On 28 July 2024, a proposal to move 499,000 COMP, worth about 24 million dollars, out of Compound’s treasury and into a vault run by a group calling itself the Golden Boys passed by 682,191 votes to 633,636. Rekt, an anonymous crypto outlet, counted only 57 addresses on the ballot. Critics called it a governance attack. Within two days the backers were retreating. That is what a governance token can do: the COMP in those wallets carried real power over real money, and a small group of holders could use it.
Seventeen months later another vote went the other way. On 25 December 2025, Uniswap holders backed the UNIfication proposal with more than 125 million UNI for and 742 against. It switched on protocol fees and scheduled the burn of 100 million UNI from the treasury. For more than five years, UNI had carried no claim on a dollar of what Uniswap traders paid.
A governance token is a ballot paper with a market price. For years most of them gave a vote and nothing else, and the fee switches, buybacks and burns of 2025 and 2026 are attempts to attach the ballot to the money a protocol earns. If the difference between a coin and a token is new to you, read our explainer on coins and tokens first. The previous lesson in this path, vote-escrowed tokens and the Curve wars, covers locks and gauges, so they are only mentioned here.
A holder needs to know what the vote really covers, whether the token gets paid, what the supply schedule will do, and who is in control when it matters. Every figure below carries its date. Nothing here ranks tokens or recommends buying one.
What a governance token lets you vote on, and how the count works
A DAO, or decentralised autonomous organisation, is a group of token holders who control a protocol’s settings and treasury through votes. The SEC’s March 2026 interpretive release gives a useful plain definition: such a token typically allows holders to “vote on certain technical or governance matters, such as software upgrades and treasury expenditures.” In DeFi that usually means interest-rate parameters, listing a new asset, paying a contributor, or changing how fees flow.
It does not mean ownership. A token holder has no share of a company, no right to inspect accounts, and no automatic claim on profits. What the ballot can do is limited to what the smart contracts allow the vote to change, plus whatever people outside the contracts agree to respect.
Forum first, Snapshot next, then a binding vote on chain
Most large DAOs split a decision into stages. Discussion happens on a forum. A straw poll follows on Snapshot, an off-chain tool where wallets sign a message. It costs no gas and moves no funds, so it measures opinion without executing anything. The binding step is an on-chain vote through a contract, usually a version of the Governor design. If it passes, the code runs. Uniswap publishes its process with numbers, which makes it a good model to read:
| Stage (Uniswap, as of 31 October 2026) | Where | Duration | Threshold |
|---|---|---|---|
| Request for comment | Governance forum | At least 7 days | None set; support must be clear |
| Temperature check | Snapshot | 5 days | 10M UNI voting for |
| On-chain vote | Agora or Tally | 2-day wait, 7-day vote | 1M UNI delegated to submit; 40M UNI for to pass |
| Execution | Timelock contract | 2 days | Passed proposals only |
A forum post costs nothing and reaches no contract. When a post on social media says “the DAO voted”, the first check is which stage it means. A Snapshot result has force only if the DAO’s own rules and its multisig signers treat it as binding.
Delegates, quorum and the timelock
Three mechanisms decide how safe a vote is. The first is delegation. Most contracts require you to delegate voting power, to yourself or to someone else, before it counts. Delegates are often professional voters who publish a platform on the forum. Your tokens stay in your wallet, and you can change the delegate later.
The second is the quorum, the minimum support a proposal needs. OpenZeppelin, whose Governor contracts many DAOs reuse, shows a 4 percent quorum of total supply in its example, a figure it says most Governors use, with a voting delay of about one day and a voting period of about one week. Uniswap’s 40 million UNI is the same idea, set at 4 percent of the original one billion.
The third is a time gap. OpenZeppelin’s token standard reads voting power from “past snapshots rather than current balance”, so tokens bought or borrowed after a proposal opens carry no weight. A timelock then holds a passed proposal for a set time before it executes, which, in OpenZeppelin’s words, “allows users to exit the system if they disagree with a decision”. Keep both in mind for the Beanstalk case below, which turned on the first one.
