On 27 October 2026 DefiLlama put the total staking or liquidity, measured at current market prices. It is the main size metric in DeFi.">value locked in Aave at about $17.45 billion. On 28 October 2025 the same series read $39.38 billion, and its peak was $45.81 billion on 7 October 2025. Did the largest lending protocol lose more than half of its users’ money in a year? Not quite. Part of the drop is the price of ether, part is people leaving, and part is simply how DeFi TVL is built.
TVL is a headline, not a verdict. This lesson takes the number apart, then adds the four measures that tell you more about a protocol: volume, fees, revenue and users. It walks through three free or partly free tools, DefiLlama, Dune and Token Terminal, in words rather than screenshots, and ends with a worked example on Aave and Uniswap where every figure carries its date. All dollar figures below come from DefiLlama’s public API, retrieved on 27 October 2026 around 09:44 UTC, unless another source is named.
You should already know what a lending market, a blockchain, where users swap tokens directly from their own wallets without handing custody of their funds to a company.">DEX and a proof of stake coin with a protocol that stakes it for you and issues a transferable token (such as Lido stETH or rETH) representing the deposit plus rewards, which you can use or sell while the coins stay staked.">liquid staking token are. If not, our lesson on the layers of the DeFi stack explains each one, and the DeFi prerequisites checklist covers wallets, gas and stablecoins.
What DeFi TVL counts: tokens held by smart contracts
DefiLlama, the most quoted aggregator, defines TVL as the value of the tokens locked in a protocol’s smart contracts, priced in dollars. It is a stock, a snapshot of balances at one moment, not a flow of money over a period. Our glossary entry on total value locked gives the short definition; here we look at the rules behind it.
DefiLlama does not simply take a protocol’s word for its number. Each protocol is tracked by an adapter, a small piece of open-source code in DefiLlama’s public GitHub repository that reads balances from the chain and multiplies them by a price; the project’s own team usually writes it, following DefiLlama’s listing guide, and anyone can read it. According to DefiLlama’s methodology page, “Almost all tokens are priced using CoinGecko’s API.” That detail matters later. If a price feed is wrong, or a token barely trades, the TVL inherits the error.
The same page lists what DefiLlama leaves out of its default figure. The list explains many of the gaps you will see between sites:
- Tokens that are not circulating yet or not issued yet, such as locked team allocations, are not counted.
- Native staking on a proof-of-stake chain (ether deposited directly with validators, for example) is not counted.
- Liquid staking protocols are tracked, with their own TVL, but are not added to a chain’s TVL by default.
- Bridges have a TVL of their own, which is not added to any chain.
- Funds sitting in smart contract wallets such as Argent or Safe are not counted.
- Within one protocol, a token is counted once even if a receipt for it is deposited again in the same protocol.
Add those choices together and you get the headline for the whole sector. The sum of chain TVL on DefiLlama was $87.98 billion on 27 October 2026 (the daily history series, updated a little earlier, closed the day at $87.78 billion), against $154.80 billion a year earlier and a record of $177.49 billion on 9 November 2021. Our history of DeFi tells how the 2021 bubble and the 2022 crash got there.
| Date | DeFi TVL, sum of chains (DefiLlama) |
|---|---|
| 9 November 2021 (record) | $177.49 billion |
| 31 December 2022 | $38.29 billion |
| 31 December 2023 | $52.79 billion |
| 31 December 2024 | $115.89 billion |
| 31 December 2025 | $113.31 billion |
| 27 October 2025 | $154.80 billion |
| 27 October 2026 | $87.78 billion |
Ethereum alone held $48.18 billion of that total on 27 October 2026, followed by Base ($6.00 billion), Solana ($5.71 billion), BNB Chain ($5.49 billion) and Tron ($5.40 billion). A later lesson compares those chains for DeFi users.
