Decentralized autonomous organizations collectively control more than $26 billion in on-chain treasuries as of Q1 2026, according to DeepDAO data. Voter participation in the largest of these organizations averages below 5% of circulating token supply. The result: a small number of wallets — often...
"Decentralization is a performance metric, not a philosophical ideal." — Ifigenia Georgiou, Associate Professor of Finance, University of Nicosia (Forbes, April 2026)
Decentralized autonomous organizations collectively control more than $26 billion in on-chain treasuries as of Q1 2026, according to DeepDAO data. Voter participation in the largest of these organizations averages below 5% of circulating token supply. The result: a small number of wallets — often fewer than 10 — routinely pass proposals governing billions in assets with negligible opposition.
The structural imbalance has moved from theoretical concern to active exploit vector. In July 2026, an attacker spent approximately $4 million to buy BONK tokens, accumulated 99.878% of votes cast on a single proposal, and drained $20 million from BonkDAO's treasury. In August, Binance's security team intercepted a separate governance attack targeting an unnamed DAO's $1.2 million treasury with fewer than 48 hours remaining before execution. Neither attack exploited a smart contract bug. Both exploited governance design.
A peer-reviewed study published in April 2026, analyzing 4,848 Ethereum-based DAOs, found that 81.3% have more than 50% of voting power concentrated in their top 10 token holders. The data raises a question the industry has avoided: whether token-weighted voting, the default governance mechanism across DeFi, is structurally incompatible with decentralized decision-making at scale.
As of March 2026, DAOs collectively manage more than $26 billion in on-chain treasuries, per DeepDAO. The five largest individual treasuries: Uniswap ($4.8B), Sky/MakerDAO ($3.9B), Optimism ($2.1B), Arbitrum ($1.7B), and Lido ($1.4B). More than 5,000 DAOs are tracked across analytics platforms, with 6.5 million governance token holders worldwide.
The composition of these treasuries introduces additional risk. According to CoinLaw data, native governance tokens comprise 67.3% of total DAO treasury assets, while stablecoins account for just 18.2%. A governance failure at any major DAO would simultaneously depress the very asset backing the treasury, creating reflexive downward pressure.
A peer-reviewed study by researchers analyzing 4,848 DAOs on Ethereum — published in April 2026 on arXiv (2604.25959) — found the following:
The study's authors conclude that "governance mechanisms of DAOs themselves systematically reinforce centralization."
On July 6, 2026, BonkDAO — a Solana-based memecoin DAO — lost approximately $20 million in a governance attack. The mechanics were straightforward:
No smart contract was exploited. The voting system operated exactly as designed. The vulnerability was governance architecture: token-weighted voting combined with near-zero participation created a 5x return on the attacker's capital.
On August 18, 2026, Binance's Chief Security Officer Jimmy Su announced the exchange had intercepted a separate governance attack targeting an unnamed DAO. The attacker had submitted a malicious proposal with fewer than 48 hours until execution. According to Su, the threat was one "no external security provider had flagged." Binance coordinated with other exchanges to freeze deposits, and the community voted the proposal down before execution. No funds were lost.
Binance did not name the project, publish the proposal identifier, or provide on-chain transaction records. The incident remains unverified by independent sources.
In April 2025, a separate incident demonstrated the economics of vote buying. An address identified as hitmonlee.eth spent 5 ETH (approximately $10,000) through the LobbyFi platform to purchase 19.3 million ARB in delegated voting rights — worth roughly $6.5 million — and used them to influence Arbitrum DAO's Oversight and Transparency Committee election. The purchased voting power exceeded that held by well-known delegates including Wintermute and L2Beat.
Participation data across major DAOs, compiled by ChainScore Labs, shows structural disengagement:
| DAO | Avg. Turnout (Last 10 Proposals) | Top 10 Voters' Share | |---|---|---| | Uniswap | 4.2% | 62% | | Aave | 2.8% | 71% | | Arbitrum | 1.7% | 85% | | Lido | 0.9% | 58% |
MakerDAO sees less than 1% of MKR holders participate in votes regularly. Proposals routinely pass across major DAOs with fewer than 2% of circulating supply voting.