Who actually turns up
Participation is the weak point of the model. A Stanford working paper by Andrew Hall and Sho Miyazaki (October 2024) studied more than 250,000 voters and 1,700 proposals across 18 DAOs. On average about 17 percent of voting tokens were delegated, clustered on the most popular delegates, and real-world participation stayed relatively low. Smaller holders were more likely to delegate than large ones.
An ECB working paper, No 3208, which gives the authors’ views and not the ECB’s, looked at Aave, MakerDAO, Ampleforth and Uniswap using data collected in late 2022 and in May 2023. Around half or more of the token holdings were linked to the protocols themselves or to exchanges. Top voters were mostly delegates, and in many cases they could not be identified or linked to token holders. On Uniswap, the top voter in both snapshots was a16z, the venture capital firm. Delegation rates were 5 percent on Ampleforth, 14.8 percent on MakerDAO and 27.3 percent on Uniswap.
The fee switch turns a ballot into a route for revenue
A fee switch is a setting that lets a protocol keep a slice of the fees that would otherwise go entirely to its liquidity providers or lenders. Once money is kept, the DAO has three choices: burn tokens with it, buy tokens and hold them, or distribute it. All three are usually described as “value accrual”, and they are not the same thing, as the next section shows.
UNI’s launch post of 16 September 2020 already said control of a fee switch would be subject to a 180-day time lock. Nobody turned it on for more than five years. The switch came only after Uniswap had given its governance a legal wrapper, covered further down.
What UNIfication changed, step by step
Uniswap Labs and the Uniswap Foundation published the plan on 10 November 2025. Its main terms:
- On v2 pools, the fee paid to liquidity providers falls from 0.30 to 0.25 percent and 0.05 percent goes to the protocol. Governance enables or disables this across all v2 pools at once.
- On v3, protocol fees are set per pool: at first a quarter of the blockchain that runs exactly as written when it is called, holding and moving funds according to its code, without an intermediary able to change the outcome.">smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">LP fee on the 0.01 and 0.05 percent pools, and a sixth on the 0.30 and 1 percent pools.
- Unichain sequencer fees, after layer-1 data costs and the 15 percent owed to Optimism, go to the burn too.
- A retroactive burn of 100 million UNI from the treasury, meant to approximate what would have been burned had fees been on since launch.
- A growth budget of 20 million UNI a year, distributed quarterly from 1 January 2026.
- Labs sets its own interface, wallet and API fees to zero, and most Foundation staff move to Labs.
After the vote and a two-day timelock, the 100 million UNI left the treasury around 04:30 UTC on 28 December 2025, in one transaction valued at about 596 million dollars at the time according to Cointelegraph. Protocol fees went live on v2 and a selection of v3 pools on Ethereum.
Burning UNI is not a dividend
The design matters more than the headline. Fees collect in a contract called TokenJar and can be withdrawn only if UNI is burned in another contract, Firepit. Someone who wants the accumulated fees has to destroy UNI to get them. No holder receives a payment, and holding UNI in a wallet creates no entitlement. Holders gain, if they gain, only because the supply shrinks. There is no record date and no payment to wait for.
Fees, revenue and holders revenue on 31 October 2026
Three lines on a dashboard are easy to mix up. Fees are what users paid. Revenue is what the protocol kept. Holders revenue is the part DefiLlama attributes to token holders, through burns, buybacks or distributions. The table uses DefiLlama data retrieved on 31 October 2026 at about 09:45 UTC, in dollars.
| Protocol (token) | Fees, 1 year | Revenue, 1 year | Holders revenue, 30 days | Holders revenue, 1 year | Market cap |
|---|---|---|---|---|---|
| Uniswap (UNI) | 931.9M | 61.9M | 8.35M | 61.9M | 4.34B |
| Aave (AAVE) | 637.8M | 90.4M | 0 | 20.4M | 2.03B |
| Hyperliquid (HYPE) | 903.1M | 687.0M | 67.9M | 687.0M | 20.93B |
| Sky Lending (SKY) | 386.0M | 194.0M | 5.64M | 55.7M | 1.60B |
Our arithmetic on those lines: Uniswap kept 9.2 percent of its fees over the last 30 days (8.35 million of 90.9 million dollars) and 6.6 percent over the year, which includes the first two months before the switch. Hyperliquid’s holders revenue was 78.5 percent of fees over 30 days (67.9 million of 86.5 million dollars) and 76.1 percent over the year. Market cap divided by one-year holders revenue gives about 29 for SKY, 30 for HYPE, 70 for UNI and 99 for AAVE.
