> For the complete documentation index, see [llms.txt](https://docs.usd.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.usd.ai/usdai/risks-and-mitigants.md).

# Risks & Mitigants

How USD.AI manages risk in GPU-backed lending: LTV limits, insurance, collateral standards, and borrower underwriting.

USD.AI is a balance between risk mitigation in the pursuit of yield capture. By orienting risk mitigation towards asset productivity, USD.AI will facilitate growth by focusing on the asset underwriting process.

<table><thead><tr><th width="128.79296875">Risk Category</th><th>Mitigant</th><th>Details</th></tr></thead><tbody><tr><td>ECONOMIC</td><td>Equity Buffer</td><td>70-80% Loan-to-Value (LTV) ratio on each loan, with blended portfolio LTV capped at 80%, provides a safety margin against asset value fluctuations and default risks.</td></tr><tr><td>ECONOMIC</td><td>Debt Service Reserve Account</td><td>~1-3 months of peak debt service held in a restricted cash account (or in USDai directly) to cover potential ramp up delays, cash flow lumpiness of borrower. Structural protection to ensure borrower maintains healthy cash reserves.</td></tr><tr><td>ECONOMIC</td><td>Amortization</td><td>Regular principal repayment schedule reduces exposure over time, with first successful implementation complete.</td></tr><tr><td>ECONOMIC</td><td>Queue Extractable Value</td><td>A market-based pricing mechanism for liquidity bottlenecks for each amortization loan repayment.</td></tr><tr><td>ECONOMIC</td><td>Residual Value Protection</td><td>Required on any loan where the offtaker is not investment-grade or the compute is sold as on-demand rental. Where a deal includes a residual value warranty on the servers, the provider covers part of the shortfall if the servers sell below an agreed floor, subject to the warranty’s conditions (including suspension during a market-wide shock). It applies to a sale of the servers, not to a sale of the SPV as a whole. In a default, the Agent first seeks to keep the SPV running under a replacement operator or to sell the SPV whole at a price that repays lenders in full; the servers are sold individually through ITAD firms only if that fails.</td></tr><tr><td>ECONOMIC</td><td>Portfolio Limits</td><td>New loans are originated within portfolio-level limits: a maximum blended portfolio LTV of 80%, a minimum quarterly portfolio DSCR of 1.15x, and a maximum sUSDai utilization of 90% for deploying new loans.</td></tr><tr><td>LEGAL</td><td>Bankruptcy Remote SPV</td><td>Clear legal structure secures asset ownership and collateral rights throughout the financing period, assets and contractual revenue split from parent corporate balance sheet prevents other creditors from attempting to claim rights to assets in event of default.</td></tr><tr><td>LEGAL</td><td>Onboarding Diligence</td><td>Thorough vetting of counterparties' financial status and compliance requirements before participation.</td></tr><tr><td>LEGAL</td><td>Software Oversight</td><td>Ongoing software oversight into machine usage, uptime and location.</td></tr><tr><td>MODULAR</td><td>Underwriting Process</td><td>Each loan must reach a minimum Compute IQ underwriting score of 60%. ComputeIQ assesses the operator (track record and financial strength), the offtaker (credit metrics and operating history) and the offtake contract (tenor versus debt term, coverage, termination protection and payment unconditionality).</td></tr><tr><td>MODULAR</td><td>Risk Incentive</td><td>Interest rates range based on cash flow qualities (ie, on demand rentals charged higher rate than contractual multi year offtake agreements). Higher yield compensates underwriter economic interests, which align with platform success.</td></tr><tr><td>MODULAR</td><td>Controls</td><td>Built-in safeguards include a debt service reserve; a DSCR test that, if breached, stops distributions to the borrower and uses excess cash to prepay the loan; a payment waterfall that sends all cash to lenders once an Event of Default exists; and the most recent server valuation as the reference point for judging any sale.</td></tr><tr><td>MODULAR</td><td>Approved Collateral</td><td>NVIDIA enterprise-grade GPUs (B200, B300, GB200, GB300, VR200) and AMD enterprise-grade GPUs (MI455X, MI430X). Non-core inference chips (e.g., Cerebras, Etched) are approved only alongside contracted offtake and a passing Compute IQ score.</td></tr><tr><td>TOKENOMIC</td><td>Governance Whitelist</td><td>Community oversight of loan terms and asset selection promotes long-term sustainability.</td></tr></tbody></table>

{% hint style="info" %}
Verify current backing in [View Proof of Reserves](https://app.usd.ai/reserves). Review smart contract reports in [Security Audits](/technical-overview/audits.md).
{% endhint %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.usd.ai/usdai/risks-and-mitigants.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.