USDai

What is USDai?

USDai (often styled as USD.AI) is a synthetic, onchain dollar system that splits “cash-like” liquidity from “credit-like” yield: USDai is designed to trade close to $1 and remain instantly redeemable, while a separate token, sUSDai, is the yield-bearing instrument whose return is sourced from a blend of short-duration U.S. Treasury exposure and collateralized loans to revenue-producing AI/compute and DePIN infrastructure operators.

The core problem it targets is the mismatch between what DeFi wants (highly liquid, composable dollars) and what real-world productive assets look like (illiquid equipment with amortizing cash flows and underwriting risk). USD.AI’s claimed moat is that it makes those infrastructure cash flows legible to DeFi through a structured two-token architecture, standardized underwriting/curation and oracle plumbing, and an explicit redemption/queue design intended to prevent a run on the yield token from forcing disorderly liquidation of illiquid collateral.

The project’s own documentation is explicit that USDai is not the yield instrument and that sUSDai, not USDai, is where infrastructure-credit risk is warehoused. This separation is described in the protocol docs and FAQ, along with the role of Treasuries as an “idle capital” baseline while infrastructure lending scales.

See the project’s documentation and its explanation of USDai vs sUSDai.

In market-structure terms, USD.AI sits closer to the “RWA/credit” corner of DeFi than to mainstream payments stablecoins: it is not competing to be a unit-of-account for retail commerce so much as to be a composable collateral and yield primitive that can be plugged into Arbitrum-native DeFi and, per its own technical materials, bridged more broadly.

As of early 2026, third-party dashboards put the protocol in the mid-hundreds of millions of dollars of TVL, concentrated on Arbitrum, with fee and revenue estimates that resemble a credit intermediation business more than a pure AMM or L1.

DefiLlama’s USD AI protocol page provides a current snapshot of TVL and fee/revenue series, while major price trackers (which can disagree on rank and market cap methodology for stablecoins) list USDai among mid-cap onchain dollars with several hundred million units outstanding, as reflected on CoinMarketCap and CoinGecko.

Who Founded USDai and When?

USD.AI has been developed by Permian Labs, the entity name shown on Arbiscan for the Arbitrum USDai token contract, and the project’s public communications frame the protocol as launching into the 2024–2026 cycle where tokenized T-bills, “RWA” lending, and AI compute financing all grew rapidly in parallel.

The team has used a relatively conventional venture-backed growth path alongside the DeFi-native product rollout: in August 2025 it announced a Series A led by Framework Ventures with participation from several large crypto funds, positioning the raise explicitly around scaling GPU-backed infrastructure finance via a yield-bearing stablecoin model.

That announcement is documented in the project’s own post, “USD.AI Raises $13M to Scale AI Infrastructure”.

The project’s narrative has also evolved in a way that is typical for RWA-credit protocols: early messaging emphasized a high-level “real yield from productive assets” framing, while later materials increasingly specify a structured credit stack (Treasury base layer plus infrastructure loans), governance/legal scaffolding, and mechanisms meant to manage liquidity under stress.

In January 2026 the team announced the creation of the USD.AI Foundation and a planned $CHIP governance token, describing the foundation as an off-chain steward for legal/regulatory and contractual matters and positioning CHIP as the eventual parameter-governance and insurance-staking asset.

That post also underscores the protocol’s attempt to formalize an off-chain/on-chain boundary for credit origination, servicing, and enforcement—an area where “decentralization” claims are often weakest in practice for RWA systems.

How Does the USDai Network Work?

USDai is not a standalone L1 or L2 network with its own consensus; it is an ERC-20 token system deployed on existing smart-contract platforms (notably Arbitrum, per the contract address provided and the chain concentration visible on DefiLlama).

Its security model therefore inherits the underlying chain’s consensus and decentralization properties (e.g., Arbitrum’s rollup security assumptions and Ethereum settlement), plus the correctness and governance of USD.AI’s own contracts and privileged roles.

The onchain component is best described as a vault-and-strategy system that mints and redeems USDai and issues sUSDai as a yield-bearing receipt token, with asynchronous redemption mechanics for sUSDai designed to match the liquidity profile of the underlying strategies rather than promising continuous par withdrawals.

