You open a DeFi dashboard and the first number on the screen is TVL: $2.4 billion, big font, green arrow. It behaves like a bank balance in your head. In reality it’s a receipt count: how much stuff is parked inside the protocol’s contracts at this exact second. Once you know what’s in the pile, and what got counted twice, the number starts doing real work for you.
What the number is counting
Total value locked is the dollar value of assets sitting in a protocol’s smart contracts. The DefiLlama docs lay out the method most dashboards copy: pull every contract address the protocol controls, add up the tokens inside, price them at the current market rate. Deposit 10 ETH and 20,000 USDC into a lending pool and both count. Borrow what’s been lent out and the coins leave the contract, so they drop out of TVL and appear in a separate borrowed figure.
That last part matters more than it sounds. In decentralized finance lending markets, deposits and borrows sit on different lines. A protocol holding $1 billion with $800 million lent out has $200 million of idle liquidity. Check which line the headline number came from.
Scale context: all of DeFi crossed $180 billion in TVL at the peak in December 2021, according to DefiLlama’s historical charts. By July 2022 the total sat below $60 billion. Token prices fell, and TVL is denominated in those same tokens, so the headline shrank with them.
TVL also tells you where a token’s supply actually lives. A token with a $500 million market cap and $350 million of it deposited in farms is a very different asset from one whose float sits on exchanges. Same market cap, completely different sell pressure.
Where the same dollar gets counted twice
Stake 100 ETH with Lido and it lands in Lido’s TVL. You get stETH back, a receipt token. Drop that stETH into Aave as collateral and it counts again in Aave’s TVL. One pile of ETH, two line items. Sum every protocol on Ethereum and the same coins can show up three or four times.
Borrowing loops add more mirrors. Deposit stETH, borrow stablecoins against it, buy more stETH, deposit again. Every lap adds TVL and adds zero new money. Restaking pushed the trick further in 2024: EigenLayer’s TVL peaked around $20 billion that June, and a large share arrived as liquid restaking tokens being looped straight back into lending markets.
Two more inflation sources sit in plain sight. Bridged copies: the same USDC on Ethereum and on a layer 2 gets counted in both chains’ totals. And price: TVL is a dollar figure, so ETH doubling from $2,000 to $4,000 doubles every ETH pool’s TVL without a single new deposit.
Anchor Protocol is the cleanest example of what hides under a big TVL number. It held around $14 billion in deposits in April 2022, nearly all of it chasing a roughly 19% yield paid out of reserves rather than borrower fees. When the yield engine stopped, the deposits left within days. The TVL had been real money the whole time. It had also been there for exactly one reason.
The numbers that pair with TVL
Four checks turn the headline figure into something you can trade on.
Fees and revenue first. A protocol collecting $30 million a year in fees treats its deposits differently than one renting them with token emissions.
Volume against TVL for DEXs. A $50 million pool doing $4 million of daily volume earns its keep. A $50 million pool doing $40,000 is a parking lot.
Composition of the pile. If most of a pool’s deposits are the protocol’s own reward token, the TVL evaporates the day emissions drop.
Withdrawal terms. Capital locked for a year and capital withdrawable in a minute carry identical weight in the TVL figure.
Read the four together. A lending market with $800 million in TVL, $40 million of annual fees, and 70% stablecoin deposits tells you far more than a farm flashing $1.2 billion while paying for every dollar of it with its own reward token.
The short version
TVL measures size, and size alone. It counts what’s in the contracts, counts receipt tokens again when they move, and can’t tell a committed depositor from a yield tourist. Use it to compare protocols doing the same job, strip out the double counts, and check what the deposits actually earn before you treat the number as demand.
