DAI is a protocol liability, not an “asset” the protocol tries to make scarce
DAI’s tokenomics are easiest to reason about if you treat it as Maker credit that happens to be packaged as an ERC-20. It is the external representation of the Maker Protocol’s internal debt unit. That framing is not ideological. It is literally how the core DAI module describes it: DAI was designed to represent “any token that the core system considers equal in value to its internal debt unit.”
That single design choice pushes DAI into a different bucket than most Web3 “token economy” assets. There is no fixed supply curve to admire. There is no emission schedule to front-run. There is only a balance sheet that expands when the protocol underwrites more collateralized debt and contracts when borrowers repay and DAI is burned. The entire game is peg stability, solvency, and governance discipline.
Mechanism-design-wise, DAI is a clean separation of concerns. The ledger is deterministic. The monetary policy is not. The core accounting and rate accumulation are implemented in predictable formulas, but the parameters that matter most are adjustable by governance. That trade-off is the throughline of DAI’s tokenomics.
Onchain, DAI exposes standard token metadata: name “Dai Stablecoin,” symbol “DAI,” decimals 18.
Supply: elastic issuance via Vault debt, plus “special mint” paths
DAI is minted and burned through the Maker system’s adapters. The DAI token contract supports Mint and Burn capabilities, and the DaiJoin adapter is the canonical bridge between internal DAI balances in the core ledger (the Vat) and the external ERC-20 you transfer around.
The baseline issuance path is still what most people mean by “Maker.” You lock governance-approved collateral into a Maker Vault and generate DAI as debt against it. The MCD system paper describes Multi-Collateral Dai as allowing users to generate DAI by leveraging collateral assets approved via governance, with risk parameters set per collateral type.
DAI supply also expands and contracts through a few “special” modules that matter for tokenomics because they alter where marginal supply comes from:
Peg Stability Modules (PSMs) act like authed vaults that let users swap between DAI and an external “gem” (often a centralized stablecoin) via sellGem() and buyGem().
Flash mint lets anyone mint DAI up to a governance-set limit, provided they repay in the same transaction and pay a fee. This is supply expansion with a hard atomicity constraint.
From a distribution standpoint, DAI does not have a founder allocation, vesting schedule, or protocol-owned “treasury supply” in the way typical volatile tokens do. The “distribution” of DAI is the distribution of who chose to borrow it, who acquired it in markets, and which integrations accumulated it as working capital. The relevant analytics are balance sheet composition and concentration, not unlock calendars.
Concrete supply data is still useful as a reality check. CoinGecko’s circulating supply data shows that on March 7, 2026, DAI has circulating supply 4,274,257,951, total supply 4,274,257,951, and max supply ∞.
CoinGecko also flags a market-level structural change: DAI “has been rebranded to USDS” in the Sky ecosystem context, and it notes that DAI’s market cap handling differs from global aggregates on their site.
Fees and fiscal flows: stability fees in, savings out, auctions as the balance-sheet backstop
DAI tokenomics are not about rewarding DAI holders by dilution. They are about charging DAI borrowers and optionally paying DAI holders, while keeping the protocol solvent under collateral stress.
Stability fees accrue on Vault debt via a cumulative-rate mechanism. The Maker documentation emphasizes the constraint that fee accumulation must be constant-time relative to the number of Vaults, so the system updates global cumulative rates and multiplies them by normalized balances when needed.
The Dai Savings Rate (DSR) is implemented in the Pot contract. Users deposit DAI to earn savings, and the DSR is explicitly set by governance.
The deterministic part is the rate accumulator math: chi (the cumulative savings rate) evolves as a product of per-second rates, and user balances are represented as pie[usr] so that interest can be applied globally without iterating accounts.
The uncomfortable part, tokenomics-wise, is where DSR “comes from” at the mechanism level. The documentation describes Pot.drip calling Vat.suck such that the additional DAI created for savings is credited to the Pot while the protocol’s unbacked debt tally increases by the same amount.
This is not a flaw. It is a deliberate accounting design. It means DSR is a protocol liability that must be covered by the rest of the system’s net income and buffers over time. In other words, DSR is “real” only to the extent Maker’s broader balance sheet remains net positive.
That brings us to the system stabilizers:
Surplus buffer and surplus auctions. Maker accumulates surplus (from stability fees and other revenue) into a buffer, then auctions surplus DAI for MKR and burns the MKR. The Flapper documentation describes surplus auctions as selling a fixed amount of DAI for MKR, and burning the MKR received.
