# Comparing the price of AI debt

The evidence file records dated price metrics, not comparable all-in borrowing costs. The cases and model are a separate hypothetical experiment; they do not fill missing IREN, Lambda or Volta contract terms.

## Files

- `debt-price-evidence.json`: sources, price definitions and unknown actual-cost inputs.
- `debt-price-cases.json`: explicit hypothetical loan and ticking-fee inputs.
- `debt-price.mjs`: standalone calculator with no external dependencies.
- `debt-price-results.json`: generated cash flows, rates and fee amounts.
- `verify-debt-price.mjs`: reproducibility, cash-flow and boundary checks.
- `README.md`: this guide.

Save all six files together. With Node.js:

```text
node debt-price.mjs
node verify-debt-price.mjs
```

After changing assumptions, `node debt-price.mjs --write` replaces only the local generated results file. The verifier contains checks for the original published examples; changing those examples may intentionally make those checks fail.

## What the rates measure

One full draw occurs at year zero. Coupons and principal are paid at the end of equal one-year periods. The annual effective rate solves initial net cash equal to the discounted future payments. Principal determines weighted average life; coupon cash alone does not measure total financing cost.

The investor pays face times the issue-price percentage. In these examples the borrower bears the discount and pays any separately shown third-party fee at year zero. That fee reduces borrower cash without changing the modeled investor's cash flows. This attribution is an assumption, not a finding about a real syndication.

No default, prepayment, refinancing, tax, reserve, guarantee price or hedge is modeled. The stipulated benchmark sensitivities are not forecasts. Unknown rates or fees are not treated as zero. If only the third-party fee is unknown, the investor yield and payment schedule remain known while the borrower's net proceeds and rate remain unknown. A missing rate path still preserves known proceeds and principal-weighted life.

The ticking-fee calculation is separate. It assumes a constant undrawn base, borrower payment, a non-retroactive holiday and the specified day-count basis. It does not establish Volta's actual fee bill or silently add pre-draw fees to the loan IRR.

## 中文说明

六份文件分开真实交易的报价口径与独立假设实验。真实借款人成本未知时保留null，不用模型假设补齐合同。

算例假设期初一次性放款，随后按等长一年期支付利息和本金；借款人承担折价，另列费用在期初交给第三方。年化现金成本由现金流折现等式计算，不是风险调整收益率。提前还本降低资金占用，也会改变折价的年化影响。若只有第三方费用未知，仍保留可知的投资者收益率与还款表，只将借款人净款和成本留为未知。

未提款费另行计算，假设固定基数、由借款人支付、宽限期不追溯计费及指定日计数。它不是Volta实际费用，也没有未经说明并入贷款年化成本。

检查覆盖已声明的算例、未知值和输入边界，不证明实际发行人资金成本或信用支持的因果价值。