Technically, the project documents sUSDai as an ERC-4626-style vault on deposit and an ERC-7540-style asynchronous redeem flow on withdrawal, including a FIFO redemption queue and a “service redemptions” function that can be called when liquidity is available.

The docs also acknowledge an administrative role that periodically services the queue, which is a material centralization vector even if the queue rules are enforced onchain, because operational latency and discretion over when strategies unwind can affect realized liquidity under stress.

These mechanics are described in the technical protocol overview and the more specific technical overview. For pricing and accounting, USD.AI states it uses Chainlink-based oracle interfaces to convert positions back into USDai terms, and it publicly announced adoption of Chainlink price feeds for USDai/sUSDai rates and USDai/USD market rates. See “USDAI Adopts Chainlink As Its Official Oracle” and the listing on the Chainlink ecosystem directory.

What Are the Tokenomics of usdai?

USDai’s tokenomics are structurally closer to a balance-sheet liability than to a capped-supply cryptoasset: supply expands and contracts with minting/redemption demand, and major trackers list no fixed maximum supply. As of early 2026, third-party sites reported circulating supply in the hundreds of millions of tokens and described supply as effectively uncapped, consistent with a stablecoin-like instrument rather than a scarcity asset.

See the supply fields on CoinMarketCap and CoinGecko. In that sense, USDai is not meaningfully “inflationary” or “deflationary” in the way an L1 token is; its supply is endogenous to the protocol’s collateral and demand for dollars, and the relevant question for risk is not emission rate but the quality, duration, and liquidity of backing assets, plus the legal enforceability of claims on those assets.

Utility and value accrual are also atypical compared with gas tokens.

USDai is positioned as the liquid, non-yield-bearing unit used for composability and secondary market liquidity, while the economic incentive to hold within the system is concentrated in sUSDai, which accrues yield sourced from the strategy set (Treasuries plus infrastructure-credit spreads, net of fees).

The protocol explicitly frames USDai as instantly redeemable and “fully backed” through its chosen stablecoin plumbing, while reserving redemption periods/queues for sUSDai given the less liquid nature of the underlying loans. This split is central to how USD.AI argues it can keep USDai close to peg without forcing all users to take credit-duration risk.

See the protocol’s core docs and the app guide describing buy/redeem and staking.

A further nuance is that USD.AI publicly stated it is “powered by M0,” indicating that USDai functions as an extension on top of M0’s $M stablecoin building block backed by Treasuries, using M0 for shared liquidity and mint/redeem infrastructure; this architecture is described in USD.AI’s own post, “USD.AI is Powered by M^0”, and in M0’s materials on $M and extensions and Build with M.

Who Is Using USDai?

Onchain usage splits into two very different cohorts: (1) traders and liquidity providers using USDai as a dollar leg in DEX pools and as collateral in DeFi, and (2) yield-seekers allocating to sUSDai for returns that are marketed as stemming from Treasury carry plus infrastructure lending.

These cohorts can be correlated in benign regimes, but they behave differently in stress: trading liquidity can vanish quickly, while queued redemption demand for the yield token can spike.

As of early 2026, public token tracker pages show a non-trivial transaction history on Arbitrum and thousands of holders according to major aggregators, but “real usage” should be evaluated by where USDai sits in DeFi (DEX pools, money markets, structured products) and by whether sUSDai growth is accompanied by transparent loan origination and servicing data rather than purely Treasury carry. CoinMarketCap reports holder counts and basic market data on its USDai page, and Arbitrum activity can be inspected directly via Arbiscan.

For institutional or enterprise-facing adoption, the most concrete, non-rumor signals are infrastructure-provider integrations and stablecoin plumbing choices rather than “partnership” headlines.

USD.AI has formally announced its integration with M0 for stablecoin infrastructure and with Chainlink for oracle feeds, both of which are credible ecosystem dependencies with real technical surface area. It has also publicly disclosed venture investors in its Series A, which is not “adoption” per se but does indicate some institutional diligence and a runway for underwriting and legal build-out.

