FLOPS · Onboarding (Finance)

Overview For finance teams For engineering teams

Settlement-grade compute prices, audit-ready

FLOPS publishes independent reference prices for GPU compute the same way Platts publishes oil and ICE publishes power. Every value pins to a methodology version, carries a cryptographic recompute receipt, and meets IOSCO Principle 7 (k-anonymity ≥ 3). Suitable for mark-to-market, covenant true-ups, and contract pricing.

What you get — the seven guarantees
SPECWhat it means
Methodology pinEvery published value carries a methodology version (e.g. flci-h100@v0.9). Methodology changes never silently rewrite history — the new version pins at the change moment.
k-anonymity ≥ 3Settlement-grade values reflect at least 3 independent source quotes. Below-3 ticks exist in the inspection table but never appear on the published surface.
Recompute receiptEach value emits an SHA-256 observations_hash + computation_proof. Verifier is one request away — GET /v1/audit/recompute/{run_id}.
MAD trimMedian Absolute Deviation outlier suppression on the source vector before the weighted median. Single bad source can't move the published value.
Source weighting5-tier authority weighting (Tier 1 first-party APIs → Tier 5 scraped aggregators). Weights documented in the methodology version.
FX pinningNon-USD sources convert via the daily Frankfurter ECB reference rate, pinned to the publication date.
Tier disciplinePublic surface is LIVE-only. SEED-mixed indices live in the partner-tier catalog and carry a clear tier chip. No silent reference fallbacks.
1 Excel / Google Sheets — formula in a cell 2 min

The fastest path for treasury, FP&A, and risk teams. Six formula functions cover lookup, time-series, methodology citation, and the audit verifier link.

Install

# Excel for Mac / Windows
Insert → Get Add-ins → Manage → Upload My Add-in →
  https://app.flopsindex.com/excel-addin/manifest.xml

# Google Sheets
Extensions → Add-ons → Manage Add-ons →
  https://app.flopsindex.com/sheets-addin/manifest.json

The five cells finance teams use most

=FLOPS.PRICE("FLOPS-H100-OD")
→ 6.155                            current settlement value

=FLOPS.PRICE("FLOPS-H100-OD", "2026-05-15")
→ 6.213                            historical settlement

=FLOPS.METHODOLOGY("FLOPS-H100-OD")
→ "flops-h100-od@v0.9 · k-anon ≥ 3 · MAD trim"

=FLOPS.RECOMPUTE("FLOPS-H100-OD", "2026-05-15")
→ "https://app.flopsindex.com/v1/audit/recompute/{run_id}"

=FLOPS.GPU_CAPEX("h100_sxm5")
→ 26000                            acquisition-layer reference price
=FLOPS.GPU_CAPEX("h100_sxm5", "tier")
→ "LIVE_USASPENDING"                federal procurement (real new)
=FLOPS.GPU_CAPEX("h100_sxm5", "seed_price")
→ 28500                            audit fallback (quarterly seed)
The killer cell: drop =FLOPS.RECOMPUTE() next to a settlement value in your mark-to-market model. The IOSCO audit trail follows the row — your auditor opens the URL, sees the source vector, the trim, the weighted median, the published value.
For acquisition modeling: =FLOPS.GPU_CAPEX("h100_sxm5") returns the highest-authority reference price from a 3-tier cascade — T2 federal procurement contract median (real new prices), T1 eBay completed-listings × 1.15 markup, or T3 quarterly seed. The tier field tells you which one fired. Methodology at /methodology/GPU_CAPEX_LIVE_METHODOLOGY.md.
For your auditor: =FLOPS.GPU_CAPEX("h100_sxm5", "n_obs") returns the observation count behind the LIVE primary; GET /v1/refdata/gpu-capex/h100_sxm5/observations returns the raw rows (federal-contract award IDs + agency + amount, or eBay item IDs + sold-date + condition) that fed the median. Drill from the published value to the underlying records — IOSCO Principle 17 auditable inputs, end-to-end.
2 Direct cite — methodology surface for contracts 3 min

For ISDA-style or bilateral contracts where the settlement reference must be named and version-pinned in the docs. Both the slug and the methodology version are addressable URLs that don't change.

Contract clause template

Settlement Reference: FLOPS-H100-OD
Methodology Version: flci-h100@v0.9
Source: https://app.flopsindex.com/v1/methodology/flci-h100/v0.9
Publisher: FLOPS Compute Intelligence
Receipt verifier: https://app.flopsindex.com/v1/audit/recompute/{run_id}

Permalink surface (citation-stable)

  • /i/{INDEX_ID} — schema.org Dataset JSON-LD, suitable for footnoting
  • /v1/methodology/{slug}/v{N} — immutable per published version
  • /v1/audit/recompute/{run_id} — receipt + source vector + math
3 REST — for treasury automation 5 min

Auth-free public endpoints serve LIVE indices for prototyping. Partner-tier X-FLOPS-Api-Key unlocks the full catalog, history, spreads, and the recompute substrate.

Quick start

curl https://app.flopsindex.com/v1/price/FLOPS-H100-OD
{"value": 6.155, "tier": "LIVE", "confidence": "HIGH",
 "unit": "USD/GPU-hr", "as_of": "2026-05-18T13:00:00Z",
 "methodology_version": "flops-h100-od@v0.9"}

The full set of finance-relevant endpoints: /v1/price/{slug}, /v1/ts/{slug}?range=30d, /v1/derived/spread, /v1/derived/compute-margin, /v1/audit/recompute/{run_id}. See the engineering onboarding page for full schemas.

? Common questions — finance
Is FLOPS suitable for mark-to-market and covenant true-ups?
Yes — the SETTLEMENT family is designed for that use. 24h-window publication, methodology-pinned, k-anon ≥ 3, MAD-trimmed. Most pilots run mark-to-market weekly; the upstream feeds publish hourly so the underlying observability is finer than the contractual covenant.
What's the difference between SETTLEMENT and LIVE INTRADAY?
SETTLEMENT is a 24h-window publication for mark-to-market and contract pricing. LIVE INTRADAY is a 1-hour rolling reference for intraday workflows and agents. Both use the same math — different observation windows. The F4 dashboard panel shows both side-by-side; sustained spreads >10% indicate the settlement reference is lagging.
How does an auditor verify a published settlement value?
Three artifacts: (1) the methodology version URL pins the math, (2) the observations_hash identifies the source vector, (3) GET /v1/audit/recompute/{run_id} replays the computation from canonical inputs and returns a green/yellow/red status. The Recompute Audit panel on F13 GOVRN shows the daily run.
What does "k-anonymity ≥ 3" mean in practice?
Every settlement-grade published value reflects at least 3 independent source quotes. Below-3 ticks exist in the partner-tier inspection table but never appear on the public surface. This protects contributor confidentiality (no single source can be back-derived from the published value) and aligns with IOSCO Principle 7.
Is there a published change log for methodology?
Yes. Each methodology has an immutable version history at /v1/methodology/{id}/versions. Changes never rewrite history — the new version pins at the change moment and old published values keep their old version pin.
How do I get a paid pilot started?
Email team@flopsindex.com with your use case (lender residual-value, derivative settlement, transfer pricing, capex IRR, etc.). Pilots are typically 60-90 days with weekly published indices under contract; we co-design the slug set with you in week 1.

Ready to put it in your model?

Email us Read the methodology Browse the catalog