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

This episode unveils USD.AI, a synthetic dollar poised to revolutionize AI and DePIN hardware financing by bridging the gap between on-chain capital and real-world, capex-intensive infrastructure.

1. Introduction to USD.AI and Team Background

• Jordan from Delphi Research hosts David Choi and Conor Moore from USD.AI.
• USD.AI is introduced as a synthetic dollar backed by AI and DePIN (Decentralized Physical Infrastructure Networks) hardware assets. DePIN refers to networks that use token incentives to build and operate real-world physical infrastructure like compute, storage, or wireless networks.
• The team's background combines traditional finance (TradFi) and deep crypto expertise:
  • David Choi: Met Conor at Deutsche Bank (real estate investment banking). Moved into the early MEV (Maximal Extractable Value) scene in 2017-2018, focusing on DeFi and redemption design for USD.AI. MEV refers to the maximum value that can be extracted from block production in excess of the standard block reward and gas fees by including, excluding, or re-ordering transactions within a block.
  • Conor Moore: After Deutsche Bank, focused on real estate private equity and structured finance. His expertise lies in physical assets, cash flows, modeling, and credit structuring for USD.AI's lending primitives.
  • A third co-founder, McCall (not present), worked at 21.co during Bitcoin's shift from GPU to ASIC mining and later at DRW on FPGAs (Field-Programmable Gate Arrays) for HFT (High-Frequency Trading). FPGAs are integrated circuits that can be configured by a customer or a designer after manufacturing. HFT is an automated trading platform used by large investment banks, hedge funds, and institutional investors that utilizes powerful computers to transact a large number of orders at extremely high speeds.
• The team came together in late 2021/early 2022, initially focusing on low-liquidity assets.

2. The Genesis of USD.AI: From Illiquid Assets to DePIN Financing

• Conor Moore explains their initial focus at MetaStreet (started in 2021, $400M loan volume) was on bringing financial innovation to illiquid assets on crypto rails, moving beyond liquid token-centric DeFi.
  • They observed that DeFi innovations like AMMs (Automated Market Makers) and CDPs (Collateralized Debt Positions) primarily catered to liquid tokens. AMMs are decentralized exchange protocols that rely on mathematical formulas to price assets, while CDPs allow users to generate stablecoins by locking up collateral.
• A key challenge was finding the right collateral pair, noting adverse selection problems in the RWA (Real-World Asset) space. RWA refers to tangible assets like real estate or intangible assets like intellectual property that are tokenized and brought onto the blockchain.
  • Conor Moore: "If someone's tokenizing real estate and trying to get a loan on chain, you know, you typically would ask why are they not getting a loan from the most robust debt market in the world, which is like real estate debt."
• The realization: DePIN networks, and their underlying hardware, are a natural fit for on-chain financing of illiquid assets.
• They started Tactical Compute, an AI mining network with Aethir and Beam, to buy GPUs and earn yields from decentralized AI networks.
• This venture highlighted a critical problem: financing the vehicle was nearly impossible despite good cash flows because on-chain earnings need on-chain capital sources. This was the inception point for USD.AI.
• USD.AI is a yield-bearing synthetic dollar where yield comes from interest paid on loans collateralized by hardware assets (GPUs, telecom, energy).
  • Strategic Implication: The difficulty in financing on-chain cash-flow generating hardware with traditional finance presents a clear market opportunity for crypto-native solutions like USD.AI.

3. Market Opportunity: Focusing on High-Value DePIN Hardware

• David Choi outlines their market focus, starting from crypto-native RWAs and narrowing to the DePIN sector, specifically targeting "good credit but underbanked" areas.
• They address the "DePIN trilemma": balancing scale, price, and data fidelity/quality. Many protocols lower price to scale, sacrificing quality.
• USD.AI focuses on sectors with expensive, productive hardware (e.g., H100s for LLMs - Large Language Models), not consumer-grade hardware. LLMs are AI models trained on vast amounts of text data to understand and generate human-like language.
  • The underwriting is primarily on the hardware itself (semi-fungible) rather than the DePIN network, though network cash flows bolster this.
  • Target assets are typically above $10k per unit due to scalability and capex intensity.
• Actionable Insight: Investors should look for DePIN projects with high-value, productive hardware, as these are more likely to generate sustainable cash flows suitable for collateralization and attract financing solutions like USD.AI.

4. USD.AI's Oracle-Free DeFi Structure: The Three Pillars

• Jordan highlights the unique problem of hardware not marking to market, leading USD.AI to create an oracle-less lending structure. This means the system doesn't rely on external price feeds (oracles) to determine collateral value, which can be manipulated or unreliable for illiquid assets.
• David Choi explains the three pillars supporting this structure:
  1. Caliber: A system for tokenizing physical assets.
  2. Modular Underwriting System: To scale underwriting for diverse hardware globally.
  3. QEV (Queue Extractable Value) Design: An exit mechanism for redemptions, inspired by MEV Boost.