See the Series A announcement on USD.AI’s site. Claims about specific credit facilities and borrower pipelines should be treated as directional unless accompanied by verifiable onchain loan data, audited statements, or enforceable disclosures; the project does discuss facilities and a pipeline in its foundation/governance announcement, but those are still forward-looking. See “Announcing the USD.AI Foundation and $CHIP”.

What Are the Risks and Challenges for USDai?

Regulatory exposure is unusually multi-layered for an onchain dollar because the product straddles stablecoin issuance, yield products, and real-world credit origination secured by physical assets.

Even if USDai itself is framed as a non-yield-bearing, Treasury-backed instrument, the system’s economic center of gravity is sUSDai and the associated credit underwriting, which can resemble a securitized note or fund interest from a regulator’s perspective, depending on jurisdiction, disclosures, and distribution.

The protocol’s move to create a formal foundation and a governance token explicitly meant to influence rates, collateral parameters, and an insurance module is a governance milestone, but it also sharpens questions about who is operating the credit business and what obligations attach to that role.

The project’s own description of the foundation’s remit is in its foundation and CHIP announcement.

Separately, USD.AI’s dependency on privileged roles for redemption servicing and strategy operations, as described in its technical overview, is a centralization vector that can matter in a liquidity event even if smart contracts enforce queue ordering. See the technical protocol overview.

Competitive threats come from both directions: at the “cash” layer, USDai competes with deeply liquid fiat-backed stablecoins and with Treasury-token wrappers that already have entrenched exchange and DeFi distribution; at the “yield” layer, sUSDai competes with the broader universe of onchain credit and RWA yield products that offer varying combinations of liquidity, transparency, and underwriting discipline.

The hardest economic threat is not another stablecoin’s peg but the cyclicality of AI/compute economics: if GPU rental rates compress, counterparty defaults rise, or hardware collateral values gap down, the credit spread that supports sUSDai can evaporate precisely when redemption demand increases.

Mainstream coverage has flagged this “AI boom/bust” sensitivity as an explicit risk case, even while describing the protocol’s attempts at custody, tokenization, and withdrawal management.

See, for example, the third-party write-up in Tom’s Hardware, which, while not a primary financial filing, highlights the model’s dependence on sustained AI compute demand.

What Is the Future Outlook for USDai?

Near-term roadmap items that are actually verifiable from first-party materials revolve around governance formalization and liquidity mechanics rather than a new chain or consensus upgrade.

The January 2026 announcement sets expectations for an ICO/TGE of the $CHIP governance token in Q1 2026 and frames CHIP as the control plane for parameters, curator approval, and an insurance module; if executed, that would meaningfully change the protocol’s decentralization narrative, but it also introduces classic governance-token risks: voter apathy, capture by large holders, and mispricing of tail risk in any insurance backstop.

This plan is laid out in the project’s foundation/CHIP post. On the product side, USD.AI’s own technical materials point to a future auction mechanism for redemption queue position, which is essentially a market-based way to allocate liquidity when redemptions are constrained; the documentation references a move toward an auction in the redemption queue design.

See the technical overview and the broader description of queue optimization concepts in its docs discussing redemption and liquidity design.

Structurally, the main hurdle is whether USD.AI can scale infrastructure lending while keeping USDai’s “cash-like” properties credible, without pushing too much correlated risk into opaque offchain processes.

The protocol’s stated approach—base-layer Treasuries via an M0-powered extension for scalable liquidity, plus gradually increasing allocation to equipment-backed loans as origination matures—can work in principle, but it will be judged by disclosure quality, loss experience through a downcycle, operational robustness of servicing and enforcement, and the degree to which “instant redeemability” is preserved under stressed secondary market conditions.

The credibility of its oracle and accounting layer (e.g., Chainlink feeds for rate and market pricing) is a necessary but not sufficient condition; the binding constraint is ultimately the liquidity and recoverability of the underlying credit book. The project’s own articulation of the M0 integration is in “USD.AI is Powered by M^0”, with M0’s extension model described in the M0 documentation.