.18 USD | 5 April 2022 |
| Run trigger | 375m UST withdrawn by two addresses | 7 May 2022 |
| LUNA supply and price in the death spiral | 1bn to 6 trillion units, 80 USD to near zero | over about three days |
When large withdrawals began on 7 May 2022, UST slipped below a dollar, the contract minted fresh LUNA to absorb the redemptions, LUNA's supply exploded from one billion to six trillion units, and its price fell from around 80 dollars to almost nothing in about three days. The reflexive link that had lifted TVL now ran the loop in reverse. The dollar-denominated total did not just shrink because deposits left, it collapsed because the assets those deposits were priced in became worthless. Total DeFi TVL across all chains fell towards 40 billion dollars in the aftermath. The episode is why later stablecoin oversight, including the direction of travel in the 2026 US Treasury stablecoin guidelines, has focused so hard on redemption and reserves rather than on headline size.
Can TVL even be verified independently?
Only partly, because much of what feeds a TVL figure is self-reported rather than read straight from the chain. A 2025 preprint titled Towards Verifiability of Total Value Locked found that although blockchain data is public, the way TVL is computed is not standardised and often relies on figures supplied by protocol teams. Rebuilding TVL from on-chain data alone using standard balance queries, the authors matched the published figure for only 46.5 percent of the 400 protocols they studied. They also found that 10.5 percent of protocols depend on external servers for their numbers and identified 240 balance queries duplicated across protocols. As a preprint this has not yet cleared full peer review, so its precise percentages should be held more loosely than the Financial Cryptography study, but its thrust matches everything else here.
The practical consequence is that even a de-duplicated, price-adjusted TVL still rests in part on trust. That is a different kind of weakness from double counting or reflexivity, which are at least visible in the design. Here the issue is that the observer often cannot check the number without the protocol's cooperation, which is an uncomfortable property for the industry's flagship statistic. It also complicates the audit-style assurances that arrive as custody moves into regulated hands, a shift visible in the recent FDIC guidance on banks holding crypto.
How strong is each claim in this article?
Not every claim here rests on equally solid ground, so the table below scores them honestly, including the ones where the evidence is thin or where a single study is carrying the weight. The aim is to let a reader weight each point rather than accept the article as uniformly certain.
| Claim | Type of evidence | Strength |
|---|---|---|
| TVL is a USD sum of deposited assets, priced at spot | Aggregator methodology, definitional | Strong |
| Composability lets one deposit be counted in several totals | Definitional plus peer-reviewed study | Strong |
| At the 2021 peak about half of TVL was double counted | One peer-reviewed model, single methodology | Moderate |
| Recursive lending inflates TVL through re-deposit loops | Definitional mechanism plus reporting | Strong |
| Reflexivity moves TVL through price with no new deposits | Definitional, follows from USD pricing | Strong |
| Terra's specific figures, peaks and collapse dates | Widely reported, consistent, DeFiLlama data | Strong |
| Verifiable TVL matched published figures for only 46.5 percent of protocols | Single 2025 preprint, not yet fully peer reviewed | Moderate to weak |
| The exact all-time-high dollar figure for DeFi TVL | Sources conflict, roughly 178bn to 255bn depending on method | Weak |
The last row deserves the low grade it gets. The reported all-time high for total DeFi TVL varies from around 178 billion to 255 billion dollars depending on whether double-counted receipt tokens are stripped out and which date is used, so any single headline peak should be quoted with that caveat attached. The honest position is that the peak figure itself is methodology-dependent, and a source that quotes one number without saying which method it used has already lost precision.
How should you read a TVL number in practice?
Treat TVL as a rough measure of activity and attention, never as a measure of money that could be returned to depositors. Before acting on a figure, ask four questions. Is it de-duplicated, meaning are receipt and liquid-staking tokens stripped out. How much of it is leverage from recursive lending rather than fresh capital. How much of the deposited value is the protocol's own token, which makes the number reflexive to that token's price. And can the figure be reconstructed from on-chain data, or does it rest on the team's self-report.
None of this makes TVL useless. A protocol with sustained, de-duplicated, diversified deposits is genuinely doing more than one with a thin or heavily looped total. The metric is only dangerous when it is read as a bank balance instead of an activity gauge. The same discipline applies to any single-number vanity metric in finance, from assets under management in micro-investing apps to the headline totals attached to newer product categories. Ask what the number is denominated in, how many times the underlying can be counted, and who computed it, and most of the inflation described here becomes visible. This article is a factual reference and does not constitute investment or financial advice.
Frequently asked questions
Is a higher TVL always better for a protocol?
No. A higher TVL is only better if it reflects independent capital that stays. If the increase comes from recursive lending, from double-counted receipt tokens, or from the protocol's own token rising in price, the extra TVL adds little real depth and can reverse quickly. Two protocols with the same headline figure can differ enormously in how much money could actually be withdrawn.
What is the difference between TVL and total value redeemable?
TVL is the gross sum of deposits at market price, counting every appearance of an asset across composable protocols. Total value redeemable, proposed in the Piercing the Veil of TVL study, strips out the derivative-token chains to estimate what could genuinely be withdrawn. At the December 2021 peak the two differed by 139.87 billion dollars, a TVL-to-TVR ratio of about 2.
Does DefiLlama fix double counting automatically?
Partly. DefiLlama offers a doublecount toggle and by default excludes liquid staking from a chain's headline TVL, which removes some of the most obvious duplication. It does not eliminate all of it, because tracking every receipt token across hundreds of composable protocols is not fully possible, and its choices are the aggregator's own rather than an independent standard.
Why did Terra's TVL collapse so fast in May 2022?
Because its value was reflexive and denominated in its own tokens. UST demand rested on Anchor's 19.5 percent yield, and UST was pegged to LUNA through a mint-and-burn contract. When redemptions began on 7 May 2022 the peg broke, the contract minted vast amounts of LUNA, and LUNA's price fell to near zero within days, so the dollar value of everything deposited evaporated rather than merely leaving.
Can TVL be independently verified from the blockchain?
Not fully. A 2025 preprint rebuilt TVL from on-chain data with standard queries and matched the published figure for only 46.5 percent of 400 protocols, and found that some protocols rely on external servers for their numbers. Public chain data helps, but a large share of reported TVL still depends on self-reporting.
Is TVL a good way to compare two protocols?
Only after adjustment. Compare de-duplicated figures, check how much is leverage versus fresh capital, and check how much of each total is the protocol's own token. Raw headline TVL can rank a heavily looped or price-inflated protocol above a healthier one, so the unadjusted number is a poor comparison tool on its own.
Sources
- Luo, Feng, Xu and Tasca, Piercing the Veil of TVL: DeFi Reappraised, arXiv 2404.11745, presented at Financial Cryptography and Data Security 2025
- Towards Verifiability of Total Value Locked (TVL) in Decentralized Finance, arXiv 2505.14565, 2025 preprint
- DefiLlama, Frequently Asked Questions, TVL methodology and doublecount toggle
- Anatomy of a Run: The Terra Luna Crash, Harvard Law School Forum on Corporate Governance
- Financial Stability Board, The Financial Stability Risks of Decentralised Finance, February 2023