Even the trigger logic is parameterized. Governance sets the buffer (Vow.hump) and lot sizing (Vow.bump) that gate when auctions can occur.
Debt auctions are the last-resort recapitalization path. The Maker system is designed so that if collateral auctions do not raise enough DAI to cover obligations and the buffer is insufficient, the protocol can trigger a debt auction where MKR is minted and sold for DAI.
From a mechanism-design lens, this is the core equilibrium claim Maker makes: DAI holders get a stable unit because the system can socialize tail losses onto the governance token if needed, while governance token holders are incentivized to avoid that outcome by choosing prudent collateral and parameters.
Peg mechanics: arbitrage is the engine, liquidations are the enforcement
The peg is not maintained by a single lever. It is maintained by a set of arbitrage rails, each with different latency and different trust assumptions.
Vault minting and repayment is the “slow but principled” rail. If DAI trades above target, creating DAI by borrowing against collateral becomes attractive. If DAI trades below target, repaying debt to extinguish DAI becomes attractive.
PSMs are the “fast rail.” Mechanically, they support trading ERC-20 DAI for an external stablecoin (or vice versa) through sellGem() and buyGem().
Mechanically, that’s powerful. Economically, it is a trade. The PSM tends to improve peg responsiveness, but it can also increase the portion of DAI supply that is effectively intermediated by a centralized asset, depending on which gems are active and how governance sizes and prices the module. DAI becomes more stable by leaning on collateral that is less reflexive than volatile crypto. It also becomes more exposed to asset-level constraints that are not enforceable on Ethereum.
Liquidations are the enforcement path that keeps DAI solvent, which is upstream of peg stability. Modern Maker liquidations (Liquidations 2.0) are documented across the Dog and Clipper modules. In 2.0, calling Dog.bark takes all the Vault’s debt into liquidation (subject to global and per-collateral limits), and auctions can sell collateral in partial fills.
The liquidation pipeline has explicit throughput constraints. Liquidations 2.0 introduces a global limit Hole and per-collateral limits ilk.hole that cap the total DAI needed to cover the debt and penalty of active auctions.
Oracle latency is part of the design, not a footnote. The docs note that the initial auction price reads from the Oracle Security Module (OSM) and that the OSM price is delayed roughly one to two hours relative to the market.
When people say “DAI is overcollateralized,” they are implicitly assuming liquidations clear in stressed markets. The actual guarantee is weaker and more honest: Maker has a specified liquidation mechanism with bounded throughput, governance-tunable incentives, delayed oracle inputs, and an MKR backstop if the system ends up with bad debt.
History that changed the mechanism: SCD → MCD, then Maker → Sky
DAI has had at least two structural regime changes that matter for tokenomics modeling.
Single-Collateral Dai to Multi-Collateral Dai. The upgrade to Multi-Collateral Dai became reality in November 2019, expanding collateral beyond ETH and formalizing governance-set risk parameters per collateral type.
Maker’s transition into the Sky ecosystem. The Sky ecosystem launched in 2024, introducing token upgrades including USDS as the upgraded version of DAI and SKY as the governance token upgrade. For a contrasting stablecoin mechanism, compare this with our USDe tokenomics review.
This matters because “DAI tokenomics” now include an explicit migration and parallel-asset story. DAI can remain in circulation, but governance and product focus can shift marginal demand to USDS via new features and incentives.
Sky’s USDS documentation describes a DAI↔USDS converter that performs 1:1 conversions, with no fees, and states that fees cannot be enabled on this route in the future. That last clause is rare and important because it is a hard constraint that reduces governance surface area for that conversion path. USDS converter docs
Governance: DAI’s monetary policy is discretionary, executed through deterministic spells
DAI holders do not directly govern DAI. MKR (and in the Sky framing, SKY) governs the protocol that issues it. Maker’s governance model is onchain and parameter-driven, but it is not algorithmically self-adjusting in the way purely rule-based stablecoins aspire to be.
The Maker governance model includes proposal polling to establish sentiment and Executive Voting to approve changes to system state, implemented through smart contracts that execute a proposal’s actions once.
From a mechanism design perspective, this governance layer is the dominant “policy API” for DAI tokenomics. It controls:
1) which collateral types can mint DAI and at what ceilings,
2) stability fees and savings rates,
3) liquidation parameters and circuit breakers,
4) PSM configuration and capacity,
5) surplus buffer sizing and auction cadence.