Lending markets: what sits in the contract and what was borrowed
Lending protocols are where TVL gets confusing. When you borrow USDC on Aave, the USDC leaves Aave’s contract and lands in your wallet. The tokens still held by the contract, which is what TVL measures, are therefore what was supplied minus what was borrowed. DefiLlama’s API keeps the loans in a separate series called borrowed, next to two others called staking (the protocol’s own token staked in its contracts) and pool2 (liquidity pools paired with the protocol’s own token).
For Aave on 27 October 2026, that gives $17.45 billion of TVL and $12.08 billion in the borrowed series, of which $9.65 billion on Ethereum. Staked AAVE tokens added a further $0.33 billion in their own series. The token breakdown hints at the effect. USDC, the dollar stablecoin that borrowers ask for most often on lending markets, appears with only $312 million in Aave’s TVL, a small figure for the largest lender, which fits a market where much of the USDC supplied is out on loan. DefiLlama’s file does not give the borrowed series token by token, so treat that as a reading, not a measurement.
Which number is right? Both, for different questions. TVL tells you what could leave the contracts today. Supplied plus borrowed tells you the size of the loan book. The DefiLlama website lets you add some of these series to the chart through its settings; check which ones are switched on before you compare two screenshots or two articles.
Why TVL moves when nobody deposits or withdraws
TVL is a quantity of tokens times a price. If ether falls 10% and every depositor stays put, TVL still falls. That is the most common trap in DeFi coverage: a headline about money leaving a protocol when the only thing that moved was the market.
Aave’s own breakdown shows how exposed its figure is. Here are the largest tokens in its TVL at 07:38 UTC on 27 October 2026, as listed by DefiLlama:
| Token in Aave’s TVL | Value, 27 Oct 2026 | What it is |
|---|---|---|
| weETH | $3.76 billion | ether.fi restaked ether |
| WBTC | $3.06 billion | wrapped bitcoin |
| wstETH | $2.50 billion | Lido staked ether, wrapped |
| cbBTC | $1.64 billion | Coinbase wrapped bitcoin |
| WETH | $1.13 billion | wrapped ether |
| USDe | $1.03 billion | Ethena synthetic dollar |
| rsETH | $0.80 billion | Kelp restaked ether |
| sUSDe | $0.40 billion | staked USDe |
| USDC | $0.31 billion | dollar stablecoin |
By our reading of that breakdown, liquid staking and restaking tokens (weETH, wstETH, rsETH, osETH, cbETH, rETH and smaller ones) made up about $7.6 billion, roughly 44% of Aave’s TVL. Add plain WETH and about half of the total, $8.74 billion, follows the price of ether. Bitcoin wrappers (WBTC, cbBTC, LBTC, tBTC and others) account for roughly $5.4 billion more.
So a 10% move in ether, all else equal, shifts Aave’s TVL by close to $870 million in the same direction. Stablecoin deposits behave differently: a dollar of USDC stays a dollar. When you compare two protocols’ TVL over a year, ask first what each one holds. A protocol full of stablecoins can look stable while a protocol full of ether looks like it is shrinking, with the same users doing the same things.
Back to the opening question. Between 28 October 2025 and 27 October 2026 Aave’s TVL fell from $39.38 billion to $17.45 billion. Lower prices account for some of it and withdrawals for some; the KelpDAO exploit of April 2026, described below, came in the middle of that year. We found no public source that splits the fall between those causes, so we do not try.
Double counting: the same ether in three places
DefiLlama promises no double counting within one protocol. Across protocols, nothing nets out by default, and DeFi is built on stacking. Follow one ether. You stake it with Lido and receive stETH, which you wrap into wstETH. Lido’s TVL counts the ether. You then deposit the wstETH on Aave as collateral, and Aave’s TVL counts it too. Borrow stablecoins against it, deposit them in a third protocol, and the same starting ether supports a third TVL figure.
This is not a theoretical case. On 27 October 2026 Lido had $22.35 billion of TVL, and $2.50 billion of wstETH sat in Aave’s figure at the same moment. Part of one is inside the other. The same goes for weETH from ether.fi ($4.41 billion of TVL in DefiLlama’s ether.fi staking entry) and rsETH from Kelp. Restaking adds one more floor: tokens restaked through protocols such as EigenLayer are wrapped again into liquid restaking tokens, which are then lent and borrowed. Our glossary entry on liquid staking explains the receipt tokens involved.
