USD.AI (CHIP) Project Report

Project Report

I -update2026-08-21

22.4K

I. Project Overview

USD.AI is a permissionless decentralized lending protocol focused on providing financing services for AI infrastructure. The protocol allows GPU operators to tokenize their hardware assets as collateral and access financing instantly. The core token of the protocol is CHIP, an ERC-20 standard token deployed on the Arbitrum chain, with cross-chain circulation on Ethereum and Base supported through LayerZero's OFT standard. CHIP has a total supply of 10 billion tokens, with approximately 2 billion tokens (20% of the total) in circulation at the time of writing. CHIP began trading on April 21, 2026, and is currently listed on the spot markets of major exchanges including Binance, Coinbase, Upbit, Bybit, OKX, KuCoin, Gate, MEXC, and Robinhood, with perpetual futures trading available on Binance, OKX, Bitget, and Hyperliquid. The protocol is developed by the Permian Labs team, and the USD.AI Foundation was formally established in January 2026 as the off-chain governance entity for the DAO. Investors include prominent institutions such as Coinbase Ventures, DCG (Digital Currency Group), Dragonfly, and Framework Ventures, and the project has been selected for the NVIDIA Inception program.

II. Project Introduction

USD.AI positions itself in the "InfraFi" sector. Its core narrative is that the AI industry is undergoing an unprecedented capital expenditure cycle. GPU hardware has become the core productive infrastructure asset of the AI economy. It generates predictable and recurring cash flows. However, traditional credit markets cannot match the hardware refresh cycle of GPUs. GPUs depreciate by approximately 20% annually, with an effective useful life of roughly three years, while traditional bank credit approval cycles last 12 to 24 months. Even flexible private credit funds require 6 to 12 months to complete underwriting. This structural mismatch means that by the time traditional financing is in place, the GPU hardware may already be past the majority of its lifecycle.

USD.AI seeks to solve precisely this problem. The protocol creates a two-sided credit marketplace: borrowers (AI infrastructure operators, including emerging AI cloud providers and data center operators) obtain financing using GPU hardware and its generated cash flows as collateral; depositors (capital providers) earn yields derived from interest income generated from GPU-backed loans by depositing stablecoins, rather than from token emissions or trading fees. Since its launch in 2025, the protocol has executed over $225 million in loans according to official disclosures, with approved financing exceeding $1.2 billion. Borrowers include publicly traded emerging AI cloud providers and institutions across North America, Europe, Australia, Latin America, and Asia. To date, total protocol deposits stand at approximately $348 million, with over 74,000 active users and an active loan pipeline of roughly $236 million.

The project analogizes itself to the "Fannie Mae moment" for the GPU credit market—just as the housing market only truly achieved scale after the emergence of tradable mortgage-backed securities, the AI infrastructure market also requires a liquid debt product to define its benchmark interest rate. USD.AI's goal is to become the on-chain benchmark rate for GPU-collateralized credit.

III. Product & Technology

USD.AI's product system revolves around three core layers: the deposit layer, the lending layer, and the governance layer.

At the deposit layer, the protocol offers two core assets. The first is USDai, a synthetic dollar fully backed by PYUSD (PayPal's regulated stablecoin, with underlying assets of U.S. Treasuries and cash equivalents), maintaining a 1:1 peg to the U.S. dollar. Users mint USDai by depositing PYUSD. USDai itself does not generate yield and is positioned as a stable unit of account and medium of exchange. The second is sUSDai, the staked version of USDai and the protocol's core yield instrument. Users stake USDai to receive sUSDai, with yields derived from two sources: loan interest paid by GPU infrastructure borrowers, and returns from idle reserves invested in U.S. Treasuries. Yields accrue automatically as the exchange rate between USDai and sUSDai increases. The protocol currently offers a yield of approximately 7.09%, with a projected yield of approximately 13.51%.

