How Whales Swing Investment DAO Votes

A whale does not swing a DAO vote by owning most of the tokens. It swings the vote by owning most of the votes that are actually cast, and Compound's Proposal 289 shows how small that share can be. The proposal moved 499,000 COMP (about $24 million, roughly 5% of the treasury) into a yield vehicle controlled by a group called the "Golden Boys", and passed 682,191 to 633,636. According to CryptoSlate's reconstruction, 563,591 of the yes votes, 82% of the total, arrived in the final 34 minutes. The decisive block landed eight minutes before close.

The mechanism applies to every token-weighted vote, but it matters most in investment DAO votes, the ones that allocate reserves, fund programs or buy tokens with treasury money, because that is where a swing converts directly into dollars. The conclusion of this page is simple, if unwelcome: total holdings are rarely the deciding variable. Turnout and timing are.

A proposal is measured against the votes that show up, and the votes that show up are few. In a study of 48 Ethereum DAOs cited by CryptoSlate, the average share of supply registered to vote was about 21%, and the ten largest holders controlled a majority of voting power in 39 of the 48. Cointelegraph's survey of ten DAOs found that passing a proposal typically requires 1 to 4% of token supply, and that only 1 in 10,000 to 1 in 30,000 holders hold enough to pass one alone.

Delegation, meant to widen participation, mostly narrows it. An academic study of delegation networks found that 16 delegates hold more than half of Uniswap's voting power and 19 addresses control over half of Compound's tokens, with voting-power Gini coefficients of 0.99 in both. The details of how voting power is counted vary by protocol. The concentration does not.

Timing does the rest. Tokens can be bought or borrowed, delegated, and voted inside a single window. The Proposal 289 wallet started with 853 COMP; over four months it acquired 563,790 COMP on exchanges and borrowed 118,089 from Compound itself. Arriving late matters because opposition needs time to organize. SpendNode's account notes that the opposing delegates, among them Wintermute Governance, StableLab and two university blockchain clubs, technically had enough power to block the measure and simply ran out of clock.

Compound, Proposal 289. The vote above. COMP fell about 7% within 24 hours, and Compound later settled by cancelling the allocation and adding a veto role. The proponent, known as "Humpy", is alleged to have made similar plays at Balancer in 2022 and SushiSwap in March 2024.

Neutron, September 22, 2026. GoPlus Security's breakdown shows an attacker who had voted with about 100 NTRN spending roughly $20,199 in USDC on 31.62 million NTRN, delegated 11 minutes before the tally. A flaw in Neutron's wasmd governance module let that late power count. The proposal granted admin control of Astroport and Drop Money contracts, and about $9.4 million left. Staked NTRN worth around $113,000 had been guarding $9.4 million, a ratio of roughly 83 to 1. The Cosmos Hub was halted for about 24.5 hours; 1,227,121 ATOM was recovered into a multisig, and about 500,000 ATOM swapped through THORChain is considered unrecoverable.

The Compound Foundation, Proposals 580 and 582. Here the whale was an insider. According to delegate ugurmersin's complaint, the Foundation converted 8.42 million DAI of reserves into 344,780 COMP and delegated it 58 minutes before the snapshot. Without those votes, supporters of Proposal 582 would have held 45.1%; with them, just over the 50% needed. The Foundation replies that the tokens stayed DAO-owned. Reports differ on the dollar value of the purchase.

A swung vote is not an exploit in the usual sense. Of Proposal 289, Aleksandar Dukic wrote: "Nothing was hacked. No smart contract was tricked." The proposal cleared the threshold under the rules as written. Neutron is the hybrid: the contracts worked, but a governance-module flaw let power bought minutes earlier count.

The strongest objection is that large holders have the most to lose and often vote to protect the protocol. That is true. At ENS, co-founder Nick Johnson's single vote drove an 82% rejection of a Security Council renewal, on the stated view that the council should be "a narrow emergency defence mechanism". At Aave, Wintermute's founder voted against the brand-ownership proposal. Large holders are not uniformly predatory.

The objection rests on stake that lasts, though, and the costly cases lacked it. One wallet borrowed its weight from the protocol it was voting to drain. Another held its weight for 11 minutes. Alignment through ownership cannot be relied on when the ownership can be rented by the hour.

Four checks follow from the mechanism.

• Is there a quorum? BonkDAO had none and no timelock. An attacker who bought about 1% of supply for $4 million cast 99.878% of the votes and took $20 million.
• When is voting power fixed? If tokens acquired during a vote can still count, the price of control is the price of a late purchase.
• What stands between a passed vote and the money? A timelock helps but is not enough on its own. Term Finance had seven days and still lost $8.5 million to an attacker who spent about 2 ETH. A veto restores safety at the cost of a trusted body, which CryptoSlate notes recreates the centralization DAOs were built to avoid.
• Who usually turns up? At Lido, public delegates vote nearly 100% of the time and have been present for every quorum achieved. Turnout from a known group of delegates is a defense. Its absence is an opening.

The figure worth working out is not how much the largest holder owns. It is what it would cost a stranger to outvote everyone who normally votes, and how long the protocol gives everyone else to notice.