DefiLlama handles this with defaults and options rather than one true number. Liquid staking is kept out of chain totals by default, and the website offers settings that add liquid staking, double-counted positions and borrowed amounts to a chart. We could not open the live interface while preparing this lesson, so we do not quote their exact labels; look for the settings menu next to the TVL chart and read what each option says before you rely on a figure.
How big is the overlap? Researchers Yichen Luo, Yebo Feng, Jiahua Xu and Paolo Tasca tried to measure it in Piercing the Veil of TVL: DeFi Reappraised (latest version 4 July 2025). They propose total value redeemable, TVR, the value you could actually take out if every position were unwound. At the peak of 2 December 2021 they find TVL was about twice TVR, a gap of $139.87 billion. They also estimate that a 25% fall in ether cuts TVL by about $1 billion more than TVR, because liquidations of derivative tokens amplify the drop.
The widest version of the trap is the word “total”. Add up every row of DefiLlama’s protocols list on 27 October 2026 and you get about $558 billion, because the list also tracks centralised exchanges (Binance alone at $169.72 billion), bridges and liquid staking. The sum of chains, built with the defaults, was $87.98 billion. Both are DefiLlama numbers; only one is DeFi TVL in the usual sense.
KelpDAO, April 2026: when the stacked layers break
Double counting is also a risk issue. Stacked tokens carry stacked risk, and April 2026 showed it. On 18 April 2026 about 116,500 rsETH, worth about $292 million, were drained from KelpDAO’s bridge built on LayerZero, according to The Block. The attacker supplied 89,567 rsETH (about $221 million) as collateral on Aave V3 and borrowed 82,650 WETH and 821 wstETH against it.
Aave then modelled two outcomes, as reported by The Block on 20 April. If the loss were spread across all rsETH, the token would lose about 15.12% of its backing and Aave would carry about $123.7 million of bad debt. If the loss were isolated on rsETH held on layer 2 networks, bad debt would reach about $230.1 million across those markets. A coalition of protocols led by Aave, DeFi United, then raised over $300 million in ETH to restore rsETH’s backing, The Block reported on 12 May, when it put the bad debt at about $190 million. On 18 May Cointelegraph reported that Aave had returned WETH borrowing limits to their pre-incident levels on the affected networks. We found no final, audited account of how the losses were shared out, so this lesson does not state one; Aave’s governance forum is the place to check.
DefiLlama marks 18 April 2026 on Aave’s chart with the label “KelpDAO hack”. A token accepted as collateral one day could lose part of its backing the next, and every protocol that held it saw its TVL change for reasons that had nothing to do with its own users.
Volume shows what moves through a protocol
TVL measures what sits still; volume measures what goes through. DefiLlama defines DEX volume as trading volume for the period, and publishes it for every exchange it tracks. A DEX with modest TVL can carry huge volume if its liquidity turns over many times a day.
Uniswap is the clearest case. On 27 October 2026 its TVL across versions was $3.72 billion, about a fifth of Aave’s. Yet it handled $1.95 billion of trades in 24 hours, $64.40 billion over 30 days and $701.4 billion over the past year. All DEXs together traded $7.98 billion in 24 hours, so Uniswap had about 24% of the day’s DEX volume by our calculation.
Volume divided by TVL gives a turnover rate: about 52% for Uniswap that day. For an exchange it shows that the liquidity is being used. For a lending protocol trading volume is the wrong yardstick; the amount borrowed, seen above, plays that role.
Volume has its own flaw: it can be faked. A trader, or a bot, selling to itself in a loop creates volume without economic activity, often to farm rewards or make a token look liquid. Our glossary entry on wash trading explains how it works. A sudden jump in volume with no change in fees or users is the first warning sign.
