ve tokenomics in one night: a token you cannot sell, a vote you can rent
On the night of 10 November 2021, a nine-person multisig called the Curve Emergency DAO cut off the rewards of a single 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.">pool, the first time it had ever acted. According to CoinDesk, a member of the Mochi team had swapped 46 million dollars of the project’s USDM stablecoin for DAI through that same pool. The proceeds went into ETH and then into CVX, the token that decides how the largest block of Curve votes is cast. On Curve’s reading, Mochi was buying the power to pay itself. Charlie, a Curve contributor, wrote that it “constitutes a clear governance attack”.
That night is the shortest possible explanation of ve tokenomics, the design in which you lock a token for years and receive voting weight in return. Votes decide where new tokens go, so votes have a price. Once they have one, some people rent them out and others try to seize them.
This lesson follows on from our guide to yield aggregators and vaults, and the link is direct: Yearn’s vaults pooled CRV locks for their depositors. The machine has four parts: the lock (veCRV), the vote and the boost (gauges), the middleman (Convex) and the market (bribes). Its best-known copy, ve(3,3), runs up to 21 October 2026, when its largest user replaced the weekly vote with continuous allocation.
A short reminder before the mechanics. Curve Finance is an exchange built for assets that should trade near the same price, and our lesson on specialised AMMs covers its stableswap curve. What matters here is that liquidity providers on Curve earn trading fees plus CRV emissions, and that the emissions are steered by vote. Whoever steers them pays for deep liquidity with someone else’s tokens.
veCRV: how a lock turns time into voting power
The price of a vote is paid in years
You can lock CRV for anything from one week to four years. In exchange you get veCRV, a balance tied to your address. It cannot be transferred, and the lock has no early exit: your CRV comes back when the term ends, and not before. Each address holds one lock, which can be topped up or extended. The four-year ceiling is written into the VotingEscrow contract as a constant.
The formula is simple: veCRV = CRV locked × years left ÷ 4. Here is an example with round numbers. Lock 10,000 CRV for four years and you start with 10,000 veCRV. A year later you hold 7,500, because three years remain. Two years in, 5,000. Lock the same 10,000 CRV for one year and you start at 2,500. The balance falls in a straight line to zero on the day the lock ends, so a holder who wants to keep full weight has to keep extending.
A flash loan cannot buy this kind of vote, and a holder cannot vote for a bad proposal and sell the next morning. Voting weight is measured in the time you have promised to stay.
Holders have taken the deal at scale. On 31 October 2026, a read of the Ethereum contracts at block 26311960 showed 841.4 million CRV locked, or 34.6% of the 2.43 billion CRV minted so far, carrying 774.9 million veCRV. Our own arithmetic from those two numbers puts the weighted average time remaining at about 3.7 years. One large holder pulls that average up, as the Convex section shows.
What a lock pays: fees in crvUSD, votes and a boost
A lock earns three things. The first is a share of protocol revenue, from swap fees and part of crvUSD interest, paid weekly in crvUSD, the stablecoin explained in our lesson on CDP stablecoins. The snapshot is taken on Thursday at 00:00 UTC and distribution follows the next Thursday. The second is the right to vote on DAO proposals and on gauges. The third is a boost on your own CRV rewards if you also provide liquidity.
DefiLlama reports about 24.85 million dollars of revenue to veCRV holders over the year to 31 October 2026, 1.6 million of it in the last 30 days. Its definition matters: 90% of the protocol’s share of pool fees “plus all bribes”, and a bribe reaches only those who voted for the pool paying it. Divide the annual figure by the value of the 841.4 million CRV locked, at about 0.285 dollars around 10:00 UTC that day, and you get roughly 10% by our arithmetic. The last 30 days multiplied by 12 give about 8%.
Neither number is a yield, and nobody promises either. Both set past income against one day’s price. What a lock pays next year depends on volumes, on crvUSD and on how holders vote, and a four-year locker also carries any fall in CRV itself, with no way to sell during the lock.
Gauges: how a Curve vote becomes emissions
The weekly split
A gauge is a contract attached to a pool that receives CRV emissions and pays them to liquidity providers who stake there. Curve’s docs set the rule: the week’s emissions are split in proportion to gauge weights, and the weights update every Thursday at 00:00 UTC. veCRV holders vote on-chain, paying gas, and a vote on one gauge can be changed only once every 10 days.
