How Investment DAO Votes Are Weighted

In February 2026 more than 13,000 Jupiter DAO voters chose to keep the "Jupuary" airdrop, and lost. Measured by weight, the rival "Zero Net-New Emissions" option took 73.9% of the vote, because wallets holding over 1 million JUP backed it at an 81.7% rate (SolanaFloor). Investment DAO votes work this way almost everywhere. A ballot counts for the tokens behind it, sometimes adjusted for how long those tokens are locked or how long they have been committed, and the number of people voting is a curiosity. This page is for holders of governance tokens who want to know how their weight is computed, what changes it, and what it can and cannot buy.

The common mistake is to assume that secrecy is the cure for whale pressure. Encrypted ballots do help. They stop large holders from watching the running tally and sniping at the last minute, and Shutter Network reports that more than 881 DAOs have used its version. But researchers at Cornell examined 3,844 yes/no votes across 31 DAOs and found that at least one voter's choice could be inferred from the final tally in 81% of them (DL News). When a handful of wallets carry most of the weight, the result itself gives them away. The real leverage for a small holder lies elsewhere: in delegation, in voting early, and in the procedural rules that decide how long a vote runs and how long execution waits.

Every DeFi DAO turns a token balance into voting power, and then decides whether anything else should modify the conversion. There are three common answers.

Plain token weighting. One token, one vote. Aave calculates governance power from a holder's AAVE, stkAAVE (staked AAVE) and aAAVE (the Aave v3 aToken) balances on Ethereum mainnet (Aave governance forum). COMP holders on Compound either vote directly or delegate their power to a representative (SpendNode). Jupiter counts staked JUP. Nothing in the formula rewards conviction, patience or headcount.

Vote-escrow (ve). Curve's model makes weight depend on time as well as size. CRV is locked for up to four years to receive veCRV, a non-transferable token whose balance falls linearly to zero as the lock nears expiry (Curve docs). veCRV holders allocate their power across liquidity gauges, which decides where newly minted CRV goes. The docs note that the design was later copied widely across DeFi as "vote-escrow tokenomics".

Conviction and quadratic weighting. These are attempts to bend the line between capital and power. Under quadratic voting, power is the square root of tokens committed, so 10,000 tokens produce 100 votes rather than 10,000. Conviction voting grows a vote's weight the longer tokens stay behind a proposal, using tokens × (1 minus e^(-time/half-life)), which reaches about 95% of its maximum after roughly a month (LedgerMind, 2026). LedgerMind is a secondary blog rather than a protocol source, and its figures below are attributed to it accordingly.

Delegation sits on top of all three. A holder can keep custody of the tokens and hand the voting power to someone else. Aave goes further and splits the right in two: voting power and proposition power (the right to put proposals forward) can be delegated to different addresses. Delegation is how most small holders end up represented at all, and the basics of how delegation works carry over from one DAO to the next. It also concentrates weight. LedgerMind reports that Optimism reaches 73% voting-power delegation coverage against an 18% industry median, and that ENS pre-delegates 100 million tokens by default, so holders must opt out rather than opt in.

The failures the sources report are rarely bugs. They are the formula working as written while the people it was meant to serve are not watching.

The last half hour. In a Compound vote reported in September 2026, a proposal redirected about $24 million in protocol reserves after 82% of its support arrived in the final 34 minutes of the voting window (SpendNode, citing CryptoSlate). Delegates who may have held enough votes to block it had no practical time to mobilize. The proposal cleared quorum and the approval threshold and executed as coded. How to recognize it: a vote that looks settled until shortly before the close. What to do: hold delegates to voting early, and treat a DAO without an execution delay longer than its voting window as one where a late swing is final.

Sniping and herding. The ENS temp check named the "sniping blockholder" directly: large holders who wait for the running tally and then cast the deciding vote. Shutter describes the same dynamic, along with herd behavior after visible early votes and apathy when an outcome looks locked in. One ENS delegate summarized the case against live tallies plainly.

Insider weight that looks like community weight. The largest voting wallet in the Jupiter emissions vote staked 27.7 million JUP. Jupiter's founder, "Meow", confirmed it belonged "to one of our core cofounders", holding vested allocations. The top 10 wallets held 22.5% of the weight. A further 350 million JUP (5% of supply) went to stakeholders from the earlier Mercurial Finance project; 182 million had vested by February 2026 and 168 million were still to vest. How to recognize it: a vesting schedule that keeps adding weight to the same addresses. Holders should read the vesting tables before assuming a vote is a community decision.

Power pooled one level up. In ve systems, holders who want liquidity hand their locks to an aggregator, and the aggregator votes. LedgerMind reports that Convex Finance controls 51.2% of all veCRV, and that a single Curve gauge-vote round in February 2026 paid $4.7 million in bribes, $890,000 of it to the most-bribed gauge. A veCRV holder voting alone is voting against that.