Fees and revenue are not the same thing
Fees answer “how much do users pay?”. Revenue answers “how much does the protocol keep?”. The gap between them is often large. DefiLlama splits the money into four lines in its documentation:
- Fees are all the value collected from users and other sources, the gross figure.
- Supply-side revenue is the share of fees paid out to liquidity providers, lenders or stakers.
- Revenue is the share kept by the protocol itself, fees minus supply-side revenue.
- Holders revenue is the value that reaches governance token holders, through buybacks, burns or distributions.
Token Terminal’s metric catalogue uses close but not identical wording: fees are what users of the app pay across every chain it runs on, revenue is what the app retained. It also tracks token incentives, the value of tokens a protocol hands out to attract users. A protocol that earns $10 million of revenue while paying $15 million in token rewards is buying its activity, and only the incentives line shows it.
DefiLlama writes a methodology line for each Aave version. The Aave V3 adapter, which carries most of the money, says fees include borrow interest, flash loan fees, liquidation fees and penalties paid by borrowers, plus a few smaller items. Most of that interest goes straight to depositors. The table shows the split for the Aave parent entry on 27 October 2026:
| Aave, period to 27 Oct 2026 | Fees | Revenue (protocol) | Supply side (depositors) | Holders revenue |
|---|---|---|---|---|
| 24 hours | $1.16 million | $148,000 | $1.01 million | $0 |
| 7 days | $8.21 million | $1.10 million | $7.11 million | $0 |
| 30 days | $37.55 million | $4.95 million | $32.60 million | $0 |
| 1 year | $643.9 million | $91.2 million | $552.7 million | $20.6 million |
| All time | $2.32 billion | $322.3 million | $2.00 billion | $47.0 million |
The protocol kept about 12.8% of fees over the last 24 hours and about 14.2% over the year. The holders column needs care. The adapter for Aave V4 says “No revenue shared to AAVE holders.”, while the V3 adapter notes that Aave began buying back AAVE tokens with treasury funds after 9 April 2025. The data show $20.6 million of holders revenue over the past year and none in the last 30 days. Owning the token does not give you a share of the fees automatically.
Uniswap shows the same pattern more strongly. Its fees were $2.10 million in 24 hours and $927.6 million over the year, but revenue was $210,000 and $59.5 million: about 6.4% of annual fees, the rest going to liquidity providers. DefiLlama shows Uniswap revenue only for the last year, so the all-time revenue figure equals the one-year figure.
Rankings by fees also depend on what counts as a protocol. Across everything DefiLlama tracks, fees reached $67.0 million in the 24 hours to 27 October 2026. The top payers were not DeFi apps in the usual sense: Tether ($17.26 million) and Circle’s USDC ($6.82 million), whose fees are interest earned on reserves, came ahead of PumpSwap ($3.08 million) and Hyperliquid perpetuals ($2.73 million). Aave V3 ranked tenth with $1.12 million. The parent Aave figure, $1.16 million across all versions, is not a line of that list, which is why the same protocol can sit at different places in two rankings.
Active users: addresses are not people
“Active users” on any DeFi dashboard means active addresses. One person can run dozens of wallets. One address can belong to many people (an exchange, a multisig, a smart wallet) or to none (a bot). Token Terminal defines daily, weekly and monthly active users as distinct users over that window, and its documentation warns that these are distinct counts: summing seven daily figures does not give a weekly one, because the same address is counted every day it returns.
Airdrops make the problem worse. When a project is expected to reward early users, farmers create thousands of addresses to look like thousands of people, a practice called sybil farming. User counts can double before an airdrop and halve after it. Dune offers label tables that tag known addresses, such as exchanges and funds, but labelling is partial, so most addresses stay anonymous.
We did not collect user counts for Aave or Uniswap for this lesson, because the Token Terminal views that hold them asked for an account. Treat any user number you find the same way: ask whether it counts addresses or estimated people, and over what window.