5. Pillar 1: Caliber - Tokenizing Physical Assets with On-Chain Property Rights

• Conor Moore details Caliber: Collateralized Asset Ledger with Insurance Bailment Evaluation and Redemption.
• To execute a loan, the borrower sells assets to a subsidiary of Peregrine Labs (USD.AI's Delaware C-Corp).
  • This allows USD.AI to obtain a master insurance policy.
• A separate bailment contract gives the borrower rights to operate and profit from the assets. The bailor (rightful owner) is designated as the holder of an NFT.
• This NFT is an Electronic Document of Title under UCC Article 7, representing an electronically native deed to the asset.
  • Conor Moore: "What we're doing here instead is putting 100% of the property rights on chain."
• Strategic Implication: Caliber's approach aims to provide stronger, more transparent on-chain property rights compared to many RWA tokenization methods, potentially reducing counterparty risk for investors.

6. Pillar 2: Modular Underwriting System - Aligning Interests

• Conor Moore explains that USD.AI will support various hardware types (Neoclouds, GPUs, telecom, power assets, EV chargers).
• For each primary origination market, there are two tranches:
  • First Loss Position (5% of capital stack): Held by specialized underwriters/originators for that specific asset class (e.g., GPU experts, EV charger experts). They earn outsized returns but bear the first loss in case of defaults.
  • Second Tranche: Goes into USD.AI backing.
• Actionable Insight: This modular, aligned-interest underwriting model could attract specialized expertise and improve risk management, a key factor for researchers evaluating the robustness of such platforms.

7. Pillar 3: QEV (Queue Extractable Value) - Gamified Redemption Mechanism

• David Choi explains QEV (Queue Extractable Value), designed to manage redemptions and profit from de-pegs, drawing inspiration from Flashbots' MEV Boost.
• Traditional redemption queues are often first-come, first-serve or opaque. USD.AI turns redemption into a market.
• Hardware loans have high amortization (principal repayment over time), meaning significant cash is freed up monthly. Amortization is the process of gradually writing off the initial cost of an asset or repaying a loan over a period.
• Every 30 days, an auction (QEV Boost auction) occurs for this aggregated liquidity (from principal repayments).
  • Participants submit blind bids; higher bids get a larger pro-rata share of the redemption liquidity.
  • This creates an arbitrage opportunity: buy sUSD.AI (staked USD.AI) at a discount and redeem at par by bidding in the auction.
• If no one bids, the "traffic tax" (bidding amounts) is distributed as supplemental yield to all sUSD.AI holders.
  • David Choi: "What we're really trying to mimic is just understanding the perspective of effectively an arbitrageur, a stablecoin market maker, or any participant in buying these assets in the secondary but making a profit in the primary."
• Strategic Implication: QEV aims to create a more dynamic and potentially more stable redemption process by incentivizing participation and managing liquidity through market mechanisms, which is crucial for the stability of a synthetic dollar backed by illiquid assets.

8. Default Process and Loss Prevention

• Conor Moore outlines the lines of defense against defaults:
  1. Borrower Equity Cushion: Loans have 50-70% LTV (Loan-to-Value), meaning a 30-50% equity cushion provided by the borrower, who takes the first loss. LTV is a financial term used by lenders to express the ratio of a loan to the value of an asset purchased.
  2. First Loss Position: The specialized underwriter takes the next portion of loss (5% of the capital stack).
  3. Protocol Exposure: The remaining exposure is to the USD.AI protocol.
• Monthly loan amortization (around 3% for a 3-year loan) helps maintain the equity cushion as assets depreciate.
• If a borrower misses a payment, the Electronic Document of Title (the deed to the collateral) is auctioned on-chain to a network of resellers.
  • This creates a profitable default scenario, as resellers can buy assets at a discount.
• Depositors (USD.AI holders) have secondary market liquidity for their tokens and can exit via the QEV redemption process.
• Actionable Insight: The multi-layered defense system and on-chain auction of defaulted collateral are key risk mitigants for USD.AI LPs. Researchers should analyze the effectiveness and potential stress points of this model.

9. Value Proposition for Borrowers: Accessing Leverage

• Conor Moore explains that borrowers pay high interest rates (targeted 15-20%).
• The math is simple: if unlevered yield from hardware operation exceeds the interest cost, borrowing is accretive.
  • Example: 15% cost of capital with a 70% LTV loan can add ~2.5x leverage to returns (e.g., 30% unlevered yield becomes ~70-80% levered).
• Benefit for DePIN Networks: Access to debt markets allows DePIN networks to reduce inflationary token emissions used to incentivize hardware operators. If operators can get leverage, the network needs to pay out lower APRs.
• Benefit for Neoclouds: Neoclouds (companies buying GPUs to rent to AI startups) often use expensive venture capital for depreciating assets. USD.AI offers a more appropriate debt financing option.
  • Conor Moore: "The number one Neocloud will end up being the biggest buyer of assets and that's just all it takes for them to win or lose. So if you have access to debt markets then that means you can buy more assets."
• Strategic Implication: USD.AI provides a crucial capital efficiency tool for DePIN operators and AI infrastructure providers, potentially accelerating the growth of these sectors by lowering their cost of capital and reliance on inflationary token incentives or equity.