At the lending layer, AI infrastructure operators apply for financing through GPUloans.com. Loans feature a non-recourse structure (with recourse limited to the collateral itself, though elastic recourse against the borrower's corporate entity exists in cases of fraud or malicious conduct), with a loan-to-value (LTV) ratio of 70%-80%, repayment cycles every 30 days, and principal amortized over 3 years. For each loan, the borrower operates through a Delaware bankruptcy-remote SPV, with GPU assets legally isolated from the operator's overall balance sheet. Complete legal documentation includes loan and security agreements, UCC-1 financing statements, SPV agreements, pledge agreements, limited guarantees, and data center lien waivers. The loan process is divided into four stages: escrow establishment and capital injection (with Wilmington Trust serving as the independent escrow agent), OEM server build and shipment, hardware installation and verification, and escrow release and capital routing.

In terms of technical architecture, USD.AI employs a hybrid on-chain and off-chain structure. The off-chain layer provides real-world legal enforceability: perfected liens, SPV structures, bank accounts, escrow, and insurance. The on-chain layer provides transparency and programmability: loan NFTs (ERC-721) represent collateral positions, on-chain records serve as the authoritative lender registry, and smart contracts automate payments and reserve extractions. The two layers synchronize at every significant event node. sUSDai is implemented based on the ERC-4626 standard (deposits) and the ERC-7540 standard (redemptions), with redemptions utilizing a 30-day cycle FIFO queue mechanism. The protocol uses Chainlink as its price oracle and LayerZero for cross-chain token transfers.

The protocol also features several innovative mechanisms. CALIBER (Collateral Architecture for Legally Interoperable, Bankruptcy-remote, Enforceable Rights) is its legal and tokenization framework, transforming physical GPU hardware into on-chain programmable collateral through a four-layer structure: the physical asset layer (UCC Article 7 bailment agreements), the SPV isolation layer, the ERC-721 token layer, and the protocol integration layer. QEV (Queue Extractable Value) is its market-driven redemption mechanism, employing a 30-day cycle and ZK privacy bid auctions to convert withdrawal pressure into an orderly, incentive-aligned process, eliminating bank run risk while allowing remaining depositors to earn additional yields from auction fees.

On the security front, the protocol has completed multiple audit rounds, including reports from Cantina (May 2025), ktl (May, October, November 2025, and March 2026), and Quantstamp (February 2026), and maintains an ongoing Bug Bounty program on the Cantina platform. Risk mitigation measures include: a 70%-80% LTV ratio providing an equity cushion; a Debt Service Reserve Account (DSRA) covering approximately 3 months of peak debt service; property insurance provided by Alliant; hardware value reinsurance provided by Barkr; real-time hardware monitoring provided by Aravolta; and ITAD partner capabilities for hardware recovery and resale in default scenarios.

IV. Tokenomics

CHIP has a total supply of 10 billion tokens. Ecosystem Bootstrapping accounts for 27.5%, used to bootstrap protocol liquidity, covering the two core objectives of yield origination and capital formation. Of this, the first 10% was distributed in Season One (The Allo Game), with the remainder to be used for airdrops, incentive programs, and other growth initiatives.

Reserves account for 19.5%, allocated for future grants, partnerships, and R&D.

The Core Contributors allocation is for the Permian Labs builders and operations team. The vesting structure is: 0% unlock for the first 12 months, a one-time 33% unlock at month 12, with the remaining 67% unlocking in equal monthly installments over the subsequent 24 months.

The Investors allocation reflects early supporters' CHIP commitments. The unlock structure mirrors that of the core contributors: 0% unlock for the first 12 months, 33% unlock at month 12, with the remaining 67% unlocking in equal monthly installments over the subsequent 24 months.

CHIP's functional positioning spans two dimensions: governance and staking. On the governance side, CHIP holders vote on-chain to determine core protocol parameters, including collateral standards (which GPU hardware qualifies, LTV ratios, depreciation schedules), interest rate parameters (base rates, tiered rate adjustments), loan underwriting standards (minimum collateral coverage ratios, acceptable offtake contract structures, geographic restrictions, borrower qualifications), protocol fee parameters (origination fees, net interest margin, QEV fees), and protocol upgrades and integrations. On the staking side, CHIP can be staked as sCHIP. The staking module serves as the protocol's backstop—in the event of a loss incident, staked CHIP may be used to cover the shortfall, making stakers active participants in the protocol's risk framework. Unstaking requires passing a cooldown period set by governance.

