UCS Satellite Database: 7,560 Operating Satellites, 28-Plus Fields Apiece

It is the spreadsheet census of the operating satellite population - compiled by the Union of Concerned Scientists from publicly available sources, first published in December 2005, and current through 1 May 2023 at 7,560 operating satellites. One row per spacecraft, 28-plus columns each: who owns and operates it, why it flies, where it flies, what launched it and who built it.

The numbers worth repeating slowly. The fleet splits 6,768 in low Earth orbit against 590 in geostationary slots, and by operator country the US leads with 5,184 records, ahead of China at 628 and Russia at 181 - three very different space programs sitting in one identical row shape, which is precisely what makes cross-country comparisons honest.

Inside Datadory's catalog - 1,744 datasets averaging 7.81 - the cable & satellite slice holds 12 records, and this one scores 8/10: field definitions verified against the published layout, coverage global across every operator country, and a fixed cutoff that makes it the citable baseline of the shelf. Get a sample of this dataset before anything else moves.

One row, one satellite. The anatomy below is a real record from the file - the Hera Systems 1HOPSAT-TD imaging cubesat:

Name of Satellite, Alternate Names : 1HOPSAT-TD (1st-generation High Optical Performance Satellite)
Current Official Name             : 1HOPSAT-TD
Country/Org of UN Registry        : NR                # Niue registry
Country of Operator/Owner         : USA
Operator/Owner                    : Hera Systems
Users                             : Commercial        # Civil | Commercial | Government | Military
Purpose                           : Earth Observation
Detailed Purpose                  : Infrared Imaging
Class of Orbit                    : LEO               # LEO | MEO | GEO | Elliptical
Type of Orbit                     : Non-Polar Inclined
Perigee (km)                      : 566
Apogee (km)                       : 576
Eccentricity                      : 0.00151
Inclination (degrees)             : 36.9
Period (minutes)                  : 96.08
Launch Mass (kg.)                 : 22
Power (watts)                     : 0.5
Date of Launch                    : 2019-12-11
Expected Lifetime (yrs.)          : 0.5
Contractor / Country              : Hera Systems / USA
Launch Site / Vehicle             : Satish Dhawan Space Centre / PSLV
COSPAR Number                     : 2019-089H
NORAD Number                      : 44859

Read the anatomy, not just the digits. A 22-kilogram spacecraft with a stated half-year design life and 0.5 watts of power tells you exactly what class of mission this is; Users separating Commercial from Government settles funding questions in one column; and the paired identifiers close the loop - 2019-089H (the international COSPAR designator) and 44859 (the NORAD catalog number) join this row to element sets and tracking feeds without touching the name string. The Niue UN registry on a US-operated satellite is not an error; it is how smallsat regulatory practice works, and the registry column preserves it faithfully. Rows for your operators, orbits and launch windows arrive with your sample.

Twenty-seven core columns form the frame documented in the table below, all verified against the published record layout. They group into four stacks. Identity: both name variants, the UN registry entry, operator/owner and country, and the twin catalog keys - COSPAR and NORAD numbers - that make joins mechanical. Mission: the Users enum splitting civil from commercial from government/military, plus purpose and detailed purpose down to distinctions like infrared imaging inside Earth observation. Geometry: orbit class and subtype, perigee, apogee, eccentricity, inclination, period, and the GEO slot longitude that populates only above the belt. Hardware: launch and dry mass, electrical power, launch date, expected lifetime, contractor and country, launch site and vehicle.

Two quirks deserve respect when parsing. Dry mass is sparsely populated because manufacturers publish it inconsistently - treat blanks as missing, not zero. And the provenance columns riding beside each record are folded under additional fields on request, documented against delivered extracts rather than asserted blind. Ask for the sample and the complete column list arrives with it.

Geography - global, every operator country represented: the US at 5,184 records, China at 628, Russia at 181, and everyone else down to single-spacecraft operators, all sharing one column layout so cross-country work needs no harmonization pass.

