What TVL Means in DeFi and Why It Misleads

Total value locked, or TVL, is the dollar value of the tokens sitting in a protocol's smart contracts. It is the most quoted figure in DeFi mostly because it is the easiest one to compute. It ranks protocols on dashboards and fills pitch decks, and it gets cited as evidence that a lending market or exchange is safe to use. It is a reasonable measure of size and a poor measure of almost everything else.

The gap between the two can be large. Researchers at UCL's Centre for Blockchain Technologies and Nanyang Technological University rebuilt DeFi's totals using only assets that could actually be redeemed. At the market peak on December 2, 2021, TVL exceeded that figure by $139.87 billion, which made it roughly twice the real amount. A depositor who reads TVL as money that is really there is, near the top of a cycle, counting a good part of it twice.

DefiLlama, the aggregator most people consult, defines TVL as "the value of any tokens locked in the contracts of a protocol / platform." In practice it reads each contract's token balances and multiplies them by a price. Prices come almost entirely from CoinGecko. When a token is not listed there, DefiLlama falls back on on-chain methods such as comparing pool weights on Uniswap V2.

The docs exclude several things. Team tokens in vesting contracts do not count, and neither does native staking that secures a chain (ATOM on the Cosmos Hub is the example given). Funds in smart-contract wallets such as Gnosis Safe are left out. Bridges are tracked, but their TVL is never added to any chain's total. A multi-chain protocol's deposits are credited to the chain where the user actually deposited. Within a single protocol nothing is counted twice, and borrowed assets are not counted at all.

That rule stops at the protocol's edge. The team has said plainly that a chain's TVL "just involves summing the TVL of each project." The team gives its own example. A user who locks 10 ETH in MakerDAO to mint 100 DAI, then deposits that DAI in Curve, adds both amounts to the chain's figure, although only the 10 ETH is new money. In the team's words, the money from that one deposit can "easily end up in more than 10 different protocols."

The multiplication by price has its own effect. TVL rises and falls with token prices even when not a single token moves in or out. Wrapped and re-deposited tokens make the swing larger. The same paper estimates that a 25% fall in ETH would cut TVL across six large protocols by about $1 billion more than it would cut redeemable value.

Aave, April 2026. Around April 18, Aave's TVL fell from about $26.4 billion to about $17 billion. Critics argued that the earlier figure had been propped up by looping, in which users deposit, borrow, redeposit and repeat. DefiLlama's founder, 0xngmi, replied that DefiLlama "doesn't count borrowed assets in TVL." He added that duplicate Ethena-related positions flagged by users had already been removed from its breakdowns. The headline figure was defensible. The episode still showed how few depositors knew what the headline figure contained.

Staked ETH, stacked. A market report dated September 23, 2026 put deduplicated DeFi TVL at $96.90 billion. On the same data, lending (Aave, Morpho) plus liquid staking and restaking (Lido, EigenLayer, ether.fi) summed to $136.43 billion, or 141% of the whole. The explanation is layering. ETH is staked, the receipt is restaked, and the result is posted as collateral, so each layer appears on some protocol's page.

Rented deposits. Capital that arrives for token rewards tends to leave when the rewards end. A September 19, 2026 report found that 81.6% of the advertised yield on Aerodrome Slipstream came from AERO emissions rather than trading fees. The same report found restaking rewards for many operators below 1%, down from promotional rates above 20% during the 2024 points boom. TVL records that capital while it stays. It does not show why the capital came.

The strongest defense of TVL deserves to be stated in full. DefiLlama applies one published method to every project. It excludes borrowed assets, refuses to double count inside a protocol, and fixes errors when users report them. The team argues that selectively excluding protocols would create arbitrary gaps. It also points out that a figure which includes composability still says something real about how much DeFi is being used. All of this is fair. None of it helps a depositor at a single protocol. That depositor needs to know how much of the figure is new money, how much moves with the price of ETH, and how much will leave when rewards stop, and TVL answers none of those questions.

The difference matters in practice. By one April 2026 estimate, Aave held somewhere between $27 billion and $44 billion in TVL but earned about $89 million in fees in the first quarter. The same analysis found that roughly 90% of DeFi protocols fail a positive net-revenue test once the cost of token incentives is subtracted.

TVL answers one question well: is this protocol large? Before depositing, the useful reading is to set it beside three other figures. The first is fees or revenue, which show whether anyone pays to use the thing. The second is the share of the advertised yield paid in the protocol's own token, since emissions-driven yield disappears when the budget runs out. The third is whether the TVL chart tracks the price of ETH more closely than it tracks deposits. How concentrated deposits are in a few wallets matters too, and the headline figure does not show it at all. For the wider picture of how a protocol works, TVL is where the analysis starts, not where it ends. A large TVL shows that capital is present, not that it is productive, and on its own it is no reason to deposit.