David Choi & Conor Moore: USD.AI - Financing the Future of AI Infra

David Choi and Connor Moore from USDAI discuss an innovative financing model involving a synthetic dollar backed by AI and deep hardware assets. David emphasizes the integration of traditional finance with decentralized finance (DeFi) principles, while Connor focuses on cash flow modeling and the significance of hardware quality in generating cash flows for USDAI's lending primitives.

USDAI evolved from its original concept as Metastreet, which faced challenges in collateral pairings for loans in the real-world asset (RWA) space. The company identified D-PIN networks as a suitable fit for financing illiquid assets. The Tactical Compute Initiative aimed to create an AI mining network, but financing challenges led to the development of USDAI, which offers yield-bearing loans collateralized by hardware assets like GPUs and telecom infrastructure.

The market for USDAI targets RWAs and addresses the "D-PIN trilemma" of balancing scale, price, and data fidelity. The underwriting process emphasizes hardware quality, focusing on assets valued above $10,000. Unique challenges include a game-theory approach to defaults and the creation of an Oracle-free DeFi structure.

USDAI's framework is built on three pillars: CALYBER, which tokenizes physical assets; a modular underwriting system for global scalability; and QEV Design, which enhances the overall structure. Connor introduces Calibre, a system that ensures on-chain property rights through NFTs, allowing borrowers to maintain operational rights while legal ownership remains with Permian Labs.

The underwriting process supports various asset types, including GPUs and EV chargers, featuring a two-tranche structure for loan origination. The first tranche aligns interests between underwriters and originators, allowing for outsized returns while managing default risks. QEV analyzes past blue chip RWA protocols that often depeg due to collateral mismanagement, creating an arbitrage venue for participants. It introduces a unique redemption queue design, transforming traditional systems into a bidding auction every 30 days.

The auction mechanism allows participants to enter blind bids, with the highest bidder receiving a pro rata amount of liquidity for redemption. This design encourages strategic bidding, while unutilized amounts are redistributed to the market, enhancing overall yield. USDAI facilitates on-chain liquidity provision to deep-end operators, generating yield from interest on loans.

Capital markets integration aims to incorporate productive assets into money markets, addressing limitations of existing platforms. The system employs robust risk management strategies, including loan-to-value (LTV) ratios of 50% to 70%, ensuring liquidity providers are protected in case of defaults. Borrowers face a first loss position, where they lose their equity cushion before impacting liquidity providers.

Loans finance new hardware acquisitions, maintaining a 30% cushion preserved as payments are made. In the event of a default, collateral can be sold at a discount, ensuring profitability. Deep-in networks benefit from reduced capital costs by accessing debt markets, allowing them to lower APR payouts significantly. Neoclouds, which rent GPUs to AI startups, can focus on software development rather than expensive venture funding due to improved access to debt markets.

The governance token structure aims to decentralize control over protocol parameters, allowing for permissionless launching of borrowing pools. Underwriting will gradually include various collateral types, starting with compute, telecom, and energy sectors. Yield dynamics will shift from high liquidity with lower yield to hardware-based loans with higher yield but reduced liquidity.

The conversation highlights the necessity of establishing protocols to scale industries such as solar panels and data centers, drawing parallels to Bitcoin's evolution. Choi and Moore emphasize the critical role of capital expenditure (capex) in attracting traditional finance, essential for creating scalable credit. They express hope that miners and emerging platforms will seek access to both compute and credit to improve resource availability on-chain.

The project is positioned as a DeFi protocol focused on generating yield from RWAs and deep-end projects, with a clear distinction made between hardware financing and traditional business loans. The discussion includes successful DeFi models and the aim to provide competitive rates and improve efficiency in the DeFi ecosystem.

Choi and Moore address the appeal of RWAs in decentralized finance for yield generation, noting adverse selection issues that arise when RWAs transition to crypto. They highlight successful RWAs such as T-bills and Bitcoin mining, which attract interest from traditional finance. New yield concepts like delta neutral purse balancing indicate a growing recognition from traditional finance.

Plans for USDAI's next steps include building a borrower pipeline and addressing demand for borrowing solutions, with a focus on emerging use cases in the GPU market. The importance of GPUs in the AI race is emphasized, alongside ongoing efforts to generate yield through data usage. A private beta launch for DeFi participants is announced, with contracts deployed and audits completed, and integration plans with Pendle and various oracles for yield generation are outlined. The anticipated public launch is set for the end of May.