Two warnings before anyone reads those multiples as a ranking. DefiLlama defines these categories protocol by protocol, and the definitions change. And Aave’s zero over 30 days does not mean buybacks stopped: a buyback paid from the treasury may simply not be counted as holders revenue. Use the table to learn the questions, not to compare valuations.
Buybacks, burns and revenue sharing are three different promises
A burn destroys tokens for good, and the supply falls. A buyback spends protocol money on tokens in the market. What happens next decides its value to holders: tokens that sit in a treasury can be sold or granted later, so the reduction is only as permanent as the next vote. Revenue sharing pays money, or a claim on it, to people who stake or lock the token.
| Method | What holders receive | Supply effect | What to check |
|---|---|---|---|
| Burn | Nothing directly | Falls permanently | Is it funded by fees, or by tokens already in a treasury? |
| Buyback and hold | Nothing directly | Circulating supply falls, reversibly | Who can spend the purchased tokens, and when? |
| Revenue share to stakers | Payment or reward | No direct effect | Is there a lock, and who bears slashing risk? |
Note the source of funds. Uniswap’s 100 million UNI came from its own treasury, so the burn removed tokens the DAO held, not tokens bought from the market. The ongoing burn is a different matter, because it depends on fees.
Aave: from a one-month pilot to a standing budget
Aave announced on 9 April 2025 that holders had approved a first phase of 4 million dollars, about a month of purchases, inside a wider plan of 1 million dollars a week for six months. More than 99 percent voted for it. On 22 October 2025, a forum proposal set out a permanent programme of 50 million dollars a year, buying 250,000 to 1.75 million dollars of AAVE a week. Its Snapshot closed around 2 November 2025 with about 882,000 votes: quorum was reached and the proposal passed. The Defiant has reported that Aave confirmed automated buybacks are running under Aavenomics 3.0, but that article shows no date and no figures, so check the current pace on the forum.
Hyperliquid pays through burns, and holders do not vote on it
Hyperliquid is the useful counter-example, because it separates money from ballots. Its documentation says fees go entirely to the community (HLP, the Assistance Fund and deployers), and that the fund “converts trading fees to HYPE in a fully automated manner as part of the L1 execution”. The HYPE held there is burned, removing it from circulating and total supply. Spot and HIP-3 deployers may keep up to 50 percent of the fees on their own markets.
There is no holder vote behind this. Hyperliquid has no Uniswap-style process in which token holders propose and decide how a treasury is used. Network decisions run through stake-weighted validator votes on matters such as jailing, slashing and quote assets. A token can therefore capture a large share of fees with no DAO at all, which is why “governance token” and “value accrual” are separate questions. The share sent to the fund is often quoted, but Hyperliquid’s documentation gives no percentage, so the table’s observed ratio is the safer figure.
Sky and Compound in brief
Sky, formerly MakerDAO, ties its token to governance and staking: its site says the stake rate is set by Sky Governance through on-chain voting. DefiLlama counts 55.7 million dollars of holders revenue over the year to 31 October 2026. Compound holders took a different route in July 2024, when the compromise described below offered stakers 30 percent of market reserves a year. That was a reported deal, so check the forum for what actually shipped.
Vesting and dilution: the supply schedule can outweigh the fee switch
A token’s tokenomics is the set of rules for who gets how many tokens and when. For a holder, the schedule can matter more than any vote. Two numbers start the analysis: market cap, which is price times the tokens in circulation, and fully diluted valuation, which is price times the maximum supply.