Temporal - one static snapshot current through 1 May 2023. That fixity is the feature: holidays, launch rushes and constellations-in-progress are captured once, consistently, in one unchanged schema - a historical archive rather than a rolling feed.

Granularity - one row per operating satellite, 7,560 of them, keyed by name and joined to the wider world by COSPAR and NORAD numbers.

The shape serves research directly: modelers get 7,560 fully attributed training rows with no assembly step, and analysts get a baseline whose values cannot shift underneath a citation. Anything after May 2023 belongs to a living companion - which is exactly how the pairing works best.

API, files, or your warehouse. Daily, weekly, or hourly.

You choose the scope - all 7,560 rows whole, or cut to your countries, operators, orbit classes or launch windows - and the landing zone: flat extracts for analysis, a feed for running pipelines, or tables landed directly in Snowflake, BigQuery or Redshift. Because the census is fixed, most teams take it once, complete, and reserve recurring cadence for whatever live series runs beside it. Every delivery ships the field dictionary unchanged and validation rows alongside, so your first join happens against evidence rather than optimism. Get a sample of this dataset scoped to your cut before anything ships.

• Space-policy researchers fixing a baseline - one citable census of who operates what, by country, user class and orbit, immune to catalog churn mid-study.
• Analysts measuring drift - diffing the frozen 7,560 rows against living directories quantifies launches, deorbits and re-registrations since May 2023 without hand-collected launch logs.
• Industrial-policy teams reading supply chains - the contractor and contractor-country columns make national manufacturing footprints countable.
• Engineers sizing orbit regimes - 6,768 LEO against 590 GEO frames capacity, congestion and slot-economics questions instantly.
• Modelers forecasting replacements - launch date, design lifetime and mass support aging-fleet curves per operator.
• Data engineers joining catalogs - NORAD and COSPAR keys link this census to element sets and position feeds deterministically.

Journalists, academics and students hold the top relevance score here - a single stable citation answering "how many satellites does each country operate, and doing what" outlasts every redesign of a live portal. Data scientists and ML engineers score next: 7,560 clean, labeled rows across orbit, mass, power and purpose are rare raw material for fleet clustering and for benchmarking features engineered from live feeds.

At relevance 2: competitive intelligence and product teams mapping rival operator portfolios and contractor relationships at the baseline, and market researchers and consultants sizing orbital markets by regime and user class. Investors and quants round out the pack using fleet composition as context input for space-sector theses.

Within orbital data this record owns the attributed fleet census: owner, purpose, mass, power and contractor on every one of 7,560 satellites - columns no tracking feed carries. The N2YO Live Real-Time Satellite Tracking Directory holds the opposite bargain: 35,103 objects positioned right now, but thin on ownership and mission attributes. SatBeams overlays EIRP footprint maps on more than 500 GEO satellites for link-budget work, and the channel family - KingOfSat at 25,839 channels across 162 countries, SatIndex scanning 17 European positions - describes what the machines broadcast rather than the machines themselves.

They stack rather than compete: this as the deep attributed baseline, N2YO or CelesTrak element sets for everything that has moved since. Where the slice ranks sits on our best cable-satellite datasets list, with the full pooled view on the cable-satellite data hub.

Three things, stated plainly. First, the census is frozen at 1 May 2023 - that is deliberate, not neglect, and it is why citations against it hold. Pair it with a living directory when your question extends past the cutoff; successive deliveries under a Datadory cadence accumulate into exactly that panel.

Second, sparse columns are sparse in the source too. Dry mass publishes inconsistently across manufacturers, and power readings mix end-of-life and beginning-of-life conventions - weight those fields accordingly, and let the validation rows in your sample show you the fill rates for your cut before you build on them.

Third, registry surprises are real data. A satellite operated from one country may be UN-registered elsewhere, and smallsat practice makes this common; the registry column preserves the official relationship rather than flattening it to the operator's flag. Name your countries, operators and orbit classes, and the sample comes back shaped to them. Get a sample of this dataset before anything is scheduled.