Uniswap represents a partial counter-example. Since its "UNIfication" governance reform, participation has increased. According to blockful.eth, only one proposal in 2026 failed to reach quorum — the sole such failure across both on-chain and off-chain votes. Participation in UNIfication itself saw 126 million UNI cast in favor, and subsequent votes regularly exceed 70 million UNI, nearly double pre-reform averages. Sensitive proposals — such as fee-switch votes — see higher participation, likely driven by delegate coordination with Uniswap Labs.
The exception does not invalidate the rule. Even Uniswap's improved numbers represent a fraction of its 1 billion total UNI supply.
Delegation — where token holders assign voting power to trusted representatives — emerged as the primary solution to participation apathy. The mechanism introduces its own centralization.
The April 2026 arXiv study found that 20% of DAOs show a 5% or greater voting power disparity between top 10 delegated voters versus direct voters. Gitcoin exhibited the most pronounced case at 19.5% disparity. Only 0x Protocol permits multi-wallet delegation, limiting token holder flexibility.
Top 10 delegates control more than 30% of voting power across major DAOs, according to ChainScore data. In Curve Finance, approximately 70% of veCRV voting power is delegated. The result is a class of semi-professional governance actors who exercise influence disproportionate to their economic stake.
Staking mechanisms compound the concentration problem. The arXiv study documented specific cases where intermediary protocols control majority voting power:
These intermediaries aggregate voting rights from depositors who typically do not participate in the underlying DAO's governance. The staking protocol's team or its own governance process determines how these votes are cast. Individual depositors effectively forfeit governance participation in exchange for yield.
Six DAOs studied use staking with unlock-time requirements; nine use staking with lock-time requirements. Both structures further discourage governance participation by imposing opportunity costs on the act of voting.
The Ethereum Name Service DAO executed what may be the most transparent case of deliberate recentralization in August 2026.
On August 11, ENS token holders approved the "Next Era of ENS DAO" proposal with approximately 70% support. The restructuring:
The vote followed a contentious period in June 2026 when co-founder Nick Johnson self-delegated approximately 3.26 million ENS tokens — representing roughly half of active voting power at the time — and used that power to block the renewal of the existing Security Council.
ENS Labs retains protocol development responsibility, including ENSv2. The DAO retains protocol-level decision authority and majority ENS token holdings. The Foundation committed to annual audited financial statements and quarterly grant updates.
The restructuring amounts to an acknowledgment that pure token-weighted governance could not secure an organization with $65 million in assets. The DAO voted to reduce its own authority.
Several approaches to mitigate governance concentration are under investigation or deployment:
Quadratic voting reduces whale dominance by weighting votes as the square root of tokens held. A June 2026 academic paper (arXiv 2605.18990) titled "Concave is the New Linear" argues that all concave voting mechanisms — including quadratic — are mathematically susceptible to Sybil attacks. The paper concludes that "anti-plutocratic DAO governance" through token-weighting alone is impossible without identity verification.
Conviction voting weights votes by how long tokens are locked, attempting to filter for long-term alignment. Adoption remains limited to smaller DAOs.
Security councils and timelocks — as ENS implemented — add centralized checkpoints that can block malicious proposals. The arXiv study notes these mechanisms "create centralization points and contradict pure decentralization."
Proxy voting for tokenized assets represents a different model. On August 5, 2026, Kraken's parent company Payward partnered with Broadridge Financial Solutions to enable proxy voting for xStocks holders across 500+ tokenized securities. With $35 billion in volume processed and 125,000+ token holders, the platform directs votes through traditional custodial infrastructure — the opposite of on-chain governance.
The DAO governance model faces an empirical contradiction. Organizations built to distribute decision-making power have, by their own data, concentrated it. The top decile of voters controls 76.2% of voting power in a typical proposal, according to the April 2026 arXiv study — a concentration level exceeding that observed in traditional corporate governance.
The economic incentives are clear. Voting requires effort; not voting carries no immediate cost. Delegation concentrates power in a professional class. Staking protocols capture governance rights as a byproduct of yield generation. Low participation makes treasuries exploitable, as BonkDAO demonstrated.
ENS's decision to place $65 million under a five-member board, with a Security Council empowered to veto token-holder votes, may represent the honest endpoint of this trajectory: organizations that call themselves decentralized but govern through centralized structures because the decentralized alternative failed to secure assets at scale.
The $26 billion question is whether the rest of the sector will follow ENS's path explicitly — or continue operating under governance frameworks where 7 wallets can decide the fate of $20 million.