Delegates who are not who they appear. LedgerMind cites a $2.3 million delegation bribe offer on ENS and one entity that secretly ran 18 delegate identities holding 23% of a DAO's power. Its account does not name that DAO and the claim could not be checked against a primary source here. The defense it implies is ordinary: prefer delegates with a long public record and an identity that is accountable for it.

Boosts that freeze. Curve's boost is calculated at the last checkpoint and then held, while the underlying veCRV keeps decaying. A liquidity provider who keeps touching the gauge may recompute a lower boost than the one they had.

Secrecy that leaks. The strongest case against this page's conclusion is that private voting already solves whale pressure, and that case deserves to be stated fully. Shutter's threshold-encrypted voting has run for three years without a reported privacy breach, more than 881 DAOs have used it, more than 372,914 votes have been encrypted through it, and 87% of adopting DAOs still use it a year later, by Shutter's own undated count on its product page. An earlier Shutter post put adoption at over 600 DAOs, so the number is growing. The ENS delegates who discussed it favored a pilot. A joint Shutter and Ethereum Foundation PSE report compares 12 private-voting protocols on eight dimensions, and tools such as MACI let a voter quietly invalidate a coerced ballot, so that a briber cannot check what was bought.

All of that is true, and it does not move the conclusion. The Cornell paper (arXiv:2509.17871, September 2025) shows why: when voting weights are public or inferable from on-chain balances, the final tally can be decomposed. Across its 31 DAOs, the researchers identified every voter's choice in about 30% of votes, and in DAOs with fewer than 45 participants they unmasked every voter 90% of the time. Large holders stayed identifiable even in bigger DAOs. Their own summary: "Extreme voting weight concentration fundamentally limits B-privacy improvements in most real-world cases." Hiding the ballot removes the running tally, which is worth doing. It does not hide a whale whose weight is larger than everyone else's put together.

One claim in circulation should also be corrected. A secondary article describes Shutter's permanent shielded voting as already in production. Shutter's own announcement says only Stage 1, a proof of concept on a forked Snapshot interface, is complete, with testnet integration and mainnet launch still pending (Shutter). What is live on Snapshot today is the time-limited version, which decrypts every vote once the proposal closes.

A ve token rather than a plain one. Weight then depends on the calendar. Two holders with identical CRV can carry a fourfold difference in power, and both will see it shrink unless they relock.

Aave's two powers. Because voting and proposition power can be delegated separately, a holder who has delegated votes may still hold the right to propose, and the reverse. Checking only one of them gives half the picture.

A small electorate. Below 45 participants, the Cornell team could recover every choice nine times in ten even under a secret ballot. In a small investment DAO, private voting offers protection from the running tally and little else. The researchers' suggested mitigations are adding noise to tallies, which raised bribery costs by about 1.5x on average and more than fourfold when five or more wallets decided the outcome, and curbing whale influence directly.

Lido's delegate rules, by quarter. To qualify for Lido's Delegate Incentivization Program, a delegate needed at least 1,000,000 LDO held or delegated on Snapshot or Aragon as of January 19, 2026. From Q3 2026, the delegate must be delegated on both venues at once. Rewards are now paid in USD stablecoins rather than LDO, within three weeks of the quarterly report, after a Delegate Oversight Committee checks compliance.

Shielding chosen per proposal. Shutter's Shielded Voting is switched on by a Snapshot space admin in Settings, and the permanent version is designed to be enabled proposal by proposal. ENS's thread leaned toward a pilot in its next steward election, with a separate vote on broader use. A holder in an ENS-style DAO may therefore meet both kinds of ballot in the same quarter.

A second house that ignores tokens. Optimism's Citizens' House gives each citizen one equal vote regardless of holdings; LedgerMind credits it with funding 73 public-goods projects against 21 in comparable single-house token-weighted DAOs, and reports that 34% of major DAOs used hybrid mechanisms as of Q1 2026. In such systems, token weight decides some questions and not others.

The market already puts prices on voting weight, and they are not small. Lido pays about $10,714 a quarter to each public delegate who votes consistently and joins the discussion, a deliberate subsidy for the transparent, active voting that passive holders do not supply. On Curve, the price runs the other way: $4.7 million in bribes in one February 2026 gauge round, by LedgerMind's count. LedgerMind also reports that compensating delegates correlates with 3.2x higher participation. Weight that is paid for gets used; weight that is not tends to sit idle.

For a holder with a modest balance, the arithmetic is unforgiving. LedgerMind reports that in MakerDAO in March 2026 the top 10 addresses held 52% of voting power and the top 100 held 89%, leaving 8,847 other voters with 11%. A direct vote from that group rarely decides anything. Its weight pooled in a delegate who votes early, publicly and every time is a different matter, because that is the weight a late swing has to beat, and the weight a timelock gives time to organize. If the DAO under consideration has neither delegation worth the name nor an execution delay longer than its voting window, the small holder's influence is close to zero, and that belongs in the decision to deposit rather than the decision how to vote. The patterns large holders follow are covered in how whales actually vote.