The flow being divided is finite and shrinking. Curve’s token page gives a maximum supply of about 3.03 billion CRV, with about 1.88 billion set aside for community emissions and issuance falling 16% each August. Those tokens are the protocol’s main budget for attracting deposits, and veCRV holders decide who receives it.
A pool with a large weight can pay depositors more than its trading fees justify. Deposits follow the rewards, and deeper liquidity brings cheaper swaps. For a stablecoin issuer, a deep Curve pool is what keeps the peg close to one dollar, which is why votes became worth paying for.
The boost, as the gauge code computes it
Curve’s user docs print a short formula for the boost, but the rule that counts is the one in the code, here version 3 of the gauge contract. There, each provider has a “working balance”:
working balance = min( l, 0.4 × l + 0.6 × L × ve_user ÷ ve_total )
boost = working balance ÷ (0.4 × l)Here l is your liquidity in the gauge, L is the gauge total, and ve_user ÷ ve_total is your share of all veCRV. With no veCRV you count for 40% of your balance, a boost of 1x. The cap is your full balance, 2.5x, reached once your share of veCRV is at least equal to your share of the pool.
Now the arithmetic, as an example using the 31 October 2026 figures. You hold 100,000 dollars in a 10 million dollar pool, or 1% of it. Full boost needs 1% of all veCRV, about 7.75 million. Even with a four-year lock that is 7.75 million CRV, or about 2.2 million dollars at the price above (that is a lock worth 22 times the position), to boost a 100,000 dollar position. With half of that, the formula gives 0.4 + 0.6 × 100 × 0.005 = 0.7 of your balance, a boost of 1.75x.
The price falls for small pools and rises for large ones. It is also fixed at a moment in time: your boost updates only when you deposit, withdraw or claim. For almost every provider the sums are absurd, and that gap created a business.
Convex: renting the lock instead of owning it
What Convex does with your CRV
Yearn got to the idea of pooling locks before Convex. Its veCRV vault, known as the Backscratcher, kept deposited CRV “locked infinitely” in exchange for a weekly claim on Curve fees, and its depositors gave up the right to withdraw. The product has since been deprecated.
Convex took that further. DefiLlama’s data for it starts on 17 May 2021. Our lesson on composability presents it as one block stacked on another; here is the plumbing. There are two doors. Liquidity providers deposit their Curve LP tokens in Convex, which stakes them through its own veCRV position, so they get a boost without locking anything. CRV holders deposit CRV and receive cvxCRV at 1:1. That conversion goes one way: the CRV is locked as veCRV for good, and the holder can leave only by swapping cvxCRV in a liquidity pool, at whatever the market pays that day.
So a non-transferable, decaying position becomes a transferable token with no expiry. The price of that convenience is that the exit depends on a market and not on a date.
Who gets what: cvxCRV, CVX and vlCVX
cvxCRV stakers receive Curve’s veCRV fees in crvUSD, plus a slice of what Convex earns, and can pick the mix of rewards. CVX is the platform token, with a maximum supply of 100 million, half of it allocated to Curve liquidity providers in proportion to the CRV they claim. vlCVX is CVX locked for at least 16 weeks, and it holds what everyone came for: the vote on how Convex casts its veCRV. Convex’s fee page sets out how the CRV earned by its LPs is shared.
| Share of CRV earned on Curve | Current | Allowed range |
|---|---|---|
| Total platform fee | 17% | Ceiling 20% |
| cvxCRV stakers | 10% | 10-15% |
| CVX stakers | 4.5% | 3-6% |
| Treasury | 2% | 0-2% |
| Harvest caller | 0.5% | 0.1-1.0% |
Fees and ranges as listed in Convex’s documentation, read on 31 October 2026. Only CRV pays this cut: tokens from incentivised pools and veCRV admin fees are not charged.
