The market prices compute credit every day, but there is no public, free, machine-readable reference for it. SpreadBench turns FINRA TRACE secondary trades into the NCSX index family — continuous, tiered and recomputable value by value (Composite / HY / IG).
Every point is all constituents of that day weighted by face value outstanding (bp over the Treasury benchmark). Hover for daily values and changes.
“Secondary bond spreads × aggregated by Neocloud sector × free and public × machine-readable” — nobody occupies the intersection of all four.
| Neighbour | What it does | Why it cannot fill this seat |
|---|---|---|
AMComputeGPU Financing Tracker |
110 primary-market financings, terms and pricing tranches | It is an issuance-side data vendor — a free daily secondary-spread benchmark would erode the freshness premium of its paid report (incentive mismatch) |
Bloomberg / FactSetTerminals |
Everything, in a terminal | $20k+ a year; the long tail (private-credit analysts, lender risk teams, researchers, press, agents) cannot get in, and terminal data cannot be freely redistributed |
FINRA TRACERaw data |
Raw trade prints (public, delayed) | It is the raw material, not an index — no constituent screen, no tiers, no continuous series. Turning raw material into a benchmark is index-provider work |
BondTerminal et al.Single-bond tools |
Single-bond price / YTW / spread calculator (CRWV tranches already covered) | Spread and history sit behind a login; single-bond lookups, no sector aggregation, not a free public index — but it proves constituent TRACE activity is sufficient for daily updates |
FINRA-Bloomberg indicesActive bond indices |
Market-wide IG / HY active-bond indices, EOD, free | Broad market only, no Neocloud sleeve; its existence actually validates that a public TRACE-based index is a compliant, precedented form |
Index change = the sum of each constituent's spread change weighted by face value. Sorted by absolute contribution; the sum equals the day's index change exactly — attribution and level share one source, not two sets of numbers.
| Bond | Rating | Tier | Weight % | Spread bp | Δ 1D | Contrib bp |
|---|
Which is exactly why an aggregated, continuous, machine-readable public benchmark is missing.
Weekly levels, change attribution, outlier bonds, sector premium and structural snapshot. Subscribe by RSS — no account, no email required.
—
The common questions, answered once. These answers are also emitted as FAQPage structured data for search engines and LLMs to quote directly.
SpreadBench is a credit-spread benchmark for AI-infrastructure (Neocloud) bonds. It aggregates public secondary-market trades in Neocloud issuer bonds into a continuous, tiered, recomputable credit spread.
SpreadBench is the brand and product name; NCSX is the index ticker (Neocloud Credit Spread Index), used for the three tiers NCSX-Composite / NCSX-HY / NCSX-IG. Cite as “NCSX-Composite (provided by SpreadBench)” with the methodology version and the dataset DOI (10.5281/zenodo.23931330, archived on Zenodo).
Single-bond spreads come from public FINRA TRACE secondary trades (face-value weighted yield to worst); the benchmark is the US Treasury yield (currently a single-rate snapshot, tenor-matched curve scheduled for v0.2.0). No quotes, no ratings, no primary-market pricing.
T+1 aggregate levels and the constituent list are free forever, no registration required. Data is licensed CC BY 4.0 — free to use and redistribute with attribution and the methodology version. Full history and attribution detail sit in the subscription tier; we never sell raw trade records.
Target cadence: index computed at 17:30 ET each trading day and published T+1; the AI Infrastructure Credit Weekly publishes every Monday covering the prior five trading days. We are currently in the seed-series phase — a frozen snapshot with no daily updates; live T+1 daily updates begin with licensed TRACE data (v0.2.0). The homepage status card states this plainly.
Recommended: SpreadBench (2026). Neocloud Credit Spread Index (NCSX), v0.1.1. https://spreadbench.com/. The citations page offers media, Markdown and agent-JSON citation cards (values update daily); the methodology page generates APA / BibTeX / MLA in one click.
NCSX aggregates public secondary trades and performs factual arithmetic (wording limited to “spreads computed from public trades”). It issues no credit ratings and no buy or sell opinions. That is the product's hard boundary.
A rating is an agency's judgement about future ability to pay; NCSX is a price outcome the market has already transacted at. One is an opinion, the other an observation that traces back to specific trades.
One dataset: people read the page, programs read the JSON, agents call the MCP server. Methodology is fully public, so anyone can recompute the same number from their own trade data.