Scrutica
Five-tier classification of how legibly each facility traces to a beneficial owner; capacity-weighted per country.
Capacity-weighted tier distribution per country, sorted by total power capacity (descending). Click a country to drill into its facility list; click a tier segment in any country's bar to filter that country's detail page to that tier.
Pick up to 4 countries. The comparison is capacity-weighted; a country's share is the sum of disclosed power capacity per tier divided by its total. Note: Data Gap differences ≥ 20 percentage points make a direct tier comparison structurally unfair — the higher-Gap country has had less of its capacity investigated.
The Compute Visibility Index is a structural measurement of how transparently a country's known compute capacity traces to a beneficial owner through public records. It is not a national-virtue metric; a country with a more public-filing-heavy chain is not more "virtuous" than one with mixed tiers — it is more transparent in this specific dimension. The data substrate is Scrutica's ownership-chain graph: facilities resolve to operating organizations, organizations resolve through parent / subsidiary / affiliate edges to ultimate beneficial owners (or to a documented terminus, if that endpoint exists).
Operator/owner is itself an SEC filer (CIK populated), OR an ancestor in the upward-edge walk has a primary-filing-grade source: SEC 10-K Item 21, EDGAR XBRL, FactSet Revere or FactSet Workstation, WRDS-Compustat, GLEIF LEI, BIS Entity List, Federal Register designation.
Example: A facility operated by Microsoft Azure traces directly to Microsoft Corporation, an SEC filer (CIK 0000789019). A subsidiary of NVIDIA documented through a 10-K Item 21 disclosure: same tier — the chain terminates in a public filing.
At least one ancestor exists, but it is reached only through a PitchBook (private-database) source. No path to a Tier_1 source exists.
Example: A privately-held neocloud whose parent holding company appears in PitchBook's corporate-structure graph but does not file with the SEC and is not on any export-control or LEI registry.
At least one ancestor is documented, but not via a Tier_1 or Tier_2 source. In practice this bucket is rare in the current substrate (most documented ancestors come from FactSet or PitchBook).
Example: An ancestor record imported from a research-database source that is neither primary-filing-grade nor PitchBook-sourced. Reserved so future ingestions don't mis-classify.
The operator/owner is named in Scrutica's records, but the upward-walk yields zero documented ancestors. We have looked, and the chain is opaque.
Example: A national datacenter operator whose corporate structure is private and undisclosed in any database Scrutica integrates. The facility is attributed; the controlling structure beyond the named operator is not.
The facility has no operator_org_id and no owner_org_id linked. Scrutica has not yet investigated this facility's ownership; this is structurally different from Tier_4.
Example: A peering-database record imported with the building name and country but no resolved operator. Investigation queued; tier classification pending.
For each facility, the classifier starts at COALESCE(operator_org_id, owner_org_id) and walks upward through the ownership_upward_edgesview. At each hop, the algorithm picks the highest-authority parent whose edge meets confidence and source-priority thresholds; the walk terminates on (a) an SEC-filer node, (b) a government-entity node, (c) a primary-filing-grade ancestor edge, (d) a cycle, or (e) the six-hop bound. Each tier classification carries a chain that consumers can audit — the country detail pages let any reader expand a facility's row to see the per-hop edge sources, confidence ratings, and as-of dates that produced the classification.
Tier distributions are reported in capacity-weighted form. A country's tier-1 percentage is the sum of power_capacity_mwacross tier-1-classified facilities divided by the country's total documented capacity, not the count of tier-1 facilities divided by the count of all facilities. The capacity-weighted view tracks where governance leverage actually sits: a single 1 GW facility at tier 1 carries more weight than fifty 1 MW facilities at tier 4. The country detail pages also surface the facility-count distribution for completeness.
These categories are intentionally separate. Tier 4 means the operator is named in Scrutica's records but the upward-walk found no documented ancestors — we looked and the chain is opaque. Data Gapmeans the facility itself has no operator or owner organization linked; investigation has not yet reached it. Comparing two countries — one at 5 % Data Gap and another at 40 % — without surfacing this distinction is structurally unfair to both, because Tier 4 is a finding about the records and Data Gap is a finding about Scrutica's coverage. The country comparison view flags Data Gap disparities explicitly; the dashboard shows the gap percentage in the rightmost column for that reason.
power_capacity_mw is null on a facility, the row contributes zero to the capacity total. The facility-count distribution (also surfaced) acts as the complementary view that does not depend on capacity disclosure.countryfield cannot be normalized (e.g., legacy ingest pollution like "Northern Virginia" or "Unknown") appear in a separate "unrecognized" bucket. The dashboard flags this so the recognized-country totals are not silently inflated.