The protocol's revenue model is based on loan origination fees and net interest margin.

V. Team & Investors

USD.AI is developed by Permian Labs. According to publicly available information, David Choi serves as CEO of USD.AI, and Ivan Sergeev serves as CTO of Permian Labs. Official documentation notes that Ivan Sergeev witnessed the challenges of aligning early hardware financing with operational scalability during his time at 21.co (a company later transformed under the leadership of Balaji Srinivasan and Lily Liu). Permian Labs' GitHub organization is named metastreet-labs, suggesting a possible connection or evolutionary relationship with the former MetaStreet protocol (an NFT lending protocol).

The USD.AI Foundation was formally established on January 27, 2026, as the off-chain management entity for the USD.AI DAO, responsible for legal infrastructure, DAO treasury custody, and ecosystem coordination. The Foundation's actions are guided by token holder governance and it does not directly operate the protocol.

On the investment side, institutional backers displayed on the official website include Coinbase Ventures, DCG (Digital Currency Group), Dragonfly Capital, and Framework Ventures. Additionally, the project was selected for the NVIDIA Inception program and established a deep partnership with PayPal (PYUSD), with PayPal providing the protocol with a PYUSD incentive program of up to $1 billion in scale at a 4.5% annualized rate. Disclosed borrower partners include QumulusAI, Sharon AI, and Quantum SKK, with approved financing totaling over $1.2 billion.

VI. Roadmap

According to publicly available official information, USD.AI's development history and plans are roughly as follows:

In 2025, the protocol officially launched operations, completing the origination and execution of the first GPU-collateralized loans, with sUSDai annual trading volume exceeding $7.7 billion. The same year saw the completion of multiple smart contract audit rounds by Cantina and ktl, and the launch of the first season incentive program, The Allo Game.

In January 2026, the USD.AI Foundation was formally established and the CHIP token was officially announced. In Q1 2026, the ICO, airdrop, and TGE were completed, and the Quantstamp audit was finalized. On April 21, 2026, CHIP officially began trading, simultaneously launching Flatiron (Allo Game Season Two), with all rewards distributed through CHIP airdrops, running until October 14, 2026.

In terms of near-term plans, the sCHIP staking function has already launched. The team has indicated that additional utility features for CHIP will be added in the coming weeks. On the lending business side, the focus is on scaling execution—expanding the borrower base, deepening institutional partnerships, and scaling revenue backed by physical compute infrastructure. Direct minting and redemption of USDai will be restricted to KYC-verified whitelisted market makers and institutions in Q2 2026 (though holding, transferring, staking, and secondary market trading of USDai and sUSDai will remain permissionless). The full auction functionality of the QEV mechanism is also under development.

VII. Risks & Opportunities

On the risk side, first is GPU hardware depreciation and technology iteration risk. GPU hardware depreciates by approximately 20% annually, and AI chip technology iterates extremely rapidly. The launch of new-generation hardware could accelerate the decline in value of older models used as collateral, affecting loan collateral coverage ratios. Although the protocol mitigates this through Barkr's value reinsurance and conservative LTV ratios, in extreme technology leapfrog scenarios, systemic declines in collateral value remain a risk requiring attention.

Second is liquidity mismatch risk. sUSDai yields are derived from GPU loans with 3-year amortization periods, making them inherently illiquid assets. While the QEV mechanism is designed to address redemption pressure, in extreme market conditions, the 30-day redemption cycle could leave depositors facing extended wait times. The protocol has explicitly stated it will not liquidate loans early to meet withdrawals.

Third is borrower concentration risk. The currently disclosed major borrowers are QumulusAI, Sharon AI, and Quantum SKK, with $1.2 billion in approved credit concentrated among a small number of institutions. Should a major borrower encounter operational difficulties or default, the impact on the protocol could be significant.

