Recently, discussions about TVL have picked up again. Many people still consider it the only standard for evaluating DeFi projects, but in fact, what TVL is and what it represents are much more complex than they seem.

Let's review what happened earlier in the Solana ecosystem. A developer used 11 identities to stack a bunch of protocols on Solana, artificially inflating the on-chain TVL. After this was exposed, DeFi Llama changed its TVL calculation method in August last year, directly removing double counting between protocols. This incident prompted people to start reflecting—what is TVL? Can it truly represent a project's strength?

Actually, at the level of DEXs and lending protocols, the meaning of TVL is fairly straightforward. Uniswap's TVL is real liquidity—no staking, no mining incentives, just the money in the pool. But Curve and Sushi are different; they allow staking governance tokens to earn a share of fees. This staked token portion can theoretically be included in TVL, but DeFi Llama now lists this separately.

Lending protocols are even more interesting. Compound's TVL is actually the "deposit minus borrow"—the total deposits minus total loans, representing how much liquidity is still available in the protocol. Aave is similar but includes the portion of staked AAVE tokens. MakerDAO is completely different because it lends out DAI issued by the protocol itself, which doesn't affect the locked funds, so its TVL is simply equal to total deposits.

Where's the problem? Some projects are built on top of other protocols, which can lead to double counting at the blockchain level. For example, Yearn and Convex Finance are yield aggregators built on Curve. Convex holds a large amount of CRV tokens and stakes them to help users mine on Curve for higher yields. In August last year, Convex's TVL reached $4.47 billion, but in reality, all that money was in Curve, causing double counting.

Liquidity staking is also a big issue. Lido's stETH derivatives are widely used in other DeFi projects—about 21.6% are used as collateral in Aave, and about 14.7% provide liquidity in Curve. These are included in the respective projects' TVL, but in fact, it's the same funds. As an intermediary tool, Instadapp helps users operate across protocols. Its TVL peaked at around $13.5 billion, but all these funds are stored in other protocols, leading to a staggering amount of double counting.

So now you understand—what TVL is depends on which type of project you're looking at. The TVL of a single DEX or lending protocol is still relatively reliable for cross-comparison. But at the blockchain level, the previous data bubble was indeed serious. Although DeFi Llama's adjustments caused on-chain TVL to drop significantly, they also revealed more accurate data to the market. This "bubble burst" process is actually beneficial for long-term development.