One address, 54% of the vote
Convex’s voter proxy, the single address that holds its position, had 421.2 million of the 774.9 million veCRV on 31 October 2026, an on-chain share of 54.4%. The veCRV contract keeps its own history, so the same read works for past blocks.
| Date | Block | Convex voter proxy | Share of veCRV |
|---|---|---|---|
| November 2021, around the Mochi affair | 13,590,000 | 137.1 million | 41.7% |
| 22 Apr 2022 | 14,636,327 | Passes half of the total | 50.1% |
| 25 Jun 2025 | 22,780,000 | 414.8 million | 52.0% |
| 31 Oct 2026 | 26,311,960 | 421.2 million | 54.4% |
Our reads of the VotingEscrow contract on 31 October 2026 (CoinDesk counted 136.58 million in November 2021). The shares cover this one address and leave out anything delegated to it. They also explain the 3.7-year average above: the 425.1 million CRV that Convex has locked run to 17 October 2030, close to the maximum, while everyone else averages about 3.4 years by our arithmetic.
Control is less simple than the share suggests. Gauge choices go through vlCVX votes on Snapshot, in rounds held every two weeks, while Curve’s own weights update weekly; the two calendars should not be confused. A Convex multisig must still sign the outcome of those votes, on gauges as on governance. Its docs say it follows vlCVX “in almost all cases”, but that proposals “considered to be blatant attacks damaging either protocol will not be signed for”. That safeguard is a promise made by a small group. On gauges, holders of vlCVX, not holders of CRV, steer more than half of the vote weight on Curve.
Bribes: the market that prices a Curve vote
How a bribe travels
The word is the one the industry uses, and it is not a legal description. A protocol that wants emissions on its pool posts tokens on a vote market. An earlier venue, bribe.crv.finance, was already having “a major effect” according to a Curve-focused newsletter of August 2021, which also noted the “impending launch of Votium”. Votium’s first round for vlCVX holders ended on 21 September 2021. vlCVX holders who vote for that pool collect a share of the tokens in proportion to their weight. Votemarket, from Stake DAO, works on a similar idea. No one signs anything and no one needs to be persuaded, because a contract pays whoever voted the required way.
Take a toy example. A stablecoin issuer pays 100,000 dollars for votes on its pool and the votes win CRV emissions worth 200,000 dollars for its depositors. The issuer gets two dollars of liquidity incentive for each dollar spent, compared with one if it had paid its depositors directly. The extra dollar comes from CRV holders who do not lock, through dilution.
How often does that ratio hold in the real market? A Russian-language post of 25 June 2025 on the Teletype channel named Curve Finance says bribe efficiency “often” exceeds 2. The post is unsigned and promotional, gives no method, and we could not confirm that the channel speaks for the protocol, so read it as a sales argument and not as a measured average. The logic is sound all the same: if a vote costs less than the emissions it directs, the buyer profits, and as buyers compete the price should rise until that gap closes.
Real numbers, with care
Published totals for the bribe market conflict, so we went to the rounds themselves. Llama Airforce, a third-party dashboard, serves each Votium round through a public API. Adding rounds 1 to 134 on 31 October 2026 gives about 341 million dollars offered for vlCVX votes on Curve, by our arithmetic, valued as the dashboard values them. That covers one market only: no Votemarket and no other platform.
| Rounds ending in | Votium bribes for vlCVX votes on Curve |
|---|---|
| 2021 | 46.3 million dollars |
| 2022 | 194.8 million |
| 2023 | 46.7 million |
| 2024 | 26.7 million |
| 2025 | 20.1 million |
| 2026, to 27 October | 6.4 million |
The peak was round 15, ended 5 April 2022, at about 21.4 million dollars. Round 98, ended 10 June 2025, came to about 932,000 dollars, led by RSUP, the Resupply token (341,369 dollars), then INV (198,035) and CVX itself (152,508). The June 2025 post gives similar amounts per token but a total of 1.2 million. Round 134, ended 27 October 2026, came to about 190,000 dollars. The post also says Votium and Stake DAO each take 4%, and that Votemarket charges no fee on voting income.
The best independent evidence is academic. In a paper on the outcomes of the veToken model, Lloyd, O’Broin and Harrigan find that voting outcomes are strongly associated with bribes, and that the cost per vote depends on how it is acquired. Frax, in their account, plays a central part by locking indirectly through yield aggregators and buying weight on vote markets.