Fourth is regulatory and compliance risk. The protocol involves multiple regulatory-sensitive areas including real-world asset tokenization, cross-border lending, and stablecoin issuance. While the protocol employs compliance measures such as SPV structures and KYC/KYB verification, regulatory policies across global jurisdictions are still rapidly evolving, creating uncertainty.

Fifth is smart contract and technical risk. Although the protocol has completed multiple audit rounds and maintains a Bug Bounty program, the complexity of the hybrid on-chain and off-chain architecture increases the potential attack surface. Multi-signature operations, cross-chain bridges, and oracle dependencies all present technical risks.

Sixth is token economic model risk. CHIP's current circulating supply represents only 20% of the total. Future token unlocks for core contributors and investors (beginning after 12 months) could create significant sell pressure. Furthermore, CHIP does not confer direct rights to protocol revenue distribution; its value capture logic relies primarily on governance rights and the staking backstop function, leaving its long-term value accrual model largely unproven.

On the opportunity side, first is structural growth in the sector. AI infrastructure capital expenditure is in a historic expansion cycle, with massive and continuously growing demand for GPU financing. The traditional financial system has clear structural disadvantages in this domain, creating an entry point for on-chain credit protocols.

Second is the real yield narrative. Unlike most DeFi protocols that rely on token incentives, USD.AI's yields are derived from interest income generated from GPU-backed loans and U.S. Treasury yields, providing strong differentiated competitiveness in the current market environment.

Third is the institutional-grade partner network. Support from top-tier institutions such as Coinbase Ventures, DCG, and Dragonfly; deep integration with PayPal PYUSD (including a $1 billion-scale incentive program); and selection for the NVIDIA Inception program provide the protocol with strong resource backing and business development capabilities.

Fourth is broad DeFi ecosystem integration. USDai and sUSDai have been integrated into mainstream DeFi protocols including Pendle, Euler, Morpho, Aave, Curve, Balancer, Fluid, and Gearbox, providing strong composability and liquidity foundations.

Fifth is first-mover advantage. The GPU-collateralized on-chain credit market remains in its early stages, and USD.AI has established certain barriers in legal architecture, underwriting processes, and partner networks.

VIII. Conclusion

USD.AI (CHIP) is a decentralized lending protocol positioned in AI infrastructure financing. By tokenizing GPU hardware as on-chain collateral, it has constructed a bilateral credit market connecting AI operators' financing needs with DeFi capital supply. The protocol presents a highly integrated product architecture, fusing the legal frameworks of traditional structured finance (SPVs, UCC liens, escrow agents) with the transparency and programmability of DeFi to form a unique hybrid on-chain/off-chain architecture. CHIP serves as a governance token, conferring on holders voting rights over core protocol parameters and the ability to participate in risk backstops through staking, but it does not directly share in protocol revenue. The project has received investment from multiple top-tier institutions and established substantive partnerships with PayPal, Chainlink, and others. The protocol currently has real loan operations and revenue generation, which is relatively rare in the RWA and DeFi lending sectors. However, factors including GPU hardware depreciation risk, borrower concentration, liquidity mismatch, token unlock pressure, and regulatory uncertainty remain core variables requiring ongoing attention.

Trading link: https://www.hotcoin.com/en_US/trade/exchange/?tradeCode=chip_usdt

About Us

Hotcoin Research, the core research and investment arm of Hotcoin Exchange, is dedicated to turning professional crypto analysis into actionable strategies. Our three-pillar framework—trend analysis, value discovery, and real-time tracking—combines deep research, multi-angle project evaluation, and continuous market monitoring.

Through our Weekly Insights and In-depth Research Reports, we break down market dynamics and spotlight emerging opportunities. With Hotcoin Selects—our exclusive dual-screening process powered by both AI and human expertise—we help identify high-potential assets while minimizing trial-and-error costs.

We also engage with the community through weekly livestreams, decoding market hot topics, and forecasting key trends. Our goal is to empower investors of all levels to navigate cycles with confidence and capture long-term value on Web3.

Risk Disclaimer

The cryptocurrency market is highly volatile, and all investments carry inherent risks. We strongly encourage investors to stay informed, assess risks thoroughly, and follow strict risk management practices to protect their assets.