A worked example of dilution
The figures below are invented, for a fictional token called EX, to show the mechanics. EX trades at 2 dollars, has 200 million tokens in circulation and a maximum of 1 billion. Its protocol sends 20 million dollars a year to holders. Over the next 12 months, 20 million tokens are released each month to the team and investors.
| Example (fictional token EX) | Today | In 12 months, same market cap |
|---|---|---|
| Circulating supply | 200M | 440M |
| Price | $2.00 | $0.91 |
| Market cap | $400M | $400M |
| Fully diluted valuation | $2B | $909M |
| Market cap / holders revenue | 20 | 20 |
| FDV / holders revenue | 100 | 45 |
Holding the price at 2 dollars would take 480 million dollars of new demand. Without any, the price has to fall for the market cap to stay put. That is dilution in one row. Now test the burn: 20 million dollars buys about 10 million EX at the example price, while 240 million new tokens arrive in the same year. The burn offsets about one token in 24. Unlocks do not always move a price that way, since some are priced in and some recipients do not sell. The arithmetic only shows how much demand has to appear.
Cliffs and linear vesting: UNI and ARB
Vesting releases insiders’ tokens over time. A cliff is a date before which nothing is released, and a token unlock is the event when tokens become transferable. UNI’s launch allocation shows the usual split:
| UNI at genesis, 16 September 2020 | Share of 1 billion | Tokens | Terms |
|---|---|---|---|
| Community | 60.00% | 600M | 150M claimable at once by past users; 430M for the treasury; the rest for liquidity mining |
| Team and future employees | 21.266% | 212.66M | 4-year vesting |
| Investors | 18.044% | 180.44M | 4-year vesting |
| Advisers | 0.69% | 6.9M | 4-year vesting |
The same post set out a planned perpetual inflation rate of 2 percent a year after the four years. So “total supply” was never a final number, even before the burn.
Arbitrum gives the cleaner cliff. Its documentation puts team, contributors and advisers at 26.94 percent and investors at 17.53 percent of 10 billion ARB, locked for a year from 16 March 2023 and then released monthly over three years. The first release, on 16 March 2024, was a quarter of each block: 673.5 million ARB for the team and advisers, 438.25 million for investors. Cointelegraph, citing Token Unlocks, put that at about 76 percent of the circulating supply. A percentage like this means little without its baseline, so check which circulating figure it uses.
Emissions against burns: net supply is the number
On its own, a burn figure says little. Add up what enters circulation (vesting, liquidity rewards, growth budgets) and subtract what leaves. UNI’s case is a live one: fee burns reduce supply while a 20 million UNI annual budget funds growth. Lock systems such as veCRV change the circulating figure in a third way, by taking tokens out of the market for a while without destroying them.
Ten questions to ask of any governance token
This checklist needs no wallet connection. A tokenomics page, a dashboard and the governance forum are enough. It is a reading method, and it produces questions, not a verdict.
- Does the token do anything besides vote? Look for a claim on fees, a staking reward or a discount, and accept “ballot only” as a possible answer.
- Where does the money come from? Separate real fees from token emissions on a dashboard such as DefiLlama.
- How much of the fees reaches holders, and in what form? Compare fees with holders revenue, then find out whether it is a burn, a buyback held in a treasury or a distribution.
- What is the supply policy? Note the maximum supply, the circulating supply and any inflation that begins after vesting.
- Who received the token at launch? Add up team, investors, advisers and treasury, and compare them with what users received.
- What does the release calendar look like? Find cliffs and monthly releases for the next 12 months, as a share of circulating supply.
- How far apart are market cap and fully diluted valuation? Then divide each by holders revenue, knowing that this is a rough ratio and not a valuation.
- Who holds it? Look at the top wallets and separate exchanges, protocol-owned addresses and funds from individuals.
- How is a vote counted and executed? Check whether voting power comes from a past block, the quorum, the timelock, and who can pause or override.
- Who answers for it legally? Identify any foundation, DAO LLC or DUNA, the company behind the front end, and what the team has promised in its marketing.
Point nine often ends at a multisig: a small group of signers who can act faster than the DAO. Many protocols need one for emergencies. The question is what it may do, how many signers there are, and whether they are named.
