Artificial intelligence is the defining capital expenditure cycle of this decade. Data centers are expanding at unprecedented scale, and at the center of that expansion is GPU hardware — the fundamental compute resource that powers modern AI workloads. Yet the financing infrastructure for GPU acquisition remains stuck in the constraints of traditional lending: slow underwriting timelines, conservative collateral frameworks, and credit cycles that move far too slowly for hardware that depreciates at roughly 20% per year.
USD.AI is a decentralized credit protocol built to solve this structural mismatch. It connects AI infrastructure operators who need capital to acquire GPUs with depositors seeking real yield, using on-chain lending mechanisms, legal tokenization of hardware, and a governance token (CHIP) to set the rules of the system. The project has raised $13 million in total funding, built a stablecoin with approximately $280 million in market capitalization, and processed active GPU-backed loans including a disclosed $26.82 million facility against 576 NVIDIA B300 GPUs.
What Is USD.AI?
USD.AI is a permissionless lending protocol built on Arbitrum, operating on a model it describes as “off-chain organization, on-chain lending.” The protocol enables GPU operators — neoclouds, data centers, and AI compute providers — to tokenize their hardware as collateral and access financing instantly, while depositors mint a synthetic dollar (USDai), stake it for yield (sUSDai), and earn returns generated by real GPU loan interest payments and U.S. Treasury bill reserves.
The long-term vision is what the team calls the “petrodollar of AGI infrastructure” — a financial layer permanently tied to the compute stack that AI depends on, in the same way the petrodollar tied global finance to oil.
Core Architecture: Three Layers
USDai: The Synthetic Dollar
USDai is a fully-backed synthetic dollar, overcollateralized by PayPal’s PYUSD — itself a GENIUS Act-compliant stablecoin backed by U.S. Treasuries and cash equivalents. USDai is designed to maintain a stable $1.00 value and is the primary instrument depositors use to enter the protocol. It is non-yield-bearing by design, maintaining low risk and redemption flexibility.
sUSDai: The Yield-Bearing Wrapper
sUSDai is the staked version of USDai. By staking USDai, depositors gain exposure to the protocol’s two yield sources: interest payments from active GPU infrastructure loans, and returns on idle capital held in U.S. Treasury bills. The protocol targets an APR range of 10–15% for sUSDai. Unlike USDai, sUSDai is not designed as a simple stablecoin — it carries the credit risk of the underlying GPU loan portfolio and operates with redemption queue mechanics (QEV) rather than instant withdrawal.
CHIP: The Governance Token
CHIP is the native governance and utility token of the USD.AI protocol. It controls all core protocol parameters — from collateral standards and loan-to-value ratios to interest rate frameworks, curator approvals, and fee allocation. Key CHIP functions include:
- Governance: CHIP holders vote on protocol decisions, risk parameters, and treasury allocation
- Staking: CHIP can be staked for sCHIP, which earns points and serves as the protocol’s first-loss backstop module — meaning sCHIP holders absorb losses first in the event of borrower default
- Fee exposure: Governance decisions on fee splits and treasury allocation affect the value accrual dynamic for CHIP holders
CHIP does not represent direct revenue sharing or ownership of the protocol. It is an “equity-like” governance layer whose long-term value depends on protocol adoption and loan origination growth.
The CALIBER System: Tokenizing GPUs as Collateral
The CALIBER standard is USD.AI’s proprietary framework for tokenizing GPU hardware as on-chain collateral. The process works as follows: a borrower places a purchase order with an OEM (original equipment manufacturer), secures data center space, establishes an SPV (Special Purpose Entity), takes delivery of GPUs, and receives electronic warehouse receipts from the data center. These receipts are then tokenized as ERC-721 NFTs — representing legal title to the hardware — and posted as collateral in the protocol to access USDC loans. Legal structure includes UCC-1 filings (U.S. secured lending documentation) and escrow via Wilmington Trust, providing off-chain legal enforceability alongside the on-chain loan record.
This system gives depositors real-time on-chain visibility into every active loan facility, including the specific hardware collateralized, its location, and the outstanding balance. A disclosed example: Crucible Capital drew a $26.82 million facility against 576 NVIDIA B300 GPUs through the protocol in April 2026.
Protocol Metrics (as of late April 2026)
- USDai market cap: approximately $280 million
- Total value locked (TVL): approximately $283 million
- Active loans: approximately $60 million
- 30-day revenue: approximately $850,000
- Hub chain: Arbitrum, with omnichain support for Base and Plasma via LayerZero
Funding, Team, and Institutional Backing
USD.AI is developed by Permian Labs, founded in 2021 by David Choi (CEO), Conor Moore (COO), and Ivan Sergeev (hardware systems). Choi and Moore come from investment banking and private equity backgrounds; Sergeev has expertise in hardware systems and compute infrastructure. The combination of traditional finance credit experience and hardware systems knowledge is central to the protocol’s ability to structure and underwrite GPU-backed loans.