Reading DefiLlama step by step
DefiLlama’s website and its API show the same data. The API endpoints used for this lesson (protocols, chains, fees, DEX volume) answered without an API key on 27 October 2026; DefiLlama also sells a paid API with more endpoints, which we did not use. A typical reading of one protocol goes like this:
- Open the protocols ranking and find the protocol by name. Note whether it is a parent entry (Aave) or one version of it (Aave V3): the figures differ, $17.45 billion against $16.38 billion on 27 October 2026.
- Start with the TVL chart, then the breakdown by chain. Aave had $14.41 billion on Ethereum and less than $0.6 billion on any other chain that day.
- Find the settings menu near the chart and note which extra series are included. Turn options on and off one at a time to see what each one adds.
- Look at the token breakdown to see how much of the TVL is ether, bitcoin or stablecoins, and so how much it depends on prices.
- Move to the fees and revenue section of the same protocol. Check the time window (24 hours, 30 days, one year) and read the methodology note that explains what is counted.
- For an exchange, open the DEX volume dashboard and compare volume with TVL.
- Read the event markers on the chart (hacks, depegs, reward programmes) and the hacks database before drawing a conclusion from a sudden jump or drop.
For a raw check, the API returns plain JSON. The address api.llama.fi/protocol/aave gives Aave’s TVL history, the borrowed series and the token breakdown used in this lesson.
Reading Dune Analytics step by step
Dune works differently. It does not hand you a ready number; it gives you the raw blockchain data and a SQL editor, and users publish their queries as dashboards. Its documentation describes curated tables such as dex.trades (trades across DEXs on EVM networks), prices, labels, lending, staking, bridges and gas, covering more than 55 EVM chains plus Solana. The models behind those tables are open source on GitHub.
- Search Dune for a dashboard on the protocol you study, for example its lending volumes or DEX trades.
- Click through to the query behind a chart you care about. Read which tables it uses and what filters it applies, such as dates, chains or token lists.
- Fork the query, change one parameter (a date range, a chain), run it and compare the result with DefiLlama’s figure.
- Check who wrote it and when it was last updated. A dashboard built two years ago may miss a new version of the protocol.
The danger is plain: a dashboard is someone’s code, and nobody audits it. Two dashboards on the same protocol can disagree because one counts a contract the other forgot. The upside is that you can see exactly why.
Check the cost before you plan on writing queries. Dune’s pricing FAQ, read on 27 October 2026, says: “Free access is view-only and does not include monthly credits or pay-as-you-go usage.” New accounts get a 14-day trial with 2,500 credits; after that, running queries needs a paid plan, and extra credits on the Analyst plan cost $1.875 per 100 on monthly billing.
Reading Token Terminal step by step
Token Terminal applies the vocabulary of company accounts to protocols: fees, revenue, supply-side fees, token incentives, active users and TVL, defined in one catalogue and computed the same way across projects. It is the tool for comparing two protocols side by side rather than digging into one.
- Go to the project page of a protocol and choose a metric tab, such as fees or revenue.
- Set the same time window for every project you compare.
- Read the definition of each metric in the catalogue, and note when it differs from DefiLlama’s: a revenue figure that does not match between the two sites usually comes from a different definition, not a mistake.
- Compare token incentives with revenue for the same period.
We did not verify Token Terminal’s current prices or the exact limits of its free access, so we leave them out; some data views asked for an account when this lesson was prepared.
Worked example: Aave and Uniswap on 27 October 2026
Put the pieces together for two protocols that make money in opposite ways. Aave earns interest on money lent; Uniswap earns a cut of trades.
| Measure, as of 27 Oct 2026 (DefiLlama) | Aave | Uniswap |
|---|---|---|
| TVL (default) | $17.45 billion | $3.72 billion |
| Borrowed series | $12.08 billion | not applicable |
| Volume, 24 hours | not applicable | $1.95 billion |
| Fees, 1 year | $643.9 million | $927.6 million |
| Revenue, 1 year | $91.2 million | $59.5 million |
| Fees / TVL | about 3.7% | about 25% |
| Revenue / fees, 1 year | about 14.2% | about 6.4% |
| Token market cap (DefiLlama field, 27 Oct 2026, about 10:00 UTC) | $2.03 billion | $4.41 billion |
Fees divided by TVL is a rough measure of how hard each dollar works. Aave’s dollar earned about 3.7% in fees over the past year, close to a lending rate. Uniswap’s earned about 25%, because the same liquidity is traded many times. Neither is better: they are different businesses. Two cautions on the ratio. It mixes one year of fees with today’s TVL, and Aave’s TVL was far higher for most of that year, so the ratio overstates its recent efficiency. Measured against the borrowed series instead, Aave’s fees come to about 5.3%.