The Curve wars in four episodes
The name comes from the contest for influence over Curve’s emissions.
| Date | Event |
|---|---|
| Aug 2020 | CRV launches on Ethereum |
| May 2021 | Convex appears on DefiLlama (17 May) |
| 10 Nov 2021 | Emergency DAO halts rewards to Mochi’s USDM pool |
| 5 Jan 2022 | Peak TVL: Curve 24.3 billion dollars, Convex 21.17 billion |
| Feb 2022 | Solidly launches on Fantom |
| 22 Apr 2022 | Convex passes 50% of veCRV (on-chain) |
| Aug-Sep 2023 | Egorov sells 106 million CRV over the counter, then clears his Aave debt |
| 13 Jun 2024 | Egorov’s loans are liquidated, about 140 million dollars in CRV |
| Nov 2025 | Aerodrome and Velodrome merger announced |
| 21 Oct 2026 | Aero launches and retires weekly voting (22 Oct, 00:00 UTC) |
2021: the race to own votes
Yearn moved first and Convex overtook it. The prize grew with the size of the pools: when DefiLlama’s peak came on 5 January 2022, Curve held 24.3 billion dollars and Convex 21.17 billion, so Convex’s total sat close to Curve’s own. Those dollars rested on the prices of the day: CRV stood at about 6.10 dollars, according to DefiLlama’s price history.
November 2021: Mochi and the Emergency DAO
The Defiant reported that after Mochi’s launch its USDM pool on Curve grew past 100 million dollars, and that about 6,000 ETH of proceeds, some 28 million dollars then, went to buying and locking CVX. Accounts differ on the amount: Charlie’s post says 1.05 million CVX. That post describes the sequence as he saw it: Mochi paid bribes in its own token to draw votes to its gauge, minted its token freely and borrowed USDM against it at a price set by its own oracle. Locked CVX would then have let it vote more emissions to the pool, CoinDesk explained, in a loop. Andre Cronje called the tactic “amazingly scammy”. Mochi had not responded publicly when The Defiant reported.
Charlie told CoinDesk: “Mochi crossed it seven times over last night.” The Emergency DAO stopped the emissions. Banteg, a Yearn contributor and member of the group who said he did not sign that transaction, said: “We know for a fact USDM is a worthless collateral.” When CoinDesk published, the pool held more than 31 million USDM valued at 0.49 dollars each, and whoever still provided liquidity there carried that paper.
Votes were open to anyone with money, including, on Curve’s account, money borrowed against a token priced by its own issuer. A forum member pointed out that the gauge had passed both Curve and Convex votes with no debate on the forum, and that Mochi had promised voters an airdrop. And the fix was not code but people: CoinDesk recorded that the main DAO appoints the group and can reverse its decisions, and Curve’s docs describe a 5-of-9 multisig today.
The decision had critics. In the same forum thread, one veCRV holder called it “a rash and unnecessary action” that set off a run on USDM and so caused the losses it claimed to prevent, while others asked why a public warning would not have done and whether Mochi should have been heard first. Defenders answered that a pool open to manipulation should not receive CRV.
2023: the founder’s loans
The second stress case has nothing to do with voting mechanics and everything to do with concentration. Michael Egorov, Curve’s founder, had borrowed heavily against his CRV across lending markets. The Block reported that in August 2023 he sold 106 million CRV over the counter for 46 million dollars. On or about 27 September he deposited 68 million CRV (about 35.3 million dollars) in Silo Finance, borrowed 10.8 million crvUSD, swapped into USDT and cleared his Aave debt. He still had about 42.7 million dollars of debt on four other platforms, against 253.7 million CRV (about 132 million dollars).
No source says those tokens were locked as veCRV, and a loan is not a governance attack. The case matters because every part of this machine, from bribe efficiency to the reward value of a gauge, is priced in CRV, and one person’s collateral could move that price. It did in the end. On 13 June 2024, The Block reported, citing Arkham Intelligence, that Egorov was liquidated for 140 million dollars in CRV across several protocols, with the token down more than 25% in 24 hours, to 0.262 dollars.