10. Phased Approach to Scaling USD.AI

• David Choi details the scaling plan:
  • Initial Phase (Launch): Deposits (USDC/USDT) will initially mint USD.AI backed by Treasuries to reduce cash drag. Staked USD.AI (sUSD.AI) becomes eligible for infrastructure loans.
  • Initial target utilization for hardware loans: 40%, eventually moving to ~80%.
  • Maturity State: Parameters like utilization targets could be controlled by decentralized governance. QEV will be live.
  • The protocol is designed to be permissionless for launching individual borrowing pools.
• Onboarding of new collateral types (compute, telecom, energy initially) will be a stepwise process as underwriting capabilities for each sector are developed.
• Actionable Insight: The phased rollout, starting with lower-risk Treasuries and gradually increasing hardware loan utilization, is a prudent approach to managing risk and testing the system. Investors should monitor the staking ratio and utilization rates as indicators of yield potential and risk exposure.

11. Governance Token and Future Utility

• David Choi discusses the potential USD.AI governance token:
  • Fee Earning: The token would earn fees from protocol operations (e.g., interest clips, insurance bond structures).
  • Liquidity Governance: Inspired by Curve, token holders could influence liquidity direction towards specific DePIN sectors they want to support, impacting underwriting, risk tolerance, and fee participation for those sectors.
• The vision is for DePIN protocols to require debt financing to scale, similar to how Bitcoin mining industrialized with access to traditional finance.
  • David Choi: "What we hope is that a lot of these different participants...will not only want access to compute but also credit, which helps scale the wide availability of these resources on chain."
• Strategic Implication: A governance token could allow investors and researchers to actively participate in shaping the growth and risk parameters of the USD.AI ecosystem, potentially creating a "flywheel" effect for specific DePIN sectors.

12. USD.AI's Place in the Broader Crypto Landscape: DeFi Protocol with RWA Yield

• David Choi emphasizes that USD.AI is fundamentally a DeFi protocol, even though its yield source is RWA/DePIN hardware.
  • He draws parallels to Ethena and Lido, highlighting the importance of vertical control over origination, structuring, and distribution.
• The objective is to distribute competitive yields to the broader DeFi ecosystem (e.g., integrations with Pendle, money markets).
  • This involves a trade-off: higher yields for potentially longer redemption periods, which the DeFi ecosystem is becoming more accustomed to.
• David distinguishes between "money markets" (focus on liquidity, yields compress quickly) and "capital markets" (ICM - Internet Capital Markets) which focus on productive, often illiquid assets like hardware.
  • USD.AI aims to bring true capital market yields (from productive hardware) into DeFi.
• Conor Moore highlights the immense scale of the target market.
  • Conor Moore: "CoreWeave did a $7 billion loan with Blackstone...at SOFR plus 11. So basically 15-ish percent cost of capital for $7 billion. So the scale in this game is so large."
  • SOFR (Secured Overnight Financing Rate) is a broad measure of the cost of borrowing cash overnight collateralized by Treasury securities.
• Actionable Insight: USD.AI aims to be a new, less cyclical source of yield for DeFi, derived from productive real-world AI/DePIN infrastructure. Its success could significantly expand the TAM for DeFi by tapping into large-scale hardware financing markets.

13. Immediate Next Steps and Launch Timeline

• David Choi shares the upcoming plans:
  • The team has been building a borrower pipeline, finding significant demand for hardware financing even in Web2 use cases due to systemic underbanking in this new, critical commodity (GPUs).
  • Private Beta: Contracts deployed, audit finished, front-end finalization underway. Expected to launch "next week" (from podcast recording date).
  • Points Program ("Cores"): For participants in the private beta and beyond, rewarding early engagement.
  • DeFi Integrations: Planned with Pendle, oracles, and modular money markets to make USD.AI yield and points farmable.
  • Public Launch: Targeted for the end of May.
• Interested parties can join the waitlist via Telegram and Twitter.
• Strategic Implication: The imminent launch and points program offer early opportunities for investors and researchers to engage with USD.AI. Monitoring initial uptake, yield generation, and DeFi integrations will be key.

Conclusion

USD.AI introduces a novel mechanism for financing AI and DePIN hardware, offering a potentially high and stable yield source for DeFi. Investors and researchers should watch its launch closely, particularly its ability to scale capital formation for productive, real-world infrastructure and integrate within the broader DeFi ecosystem.