# Sorso View Compute Index, Issue 003, Calculation Report v1.0

Issue: SVCI Issue 003. Generated 2026-06-24. Institutional record of derivation. This report is derived from the sealed data layer and historical inputs named below; it is not a source of truth, the data layer is.

## Inputs (named and hashed)

- Issue 003 data layer: sorso-issue-003-data-v1-3.json, sha256 5374f050ec9b947430e7e1e476cb191efe35ea9e909c1a50c751d69b44b6740c
- Issue 001 official capture log: svci_issue001_official_capture_log_2026_w23_v16.md, sha256 79551ca89b7055a234b54050f8f24287da515ec31bbf60543027efebb58610c8
- Issue 002 capture data log: sorso-issue-002-capture-data-log-v1-1.xlsx, sha256 587dcc68ae80ad33a75a93ef4acdd18b343c12ed4212cd9b2413ccd48341e81e

All figures below are recomputed in Python from these inputs. Historical anchors are read from the sealed Issue 001 and Issue 002 records, not retyped from memory.

## Headline result

- SVCI Issue 003: 103.8030 (public display 103.8)
- Issue 002 level: 100.0
- Weekly change: +3.8030 points, displayed +3.8 percent

The index movement reflects coverage and composition change, principally Voltage Park no longer publishing a comparable fixed rate, not a broad increase in matched-provider GPU prices. The publication must not state that GPU prices rose 3.8 percent. See the matched-panel sensitivity below.

## Tier medians (from Issue 003 priced rows)

| Tier | N | Median (USD per GPU-hour) |
| --- | --- | --- |
| T1 | 4 | 10.53065 |
| T2 | 9 | 3.85000 |
| T3 | 2 | 4.62000 |

## TierIndex and SVCI derivation

Locked methodology: current-issue tier median divided by the sealed Issue 001 base tier median, times 100. No rebasing. No fixed-panel substitution. Weights T1 0.45, T2 0.45, T3 0.10.

| Tier | Issue 001 base median | Issue 003 median | TierIndex | Weight |
| --- | --- | --- | --- | --- |
| T1 | 10.5306 | 10.53065 | 100.0005 | 0.45 |
| T2 | 3.5500 | 3.85000 | 108.4507 | 0.45 |
| T3 | 4.6200 | 4.62000 | 100.0000 | 0.10 |

SVCI Issue 003 = 0.45 times 100.0005 + 0.45 times 108.4507 + 0.10 times 100.0000 = 103.8030

## Matched-panel sensitivity (noncanonical)

This check isolates whether the index move came from price changes among providers present in both issues, or from a change in which providers publish a comparable rate. It does not replace the canonical SVCI.

Matched Tier 2 set, eligible in both Issue 001 and Issue 003: E005, E006, E007, E008, E009, E010, E011, E020, E022 (nine providers). The only Tier 2 provider eligible in Issue 001 but not Issue 003 is E012 Voltage Park, which held a 1.99 fixed rate at launch and no longer publishes a comparable fixed rate.

| Metric | Issue 001 | Issue 003 |
| --- | --- | --- |
| Matched T2 median (N=9) | 3.8500 | 3.8500 |

Matched T2 change: 0.0 percent. Matched-panel sensitivity SVCI: 100.0 (internal 100.0002). The full Issue 001 T2 median of 3.5500 was held down by Voltage Park at 1.99; removing that single non-publishing provider lifts the median to 3.85, identical across both issues. Together AI fell from 5.49 to 4.79 but remained above the median and did not move it.

Finding: the canonical rise to 103.8 is a coverage and composition event. Among providers that published a comparable fixed rate in both issues, the Tier 2 median did not move.

## Coverage and Compute Opacity Index

Locked eligible-for-denominator counts: T1 4, T2 18, T3 5.

| Tier | Included | Eligible | Coverage | Weighted |
| --- | --- | --- | --- | --- |
| T1 | 4 | 4 | 1.0000 | 0.4500 |
| T2 | 9 | 18 | 0.5000 | 0.2250 |
| T3 | 2 | 5 | 0.4000 | 0.0400 |

Composite Coverage: 0.7150 (Issue 002 was 0.7400)
Compute Opacity Index: 28.5 (Issue 002 was 26.0)

The flat non-entry share 12/27 = 0.4444 is audit-only and is not the Compute Opacity Index.

## Companion Dollar Basket (secondary metric, not the index)

Companion Dollar Basket = 0.45 times 10.53065 + 0.45 times 3.85000 + 0.10 times 4.62000 = 6.9332925
Issue 002 basket was 6.798315. Change +1.99 percent.

The Companion Dollar Basket is a secondary dollar metric. It is not the SVCI. The SVCI is the base-100 index level 103.8.

## Panel summary

- 27 providers captured. 15 published a comparable fixed public rate that cleared the methodology. 12 were captured but did not clear the headline inclusion gate (floors, marketplaces, aggregators, sales-gated, signed-in only, or product-mode incompatible).
- 0 pending.

## Locked public interpretation

The Sorso View Compute Index rose to 103.8 from 100.0. The movement came from Tier 2, where the number of providers publishing a comparable fixed rate fell from ten to nine. Among providers eligible in both the launch issue and Issue 003, the Tier 2 median was unchanged at 3.85. The change reflects declining public-price coverage rather than a broad increase across comparable providers.