Total funding: $13+ million, raised across multiple rounds:
- Series A ($13 million, August 2025): Led by Framework Ventures, with participation from Dragonfly, Arbitrum, Bullish, and others
The Aave V3 Arbitrum governance proposals to onboard USDai and sUSDai (active as of April 2026), and Messari’s publication of a CHIP valuation report in March 2026, signal growing institutional recognition of the protocol within the broader DeFi ecosystem.
Competitive Positioning
USD.AI sits at the intersection of two established DeFi categories: yield-bearing stablecoin protocols and on-chain RWA lending. Its closest comparisons include Ethena (USDe), Ondo Finance (USDY), and Maple Finance — each offering some form of real-yield or credit-based stablecoin product. USD.AI’s differentiation lies in its specific focus on GPU-backed credit as the yield source, rather than delta-neutral crypto derivatives (Ethena), tokenized Treasuries (Ondo), or unsecured institutional crypto lending (early Maple).
As Clifford Chance’s 2026 data center infrastructure report notes, the market is increasingly treating GPU infrastructure as a capital market in its own right, with GPU-collateralized credit structures becoming a recognized institutional asset class. USD.AI is attempting to build the on-chain version of that market.
Risks and Considerations
- GPU depreciation risk: GPUs depreciate at 15–20% annually. If a borrower defaults and collateral must be liquidated, the realized value may be significantly lower than at loan origination
- Borrower cash flow opacity: The protocol does not publicly disclose full details of borrower revenue streams or offtake agreements. Investors cannot independently verify that borrowers have the cash flows to service their loans
- Redemption queue risk: sUSDai operates with async redemption mechanics. During periods of high redemption demand, users may face delays exiting their positions
- CHIP token complexity: CHIP’s value thesis requires understanding multiple layers — synthetic dollar mechanics, GPU credit markets, legal tokenization, and governance dynamics. This complexity may limit broad retail adoption
- Early-stage liquidity: CHIP is an early-stage governance token with relatively low liquidity. Price volatility is significant, and the token’s value is tightly correlated with protocol adoption rather than established cash flows
How to Trade CHIP on KCEX
CHIP/USDT is now available for trading on KCEX. Start trading: https://www.kcex.com/exchange/CHIP_USDT
USD.AI (CHIP) Core FAQ
What is USD.AI?
USD.AI is a decentralized credit protocol that connects AI infrastructure operators (GPU borrowers) with yield-seeking depositors. GPU operators tokenize their hardware as collateral to access on-chain financing; depositors mint USDai or stake for sUSDai to earn yield from GPU loan interest and Treasury bill returns. CHIP is the governance token that controls protocol parameters.
What is CHIP used for?
CHIP is USD.AI’s governance and utility token. Holders vote on risk parameters, curator approvals, fee splits, and treasury allocation. CHIP can be staked for sCHIP, which earns points and serves as the protocol’s first-loss module in the event of borrower defaults.
What is the difference between USDai and sUSDai?
USDai is a non-yield-bearing synthetic dollar backed 1:1 by PYUSD, designed for stability and redemption flexibility. sUSDai is the yield-bearing, staked version that earns returns from GPU loan interest and Treasury bills. sUSDai carries the credit risk of the loan portfolio and has redemption queue mechanics rather than instant withdrawal.
Who developed USD.AI?
USD.AI is developed by Permian Labs, founded by David Choi (CEO, investment banking/private equity background), Conor Moore (COO), and Ivan Sergeev (hardware systems). The $13 million Series A was led by Framework Ventures, with participation from Dragonfly, Arbitrum, and Bullish.
What are the main risks of USD.AI?
Key risks include GPU depreciation reducing collateral value in default scenarios, limited public disclosure of borrower cash flows, sUSDai redemption queue delays, CHIP token complexity limiting retail adoption, and early-stage liquidity resulting in significant price volatility for the CHIP token.
About KCEX
KCEX is a centralized cryptocurrency exchange established in 2021 and registered in the Seychelles. Guided by the values of “users first, security first, and innovation-driven,” KCEX is committed to providing diversified digital asset trading and investment services, covering futures trading, spot trading, and innovative financial derivatives. Our team is composed of experts in blockchain and finance, with extensive experience and technical capabilities to ensure platform stability and the security of user assets.