When the vote was captured: four cases
| Date | Protocol | What happened | Amount |
|---|---|---|---|
| April 2022 | Beanstalk | Flash-loaned voting power passed a malicious proposal | About $181M drained, about $76M kept |
| May 2023 | Tornado Cash | A proposal with an added function gave the attacker votes | About 483,000 TORN taken |
| July 2024 | Compound | A contested treasury proposal passed narrowly, then was cancelled | 499,000 COMP, about $24M |
| Dec. 2025 to April 2026 | Aave | A DAO and its founding company disputed swap fees | About $10M a year, one DAO member’s estimate |
Beanstalk: the vote that lasted one transaction
In April 2022 an attacker borrowed about 350 million DAI, 500 million USDC and 150 million USDT in flash loans from Aave, plus BEAN and LUSD, and used them to obtain a supermajority of voting power. The malicious proposal, BIP-18, had already waited out a delay of about a day. An emergency function then let it execute at once. Rekt put the loss at about 181 million dollars, with about 76 million kept by the attacker. Voting power followed balances at the moment of the vote, so loans repaid within the same transaction counted in full. A past-block snapshot would have given those loans no weight.
Tornado Cash: read what a proposal executes
In May 2023 the Tornado Cash DAO passed an upgrade that appeared identical to an earlier approved one but contained an added function. After it passed, the attacker used that function to give themselves 1.2 million extra votes and effective control. According to analyst EmberCN, cited by The Block, about 483,000 TORN left the vault. TORN fell from a high of 7.30 dollars on 20 May to a low of 3.75 dollars on 21 May, The Block reported. No loan was needed here: the title of a proposal and the code it runs are two separate things.
Compound: outvoted, or attacked?
Compound is harder to classify, because nothing was hacked. Proposal 289 used the normal process. According to Rekt, an earlier and smaller attempt had failed, and Wintermute Governance warned that withdrawals from the vault would be controlled solely by the Golden Boys multisig. Humpy, a pseudonymous whale, backed the proposal and presented it as a yield product. The community called it an attack. That is the community’s word; no court or regulator has ruled on it. Proposal 289 was cancelled on 29 July 2024 according to Rekt, and CoinDesk reported on 30 July a deal for a DAO-controlled staking product paying 30 percent of existing and new market reserves each year to staked COMP. COMP rose 5.6 percent to 51.27 dollars on the news.
The dispute ended in a negotiation. A narrow on-chain majority had been enough to bring revenue policy to the table, and the 57 addresses counted by Rekt show how few wallets that took.
Aave: the DAO against the company that built it
The fourth case is about the edge of what a vote covers. On 4 December 2025, Aave Labs changed the swap provider on the aave.com front end to CoW Swap. DAO members then said the fees were reaching an address controlled by Labs and not the DAO treasury; one of them, the pseudonymous EzR3aL, estimated about 10 million dollars a year. Labs answered that front ends had always been its own responsibility, while protocol changes stayed with the DAO. A Snapshot vote on moving brand assets such as the aave.com domain to a DAO-controlled entity failed in late December, with 55.29 percent against, 41.21 percent abstaining and 3.5 percent for. Its listed author, Ernesto Boado, said it had been started without his consent.
On 12 April 2026, a binding vote passed by 522,780 to 175,310, about 75 percent. It funds Aave Labs with 25 million dollars in a GHO-based stablecoin, 5 million at once and the rest streamed over 6 to 12 months, plus 75,000 AAVE vesting over 48 months. Under the wider “Aave Will Win” framework, revenue from Aave-branded products would go to the DAO treasury, although that vote covered only the funding. A token governs the contracts. The website, the brand and the team may belong to someone else.
What the law says about governance tokens and DAOs
This section describes the rules as of 31 October 2026. It is general information, not legal advice.
United States: SEC classes, the Howey test and Ooki DAO
The anchor is the SEC’s interpretive release 33-11412 of 17 March 2026, effective 23 March, which the CFTC joined with its own guidance. It sorts crypto assets into five groups: digital commodities, digital collectibles, digital tools, stablecoins and digital securities. The first three, it says, “are not themselves securities”, though “a non-security crypto asset can be offered and sold subject to an investment contract, which is a security.” It places governance rights under digital commodities, and it says a digital commodity “does not have intrinsic economic properties or rights, such as generating a passive yield or conveying rights to future income, profits, or assets” of a business. The release does not replace the Howey test, and it cites the 2017 report in which the SEC found that tokens of The DAO were offered and sold as investment contracts.
