A DAO treasury is the pool of on-chain assets a decentralized organization collectively owns, held by a smart contract rather than a company bank account and deployable only when a governance vote authorizes it. Treasury management is the practice of keeping that pool solvent, diversified, and productively deployed – arguably the single most consequential operational discipline a DAO has, because a treasury is what turns a voting community into an organization that can actually pay for things. Aggregated across the ecosystem, DAO treasuries hold billions of dollars; the discovery hub DeepDAO tracked assets-under-management, spending, and participation across roughly 14,000 organizations until its public site went offline in mid-2026.
The defining risk of a DAO treasury is that most of it is usually a single, highly correlated asset: the DAO's own governance token. a16z crypto's treasury management guide reports that roughly 85% of DAOs hold their entire treasury in one crypto asset, typically their native token. That figure flatters a DAO in a bull market and traps it in a bear one: a treasury that is nominally worth hundreds of millions can only be spent by selling the very token whose price the sale depresses, so the assets are worth least exactly when the DAO most needs to spend them.
The largest DAO treasuries make the risk concrete: Mantle holds much of its multi-billion-dollar treasury in its own MNT token. The standard remedy is diversification – converting a portion of native-token holdings into stablecoins and blue-chip assets so that operating expenses are funded from assets uncorrelated with the DAO's own market.
How that conversion is executed is a decision in its own right. Selling a large native-token block on the open market moves the price against the seller, so DAOs generally reach instead for OTC placements and market-maker loans – an off-market sale to named buyers, or tokens lent to a trading firm as quoting inventory. Both are voted in public, and both carry costs that the proposal format tends to hide.
Mature treasury policies borrow the language of corporate treasury desks and rank objectives in order: capital preservation first, then liquidity, then income. The a16z guide recommends keeping at least 12 – ideally 18 – months of operating expenses in accessible cash or stablecoins, sized against a realistic model of monthly burn – whose largest line is almost always contributor compensation – under conservative revenue assumptions. Runway discipline is what separates DAOs that survive a multi-year downturn from those that ratify an ambitious budget at the top of a cycle and run out of spendable assets before delivering it. Some DAOs go further and pursue protocol-owned liquidity – owning the market depth that backs their own token instead of renting it, a model OlympusDAO pioneered through bonding.
Because no individual may unilaterally move the funds, a treasury is held either in an m-of-n multisig – most commonly a Safe – or directly in a governance contract that releases assets only on a passed vote. Spending follows the DAO's governance process: a proposal names the recipient and amount, the membership votes, and an approved transfer executes on-chain. Frameworks such as Aragon bundle treasury permissions, voting, and payments together; analytics platforms such as Dune and Boardroom give members a public, real-time view of what the treasury holds and how it is being spent – the size, composition, and runway figures that feed a DAO's overall health metrics.
That radical transparency – every balance and every disbursement visible on-chain – is both a DAO treasury's greatest accountability advantage and a standing invitation to governance attackers. Because the DAO and the recipient of a transfer read the same on-chain record of it, neither keeps a private version to reconcile against the other's, which is the property Ian Grigg designed triple-entry accounting to provide; the invoices, expense categories and statutory accounts around a transfer are still kept off-chain. Large treasuries increasingly delegate day-to-day allocation to a non-custodial manager such as karpatkey, granted scoped permissions over a DAO-owned Safe via the Zodiac Roles Modifier – it can deploy funds into pre-approved strategies but can never withdraw them to itself.
That arrangement is one instance of a broader pattern: authority the DAO grants once and does not re-vote, bounded by scope, magnitude, direction and a cooldown rather than by a ballot. Standing authorizations and delegated mandates reads four production examples at contract source and sets out which bounds they share and which they omit.
Custody is often split on purpose. A DAO with separate budgets, such as payroll, grants, rewards and reserves, may hold each in its own account with its own signers, which makes the treasury harder to read as one balance sheet. Aragon's Linked Accounts, released in March 2026, join such accounts into one organisation view without merging them (Aragon). Two accounts link only when each acknowledges the other on-chain, which stops an outsider attaching itself to a DAO's page. No new contract is deployed, each account keeps its own executor and permissions, and linking gives neither account any control over the other. The app sums balances across linked accounts and offers a proposal only the actions its account is permitted to take.
On Radix the multisig is part of the account itself rather than a separate contract such as a Safe. The Radix DAO's first treasury account is a native account whose owner rule requires two of three signature badges, and in August 2026 it held 200 $XRD and the master badge of the Consultation app the DAO votes through; the share of its runway to hold in $XRD rather than stablecoins had not been decided. radix.wiki tracks the decision on its custody card.
A large, transparent, vote-controlled treasury is a prize, and several of the most-studied governance attacks target it directly: an adversary who can cheaply acquire or borrow a voting majority can pass a proposal that drains the treasury to themselves. This tension has pushed treasuries toward two safeguards. The first is timelocks and vetoes that delay execution long enough for the community to react. The second is redemption rights – a floor that lets members reclaim their pro-rata share rather than be captured. In June 2026 GnosisDAO passed GIP-151 – 157,749 GNO for against 2,500, 161,740 GNO in total on a 75,000 GNO quorum – authorizing GNO holders to redeem for a pro-rata slice of the treasury.
On the proposal’s own worked example the redeemable base was $156,420,464 rather than the ~$228M headline treasury, because the DAO’s own GNO and Gnosis Ltd’s are excluded from both sides of the ratio: 1,364,058 GNO eligible, ~$114.67 of net asset value each. It is the largest live test yet of treasury redemption as a governance backstop (GIP-151, Snapshot record). By contrast, DAOs with the largest treasuries – Arbitrum and Optimism among them – lean on delegate structures and grant frameworks to deploy capital without exposing the whole balance to a single vote (Arbitrum forum, Optimism Collective).
A caper inverts the usual treasury story: instead of raising a lump sum and then worrying about custody and runway, its treasury is a protocol-controlled vault that fills continuously from activity. Every buy and sell on the bonding curve pays a skim, a share of the trade that is 100% on an empty curve and falls as the supply sells, and 49.5% of it goes into the caper's own treasury in XRD; proposal and vote fees accrue there too – there is no signer set to compromise and no multisig to socially engineer.
Capital is deployed only through governance: a PAYOUT proposal transfers assets to a recipient, and an INVEST proposal buys into another caper's curve and books the resulting tokens back into the treasury, so a caper can diversify into other capers by vote. And the exit right is built in rather than bolted on: any holder who has cast a ranked ballot can redeem for a pro-rata share of the treasury, weighted by their canonical vote weight, so majority-capture of the treasury is bounded by every other member's standing option to walk away with their share. (Verified against the Caper contract, not aspirational.)