This is both a strength and a modeling headache. Governance adaptability allows Maker to respond to market structure changes. It also means DAI does not have a fixed, credible monetary constitution. Parameter stability is social, not cryptographic. If you want a quick terminology refresh before modeling governance-dependent flows, the tokenomics FAQ covers the common moving parts.
If you are integrating DAI into a broader system, treat governance parameter variance as a first-class dependency, not “background protocol risk.” For teams doing tokenomics design services or token economy design around stablecoin-denominated flows, the cleanest approach is to build explicit buffers and kill-switches keyed to observable Maker indicators rather than assuming DAI behavior is stationary.
Risk register: where the mechanism strains
DAI is a mature design with real battle scars. Its risks are not mysterious. They are measurable and mostly sit at the boundary between deterministic execution and discretionary policy.
Top 3 risks
-
Collateral regime drift (dominant systemic risk)
Trigger: Governance meaningfully increases reliance on collateral whose risk is not enforceable by onchain liquidation (for example, centralized stablecoins via PSMs, or governance-approved real-world-asset strategies).
Mechanism: DAI’s solvency becomes increasingly sensitive to offchain constraints and administrative actions (freezes, blacklists, redemption gates, legal process) that the Maker liquidation engine cannot “price in” or arbitrate away. The peg can remain tight right up until it fails discontinuously because the failure mode is binary, not a smooth price move.
Who bears it: DAI holders and DAI-denominated integrators first (loss of peg, impaired liquidity), then governance token holders via recapitalization paths if bad debt emerges.
Measurable indicators: Growth in PSM usage and capacity (amount of DAI supply effectively minted via PSM rails), frequency of governance changes to PSM parameters, and collateral mix shifts toward assets whose downside is dominated by administrative rather than market dynamics. -
Liquidation throughput and oracle-latency stress
Trigger: Rapid collateral drawdowns combined with constrained auction capacity (global Hole, per-collateral ilk.hole), high gas, or thin DAI bid-side liquidity.
Mechanism: Under stress, liquidation initiation and auction clearing can lag. The system uses delayed oracle prices for auction initialization, which can be stale relative to fast markets. That increases the chance of inefficient clears and bad debt when volatility spikes.
Who bears it: Vault owners (forced liquidation penalties and slippage) and, in the tail, DAI holders and MKR/SKY holders if deficits propagate to protocol debt and recapitalization paths.
Measurable indicators: Utilization of Dirt versus Hole (auction backlog pressure), number of active auctions, and frequency of governance circuit-breaker actions or emergency parameter changes during volatility. -
Governance policy volatility (rate and buffer whiplash)
Trigger: Rapid changes in stability fees, DSR, surplus buffer sizing, or auction configuration in response to market or political pressure inside governance.
Mechanism: Even with deterministic execution, parameter discontinuities can cause second-order effects: leveraged positions unwind abruptly, DAI demand shifts between holding and borrowing, and integrators who assumed slow-changing policy get surprised. Running auctions can be affected by parameter changes at any time.
Who bears it: Vault users (rate changes on outstanding debt), DAI holders using savings modules, and DeFi protocols with DAI as a base asset for pricing or accounting.
Measurable indicators: Frequency of Executive Votes affecting core parameters, magnitude of DSR/stability-fee deltas, and sustained deviations of DAI from peg that correlate with governance inaction or delayed response.
Dominant risk: Collateral regime drift is the one that breaks the clean mental model most builders carry. DAI’s original story was crypto-collateralized credit with liquidation-enforced solvency. The modern system adds rails that are excellent for peg tightness and capital efficiency, but they quietly shift the nature of the guarantee.
Liquidations are an onchain enforcement mechanism. They work best when the bad outcome is “price moved against you” and the remedy is “sell the asset.” That is exactly what Dog and Clipper are engineered to do, with explicit throughput limits and auction mechanics.
Administrative risk does not clear in an auction. If a collateral asset can be frozen, redeemed at par only under conditions, or impaired by offchain enforcement, then the liquidation engine is solving the wrong problem. Governance can respond after the fact. It can change parameters, deprecate modules, and recapitalize through debt auctions if needed.
But those are discretionary actions with coordination latency. That latency is the risk. It is why a deterministic engineer should be skeptical of treating DAI as a fully rule-governed stable unit. The execution is deterministic. The underwriting standards are not. As DAI’s marginal supply increasingly comes from “special mint” rails like PSMs, the system trades liquidation-enforceable risk for policy-manageable risk. That can be a rational design choice. It is also the main reason you should monitor collateral composition and governance posture, not just the $1 chart.
This article is part of our Tokenomics Deep Dive series.