The release never mentions fee switches or buybacks. Read with that definition, though, a token that gives a right to protocol income would not fit the digital commodity description, while a burn that pays nobody is harder to place. That is our reading, not an SEC finding. Whether such a token, or a sale built on promises of a team’s work, ends up as an investment contract stays a question of facts. It is also an interpretation, not a statute, and a future commission can revise it.
The CFTC took a different route. On 22 September 2022 it settled with bZeroX and its founders and sued Ooki DAO, its successor. The settlement order found the DAO to be an unincorporated association in which the founders were actively participating members. In June 2023 Judge William Orrick entered a default judgment: a 643,542 dollar penalty, permanent trading and registration bans, and an order to shut the website. The court held that the DAO is a “person” under the Commodity Exchange Act. Nobody defended the case, which limits its weight as precedent, and the two press releases say nothing about ordinary holders who voted.
That risk is why legal wrappers exist. Wyoming added a DAO supplement to its LLC law, effective on 1 July 2021. Its governor then signed the DUNA act, a nonprofit association form, in March 2024, and the act took effect on 1 July 2024. Uniswap used it: the DUNI proposal was executed on 12 September 2025 with about 53 million UNI for, and it requested 16.5 million dollars in UNI for legal defence and tax compliance. Under it, members are not personally liable merely for being members, while criminal liability is not covered and recognition abroad depends on foreign law. A commenter on that forum thread argued that a nonprofit cannot simply pay income to members, which would explain a burn; that is one commenter’s reading, not a legal finding.
For tax, the SEC release says it does not touch the Internal Revenue Code. A burn, a buyback and a staking reward may be treated differently, so start with our guide to US crypto taxes and check your own case.
European Union: MiCA and the fully decentralised grey zone
MiCA, explained in our dedicated guide, does not apply to crypto-assets that qualify as financial instruments, which fall under MiFID II instead. A plain governance token sits awkwardly in what remains. Article 3 defines a utility token as one “only intended to provide access to a good or a service supplied by its issuer”, and a vote on protocol settings is not obviously that. The fallback is the residual group of other crypto-assets, and the token’s features decide.
The hard question is who the rules attach to. Recital 22 says that crypto-asset services provided “in a fully decentralised manner without any intermediary” should fall outside MiCA, but the regulation does not define the phrase, and a service that is only partly decentralised stays covered. The same recital deals with tokens that have no identifiable issuer: they should fall outside Titles II to IV, the issuer chapters, while firms providing services in those tokens stay covered. A token with no identifiable issuer can still reach you through a regulated platform. Concentrated governance is exactly what the debate turns on: the Commission’s targeted consultation, opened on 20 May 2026 with comments due by 31 August, lists concentration of governance power over key functions among its criteria. The ECB paper’s concentration findings point the same way.
United Kingdom
The Law Commission of England and Wales published a scoping paper on DAOs on 11 July 2024. It said no DAO-specific legal entity is needed for now, and suggested testing a limited liability not-for-profit association with flexible governance. It made no formal law-reform recommendations. England and Wales therefore have no dedicated DAO status, and the FCA’s crypto rules are a separate matter that this lesson does not cover.
How to follow a protocol’s governance in an hour a month
You do not need to read everything. Start with the forum’s governance category and read only proposals that move money or change who holds power. Then check Snapshot and the on-chain interface the protocol names. Uniswap’s docs name Agora and Tally, and say Tally wound down in April 2026 with its app continuing under ScopeLift. Reading needs no wallet. Never connect a wallet or sign anything because a link in a chat told you a vote was closing: a fake voting page is an old way to drain one.
Before you delegate, read the delegate’s statement and voting record. An anonymous delegate can be competent, as the ECB paper’s unidentified voters may be, but you have no recourse if the delegate votes against your interest. Delegation can be changed, and that is your check on it. Holding a small balance, you will rarely swing a result. The value of following a vote lies in seeing changes to fees, emissions and treasury spending before they show up on a dashboard.
The next dates to watch are protocol-specific: the next release in the unlock calendar, the next quarter of Uniswap’s growth budget, and the European Commission’s report that follows its MiCA review.