After the peak: same machine, smaller stakes
As of 31 October 2026, DefiLlama shows Convex at 519 million dollars of total staking or liquidity, measured at current market prices. It is the main size metric in DeFi.">value locked, about 2.5% of its peak, and Curve’s exchange at 1.11 billion, about 4.6%. CRV stood at about 0.285 dollars around 10:00 UTC that day, and CVX at about 1.99. The war went quiet and the contracts kept running. A Votium round now moves around 300,000 dollars, under 2% of the April 2022 peak, and it still helps set where Curve’s emissions go.
ve(3,3): what the copy added, and what Aero changed
Solidly: the model that started the family
ve(3,3) joins two ideas. “ve” is Curve’s vote escrow. “(3,3)” is a slogan from OlympusDAO’s game theory: if everyone stakes, everyone gains, and if people sell, everyone suffers. Solidly, a decentralised exchange on Fantom launched in February 2022 and credited by CoinMarketCap’s Academy to Andre Cronje and Daniele Sestagalli, was built around that combination.
It changed three things compared with Curve. Voters earned fees only from the pools they voted for, so a vote became a bet on real trading. Locked positions were NFTs, so they could be sold or borrowed against, where veCRV cannot move. And emissions depended on how much was staked: the more of the supply was locked, the fewer new tokens were emitted, and a staker received new tokens in line with the share staked, which was meant to spare stakers from dilution.
The Block reported that the launch helped make Fantom, for a time, the third-largest DeFi ecosystem among popular layer 1 chains. It did not last. On 6 March 2022 a Fantom Foundation developer tweeted that he and Cronje were leaving, without giving reasons, and the TVL of Solidly and of Solidex, its yield optimiser, fell more than 30% in 24 hours.
Velodrome and Aerodrome: ve(3,3) at scale
Velodrome took the design to Optimism, where DefiLlama’s data begins on 2 June 2022. Aerodrome, on Base and sharing the same MetaDEX02 base as Velodrome according to its builders, begins there on 29 August 2023. In their model, as it ran until October 2026, you locked AERO or VELO for up to four years (100 AERO for four years gave 100 veAERO, one year gave 25) and voted once per weekly epoch. Voters collected the fees and incentives of the pools they backed. Liquidity providers who staked in those pools earned the emissions instead of the fees.
Lockers were protected by a rebase: rebase = weekly emissions × (1 − veAERO supply ÷ AERO supply)² × 0.5. The smaller the locked share, the larger the rebase, paid to existing veAERO holders to limit the dilution of their votes. From around epoch 67, voters could also move the emission rate by 0.01% of supply each epoch, by simple majority, a feature called Aero Fed.
Aero’s dashboard counts 582 million dollars of lifetime exchange revenue across the protocols as of 31 October 2026, the protocol’s own figure. Its best quarter was the last of 2024, at 111.2 million dollars. Q3 2026 brought in 25.8 million.
Aero, 21 October 2026: the end of weekly voting
On that date, at 8:00 PM EDT (22 October, 00:00 UTC), Aero went live on seven networks, merging Aerodrome and Velodrome on a new base from Dromos Labs. The merger had been announced in November 2025. The upgrade page states that AERO and VELO emissions ended at launch and that all gauges were deactivated; the last voting epoch closed at 22:59 UTC on 21 October. One AERO converts to one new AERO, and one VELO to 0.044. Velodrome stays open for trading on four chains, without token incentives.
For the question “what is ve(3,3)?”, this changes the tense without ending the idea. Aero’s economics page now says “No voting. No epochs.” Dromos’s FAQ of 8 May 2026 is more exact: Aero says “staking” where Aerodrome said “locking” and “allocating” where it said “voting”, and a stake still cannot be withdrawn before its term ends. Holders still commit tokens for time and still choose which pools get rewards. What went is the weekly cycle and the rebase. sAERO is, in Aero’s words, “the position you get when you stake AERO for a period you choose”, held as a transferable NFT, and under Predictive Allocation rewards and revenue stream continuously. The upgrade page sets a 47-hour cooldown between allocation changes at launch, subject to change; the May FAQ had planned 48 hours and a shorter wait over time.