Aave’s own trend shows in the fees too. Its last 30 days of fees, scaled to a full year (times 365/30), give about $457 million, against $644 million actually collected over the past twelve months. Activity fell during the year, which matches the fall in TVL.
Ratios like P/F and their traps
P/F (price to fees) divides the market value of a protocol’s token by its annual fees; P/S (price to sales) divides it by annual revenue. Both borrow from stock analysis, where a lower figure means you pay less for each dollar of business. Token Terminal and others publish them. The method is simple, and so are its traps.
Take DefiLlama’s market cap field as retrieved around 10:00 UTC on 27 October 2026: $2.03 billion for AAVE and $4.41 billion for UNI. CoinGecko’s pages showed about $2.00 billion and $4.35 billion a few minutes later, so the two sources agree within a few percent. With one year of fees, Aave’s P/F comes to about 3.2 and its P/S to about 22. Uniswap gives a P/F of about 4.8 and a P/S of about 74. Market caps move every minute; redo the sums with the figure of the day.
These figures do not say that a token is cheap or dear. A governance token is not a share: it carries no legal claim on fees, and in Aave’s case holders revenue was zero in the last 30 days. P/F uses gross fees, most of which go to depositors or liquidity providers. One-year figures hide trends, one-off liquidation fees can swell a single month, and token incentives are not subtracted. Use the ratios to compare protocols of the same kind with the same window, and read the definitions first. The market cap used is the token price times the circulating supply, not the value of the protocol.
What US rules and taxes have to do with DeFi data
Nothing on these dashboards is regulatory or tax information about you. Two recent US texts matter for readers who use the protocols they measure. On 17 March 2026 the SEC and the CFTC published a joint release, Application of the Federal Securities Laws to Certain Types of Crypto Assets and Certain Transactions Involving Crypto Assets (Release No. 33-11412, effective 23 March 2026). The interpretation in it is the SEC’s; the CFTC provides guidance relating to that interpretation. It covers liquid staking by name, and says the sale of a staking receipt token backed by a non-security asset, in the circumstances it describes, “does not involve the offer and sale of a security”. That is the stETH-type token that appears twice in TVL figures.
A month later, on 13 April 2026, the SEC’s Division of Trading and Markets issued a staff statement on broker-dealer registration of certain user interfaces used to prepare transactions in crypto asset securities, the kind of front end many DeFi users open every day. It is a staff view, not a rule.
On tax, the IRS rule that would have required DeFi front ends to report users’ trades on Form 1099-DA as brokers was overturned by H.J.Res.25, signed on 10 April 2025. Your gains are still taxable. Our guide to crypto taxes in the United States explains what you report.
A checklist before you trust a DeFi number
- Find the source and the date. A TVL figure without a date and a provider is not usable.
- Check the settings: is borrowed money, liquid staking or double-counted value included?
- Look at what the TVL holds. A protocol full of ether moves with the ether price.
- Ask whether the same tokens are counted in another protocol, especially liquid staking and restaking tokens.
- Separate fees from revenue, and revenue from what token holders actually receive.
- Compare volume with fees and users: volume alone can be washed.
- Treat active users as addresses, and watch for airdrop farming.
- Ask who publishes the number and why. A protocol announcing its own record TVL is not a neutral source.
Run those eight checks on one protocol you use, with DefiLlama open in one tab and the protocol’s own documentation in the other. The next lesson of this course compares the chains where DeFi runs, Ethereum, its layer 2 networks, Solana and others, using the measures you now know how to read.