Dromos gave two reasons, both aimed at the heart of the old design. Voters earned “historical swap fee revenue” while directing the following week’s rewards, so tokens could keep flowing to pools that had stopped being productive. And new pools waited for the next vote before they earned anything. A second mechanism, Gauge Caps, limits each pool’s rewards to a multiple of its revenue, with a long-run target of about 1.2x across the exchange. Whether this beats voting is untested: the protocol only went live in late October 2026.
| veCRV (Curve) | veAERO, veVELO (until 21 Oct 2026) | sAERO (Aero) | |
|---|---|---|---|
| How long | 1 week to 4 years | Up to 4 years | A period the staker chooses |
| Form | Non-transferable, decaying balance | NFT | NFT, transferable |
| Rhythm | Weights every Thursday, 00:00 UTC | Weekly epochs | Continuous; no epochs |
| Directing rewards | Gauge vote | Gauge vote | Allocation, 47-hour cooldown at launch |
| Dilution protection | None | Rebase | Rebase removed |
| Holders’ income | Share of protocol fees, in crvUSD | Fees and incentives of voted pools | Revenue of allocated pools, as it accrues |
Figures from Aero’s dashboard at 07:12 UTC on 31 October 2026: 1.04 billion sAERO staked, 48.8% of supply, with an average stake duration of 3.93 years. The upgrade page expects realised inflation of about 12% a year, under a ceiling of 20% at launch. The dashboard’s current TVL, 319.39 million dollars, is above what DefiLlama lists for the Aero pools, and we keep the two apart.
Why ve tokenomics spread, and where it cracks
What the model solved
Before locks, farmers took emissions and sold them the same day. ve tokenomics answered with a trade: the most valuable right in the system goes to holders who cannot sell. It gave protocols a way to rent liquidity by the week and not buy it for good, and it put a visible price on attention, since a bribe is a public quote for what a pool’s depth is worth. Tokenomics designers copied it widely, and “dominates” was a fair word for its place in DEX design at the peak. It is less so now.
Its five weak points
- Votes turn mercenary. Locks were meant to reward patience, yet the paper above ties voting outcomes on Curve to bribes.
- Power concentrates. One address holds 54.4% of veCRV, and control sits one layer up, with vlCVX. The Curve founder’s 2023 loans showed the same risk on the token side.
- Locks are illiquid. veCRV has no exit for up to four years. cvxCRV gives one, through a market price. Solidly-style NFTs could be sold, again at a market price.
- Everything depends on the token price. Emissions dilute non-lockers, and bribes and boosts are worth only what the token is worth. CRV has gone from about 6.10 dollars at the peak of January 2022 to about 0.285, and Votium’s yearly bribes from 194.8 million dollars in 2022 to 20.1 million in 2025.
- Rewards look backwards. Weekly votes pay for last week’s activity, which is Dromos’s own critique and the reason it gave for replacing the system.
Governance capture sits on top of all five. Mochi needed only money, and a nine-member multisig was what stood in its way.
Pendle and Aero both changed the recipe
Aero is not the only team to rework the model. Pendle stopped new vePENDLE locks in January 2026 in favour of sPENDLE, as our Pendle lesson explains. The two changes differ: Pendle dropped the long lock, while Aero kept long stakes (3.93 years on average) and dropped the weekly cycle. As of 31 October 2026, Curve’s veCRV still works as described above. The first test of Aero’s version is its own revenue: the dashboard showed 7.6 million dollars for the fourth quarter so far on 31 October, with a projected total of 19.9 million against 25.8 million in the third.
How to read a ve-token before touching one
None of this is advice, and nothing here ranks a token. These are the questions we would put to any vote-escrow design:
- How long is the lock, and is there really no exit before the end?
- Can the position be transferred, and if so, what is the discount on a resale market today?
- Who holds the largest share of votes, and through which layer?
- What share of what holders earn is trading fees, and what share is bribes or new tokens?
- What do emissions cost per year compared with fees, in dollars?
- What happens to bribes and rewards if the token falls 50%?
- Who can stop a pool’s rewards, and can that decision be reversed?
- Does the protocol still work the way the docs you are reading say? Aero shows that a design can change in a day.
Tax and regulation for US and UK readers
We found no ruling from the IRS, the SEC, the CFTC, HMRC or the FCA that addresses vote-escrow rewards or bribes as such. The nearest US rule is Revenue Ruling 2023-14, which treats staking rewards as income once the taxpayer has dominion and control over them. Whether crvUSD revenue, boosted CRV or bribe tokens fall under it has not been settled. See our US crypto tax guide or the UK guide, and keep records of each claim with its dollar or pound value on the day.
